航空宇宙購買の最適化：Lokadソリューションのガイド付きデモ

Back to Lokad TV

00:00:00 航空宇宙購買：不確実性と経済的優先順位
00:05:58 フリートの可視性、サービス目標、投資コスト
00:11:47 ロータブル、消耗品、経済性で順位付けされた投資
00:17:23 予算と運用制約下での優先順位
00:23:56 1ドルあたりのサービス改善で投資を順位付けする
00:29:43 データ検証と不確実な需要のモデル化
00:35:48 需要とターンアラウンドタイムの不確実性を組み合わせる
00:41:42 システム統合と売却機会
00:47:15 交換、修理優先順位、サプライヤー統合
00:53:29 データ準備と導入要件
00:58:44 履歴シミュレーションと意思決定のトレーサビリティ
01:04:10 経済的優先順位、消耗品、サプライヤー遅延
01:09:14 整備予測と推奨遵守の追跡
01:14:05 購買自動化、調達プラットフォーム、修理優先順位
01:19:52 主要な学びと締めくくり

要約

Conor DohertyとFabian Hoehnerは、Lokadが不確実な需要、修理時間、限られた予算を、順位付けされた航空宇宙購買の意思決定へ変換する方法を示します。このアプローチは、追加の1単位がサービス改善、またはaircraft-on-groundの回避にどれだけ寄与するかを、コストとの関係で評価します。同じ考え方は、売却、修理の優先順位付け、在庫移動にも適用されます。Lokadは既存システムと並行して動作し、supply chain scientistsと運用専門家が計算を洗練させます。履歴シミュレーションは過去の意思決定を説明する助けとなり、自動化の度合いは対象となる購買の価値と複雑さに依存します。

詳細要約

航空宇宙の購買では、希少な資源をどこに使うかという競合する選択が生じます。追加のスペアを購入すれば航空機を地上に留めるリスクを下げられるかもしれませんが、その資金は別のより重要な業務を守るためにも使えます。このライブデモで、Conor DohertyとFabian Hoehnerは、Lokadがそうした選択をどのように評価するかを説明します。中心となる問いは、1ドルでどれだけ追加のサービス、またはAOG回避を合理的に購入できるかです。在庫数量は、この比較と、事業が直面する制約から導かれます。

デモは、例示的な航空会社のフリートと、修理して再び使用できる部品であるrotablesのプールから始まります。可用性は需要だけでなく、修理と物流プロセスに費やされる時間にも左右されます。5個を所有していることは、すぐに利用できる5個のserviceableな部品を持つことと同じ保護を意味しません。Lokadは既存システムから取引データを取得して運用状況を把握し、そのうえで投資推奨を計算します。例では、約6800万ドルの在庫が95.7%の期待サービスレベルをもたらし、約120万ドルの追加投資で98%近くまで引き上げます。それ以上の改善はますます高くつきます。

推奨は1単位ずつ順位付けされます。特定のPNの追加1単位は、価格や需要パターンが大きく異なる他のPNの追加単位と競合します。通常、1個目は2個目や3個目より多くの不確実性をカバーします。この逓減する効果は重要です。同じ部品を複数購入するより、投資を分散した方が価値が高い場合があります。重要度も考慮されます。比較的安価なno-go部品は、不在時の影響が小さい高価な部品より高い可用性を正当化することがあります。順位付けされたリストは、予算制限、購買能力、受入制約の適用も容易にします。

こうした比較には、不確実性の現実的な扱いが必要です。航空宇宙の需要はまばらで不規則なことが多く、ターンアラウンドタイムも大きく変動します。平均値は、航空会社がまさに備えたい例外的な状況を隠してしまいます。そのためLokadは、需要と補充時間の分布をモデル化し、それらを組み合わせて将来の複数の可能な期間における需要を推定します。Hoehnerは、作業が順調な場合は修理が短い期間に集中し、問題が起きるとずっと長い期間に集中することがあると説明します。このパターンを単一の平均で置き換えると、必要在庫を誤って表す可能性があります。既知の整備イベント、確率的な部品表、変化するターンアラウンドタイムの傾向も計算に反映できます。

財務的な推定は、結果として生じるトレードオフを議論可能にします。AOGイベントのコストを確定するのは難しい場合がありますが、概算があればチームは意思決定を比較し、仮定を見直す土台を得られます。運用知識は依然として不可欠です。シートカバーやコーヒーマシンでも、技術的な耐空性分類だけでは捉えきれない影響を持つことがあります。ユーザーはitem inspectorや補助ダッシュボードで推奨を確認し、意外な結果に異議を唱え、モデルの改善を助けます。高価なrotable購買では通常、専門家の検証とサプライヤー見積もりが引き続き必要です。一方で、低価値の消耗品やexpendablesは自動実行の余地が大きくなります。

同じ考え方は、売却、修理、配分にも広がります。売却では、解放される現金と失われるサービス保護を比較します。修理優先順位付けでは、どの非稼働ユニットが最も緊急に支出や迅速化を正当化するかを特定します。移動では、供給元拠点で増えるリスクと輸送コストを、別拠点での便益や追加購入という代替案と比較します。これらは関連する経済的意思決定ですが、それぞれ固有の運用詳細を必要とします。たとえば理論上の売却機会も、買い手を見つけ、現実的な市場価値を設定できるかに依存します。

Lokadは、取引システムの横で動くインテリジェンスのシステムとして機能します。Lokadのsupply chain scientistsは、顧客のデータとプロセスに合わせてソリューションを調整します。サプライヤーが新しいプラットフォームを採用する必要はありません。有用な意思決定は、どのデータ欠陥に注意を向けるべきかを明らかにし、最初にすべてをきれいにしようとする終わりのない作業を避ける助けにもなります。Hoehnerは、導入期間の典型例を約6か月としつつ、大きなばらつきがあると述べています。過去の実行結果は調査可能な形で残り、チームは予算制約、誤った仮定、受け入れたリスクを区別できます。より広い規律は継続的です。不確実性を定量化し、結果を比較し、証拠が蓄積するにつれて意思決定を改善していきます。

Full Transcript

Conor Doherty: Welcome to the first ever Lokad live demo. Today we will show you how Lokad generates investment and divestment decisions in aerospace. Today we will show you which PN should take priority.

Today we will show you how a limited budget should be divided, and we will show you how your constraints, be they supplier lead times, turnaround times, costs, budget itself, how those influence the decisions that we generate for you. Now, this is a live event, so please get your questions in. We will answer them probably in about 30 minutes. Depends how long we banter.

But who are we? I’m Conor, marketing director here at Lokad. And joining me in studio, my very good friend and director of strategic accounts, Fabian Hoehner. Guten Tag, Herr Hoehner, wie geht’s?

Fabian Hoehner: Hello, Conor.

Conor Doherty: Das ist wunderbar. So, Fabi, as people can see in the corner of the screen, or should be visible in the corner of the screen already, we will soon look at a purchasing, an aerospace demo account. We will focus on purchasing, but some people who are in attendance and who will watch later are actually not that familiar with Lokad. Some are, but many are not.

So, let’s take a minute before we get into that. Could you answer two quick questions? One, how exactly does Lokad see the problem of purchasing in aerospace? And two, if people only watch the first few minutes, what are a couple of key concepts they should take away?

Fabian Hoehner: You didn’t ask me that in the prep, so this is—

Conor Doherty: No, I didn’t. I wanted improvisation.

Fabian Hoehner: Yeah. Okay. So, I mean, there’ll be a mixture of people being familiar or not familiar. So we try to keep it high-level but go a little bit deep on some areas. So the two things that I would like people to take away on the one hand, how we handle uncertainty.

Uncertainty being in the topics that we talk about, mainly on supplier uncertainty being turnaround times, lead times, and on the other hand, demand uncertainty. And then the second aspect being economic optimization, economic prioritization. So with a limited available resource, how do I do the best, basically? How do I purchase efficiently in that case? So these are the two main areas that I would like people to take away from today, and yeah.

Conor Doherty: So, in a sentence, for every single euro or dollar or whatever currency that you’re operating in, for every, let’s just say euro, for every euro that you invest in your inventory, what is the expected service level gain, essentially?

Fabian Hoehner: Yeah. So, to be more precise, what you want to do, and we’re going to go into detail here, is maximize the service level gain per dollar or the—reduce the AOG, so aircraft on ground, per dollar that you’re spending. So the most efficient route towards your goal, and we hope that this is going to be the takeaway for people listening today.

Conor Doherty: Cool. All right. Now that we’ve set the table, let’s let’s not bury the lede any longer. Let’s go straight to the demo account.

Max, our producer, feel free to just switch, make sure that everything’s okay. Subcto. All right. So, Fabi I’m looking at my own screen because my eyesight is terrible, but what exactly are we looking at?

Please tell people so they understand: what is the context here? Is this aviation? Is it MRO? Is this AI? Are we doing AI?

Fabian Hoehner: Yeah, we always do AI, of course. we go into, into details what what we do in terms of AI, but we are on, indeed, on a demo account. Lokad in 20 seconds: we design custom inventory optimization solutions for our clients from different areas and today we are going to focus on aeronautics. Could be MRO, so maintenance, repair and overhaul, but we are going to focus on example from basically an airline.

So the main question that we’re going to ask ourselves is: for a given fleet, how do I get the most efficient investment, potentially divestment, route for my stock? That’s the overarching question. Very quickly, so where we are right here. So you can see we are on go.lokad.com.

This is a demo account. So all of our clients have—are on the same platform. So it’s a multi-tenant application, but we write an individual solution for every client. This is what’s making demos in a way challenging because every client is very different.

Conor Doherty: Different. Yeah, of course.

Fabian Hoehner: So this is, I would say, our USP, is to customize what we’re doing. So the focus on today is going to show, the two principles that I alluded to in the beginning. On the one hand, how do we see uncertainty?

The key word is going to be probabilistic forecasting. For the people that are following us a little bit they have heard it many, many times, but today we’re going to go very practical. And then the second: economic optimization for one specific example. So if there are questions, please, of course, ask them in between.

Oftentimes my answer will be: “yes, if it’s logic, we can build it, or we have built it in the past”, but we’re just looking at one specific example today. So that’s what I propose to go into. Any further questions? Otherwise, we’re just going to get rolling.

Conor Doherty: I would say let’s just get rolling because there actually are a lot of questions already. As people know, I’ve actually communicated with a lot of people in advance of this. So, there are some very concrete questions to get to later, but as I said, if something stands out to anybody watching, feel free to comment below and we will unpack that in due time.

Fabian Hoehner: Yeah. And you do know me, so interrupt me. Otherwise, I

Conor Doherty: I will, don’t worry.

Fabian Hoehner: I talk myself into a little flow. all right. So, this is again a demo account, and we can see this is just an overview. So I could click on anything I wanted, but we are going to dive into just two, three screens today that we are going to look at.

So, in that case, we are looking at the this is an overview dashboard. We are going to do inverted pyramid. We’re going to start with the overview and then going to drill down into the very lowest level. So this here would be a management overview where we can see the pool performance.

So, in that case, you can see it’s a pool. So it could be it could be from several airlines. The bottom line, the—what we are looking at is one fleet being Triple 7. Whether that’s from one or several airlines doesn’t matter, and the parts that I need to service that fleet.

So this is going to be the overarching concept here. In that case you can see a historic cons—display, and, of course, it’s going up because Lokad is optimizing, so things are always getting better. And some, some additional views on the current operational situation. So where are—where’s my stock?

In aeronautics we are always talking about loops. This is going to be a very important concept. Obviously, most people watching today are very familiar with that. But stock isn’t equal stock in aeronautics.

It’s about having serviceable stock and unserviceable stock, knowing where it is in the process. So, having five units, having five serviceable units, having four units in a repair cycle, one unit serviceable this is a very different interpretation of your reality. So, this here just basically an overview dashboard. How would we have gotten there? Would be by integrating—

Conor Doherty: Literally about to ask you that exact question, opening my mouth to ask it.

Fabian Hoehner: Yeah. So, typically, in that case, the data, where’s it coming from? From different ERP systems, MRP systems, but bottom line, transactional data that we take onto the platform, Lokad being a platform of intelligence. So transactional system, ERP, MRP, just “where is my stuff”, just the transactional level, we take that on a daily basis, and then we do the, I want to say, the big data manipulation to get the insight of “I have 186 units that are right now in a return process.” Good, or did you—

Conor Doherty: No, it was good.

Fabian Hoehner: Perfect. So then, high-level overview, and now we are going to directly dive into, actually, the result. So we’re going to start with what could the result of an optimization look like, and in that case, what we’re looking at here is the investment optimization.

So we are looking at a perfect or an investment that we want to get for, in this case, a target service level of 98%. We can see that here’s a little simulator with a drop-down, so I can simulate all different scenarios that I’ve already pre-calculated, in this case to make it quick. We could also design that differently so that I can have a little fill-in field and try around myself, but in that case, we have pre-calculated to make it quicker. And what we can see here is that, for a given situation, so we can still see our current stock, our current stock being, that case, 68 million, and these 68 million are expected to give me a service level of 95.7%.

Conor Doherty: Mhm.

Fabian Hoehner: 95.7% availability relative to the individual lead times of all the different parts and weighted by their consumption.

Conor Doherty: Okay.

Fabian Hoehner: So, next thing that we’re doing we select, in that case let’s go to our 98 here. We select, “okay, we want to have 98% average availability weighted by consumption and so on.” And in that case, the simulator here is telling me, “okay, I need to invest 1.2—2 million”, and I’m going to reach my target, in that case, 97.99. That’s what I would have gotten.

So that’s the high level, highest level. I could play around with that, and so if we were to play around, we’re going to see something very typical in aeronautics, which is, and I’m sure that the majority of the audience is very familiar with that, the long tail is where the costs lie. Yeah. So if we are seeing, so I went from 98 or 98.5 to 99.5.

Let’s do—go to 99. So we have the 1%. What we can see is the first percentage point that we wanted to gain or the first two points cost us a million dollars in investment and then to get the next one is going to be 2.5. So this exponential increase to gain these service level points, and that is extremely important when we talk about, in the beginning, what is the technology?

Why is it important? What are we doing differently? The big fit for Lokad in this industry is the understanding on how to deal with sparse and erratic demand which is the majority of aeronautics problems. So that means how do I better understand the likelihood of the extremes, so these 90-plus percentiles because that is basically where aeronautics plays. aeronautics being an extremely risk-averse sector being out of stock costs a lot of money.

Conor Doherty: Yeah. Much more than the individual part in many cases.

Fabian Hoehner: So, traditionally, aeronautics, everyone is overstocked. Yeah. “In doubt, just buy another one.” That’s the typical approach.

So understanding how likely a 99 or 99.5 scenario is extremely important on the individual part. We’re going to get to that in on a more detailed level. Just high level. Why are we looking at all the those extremes here?

Because that’s where we play in aeronautics. Nobody wants to know what the average is. That would mean that you have enough stock in 50% of the cases. You don’t care about that.

You want to know how likely it is to cover extremes. So, when looking at that, so the—again, we’re going from high level to lower level. Here we have the investment and now this is my total investment, and we want to know, okay, what are the what are the quantities typically for the client? So, in that case, we have, very straightforward, a list of: “these are my items and those are the quantities that we are investing into.”

Conor Doherty: So, just to clarify, at this point, we’re looking at rotables or consumables right now? This is just rotables.

Fabian Hoehner: Yeah, rotables.

Conor Doherty: Yeah, absolutely. But we’ll look at consumables later, or we’ll touch on the logic there.

Fabian Hoehner: Yeah, we’ll touch on the logic, but

Conor Doherty: But these are the most expensive parts, basically. That’s why we’re looking at them.

Fabian Hoehner: You saw, basically the values that we’re seeing here are going to be quite important because yeah, we look indeed at rotables. The majority of the concepts are transferable to consumables. The main difference the rotable part, per definition, has a rotation, so has a loop, instead of several repair loops.

In theory, you can do that internally, or you have a scrap. And a consumable you just buy, and it goes out. So there there are some differences in the math but the underlying principles that I was referring to in the beginning, being uncertainty, or quantifying uncertainty, and prioritization based on economics are going to stay the same.

Conor Doherty: Okay.

Fabian Hoehner: So the question now is why do we recommend four, three, five, and so on and so on? For that, we’re going to go a little bit more into detail, but first, actually stay somewhat high level. What we can see here is, again it’s just the same list that we looked at before, but with some additional information. And the first thing that we intuitively want to see—so I’m going to show you quite a few screens with lots of numbers.

I’m going to you and the audience what things are that you should be looking at. So in that case, this is my list and these are the parts that I’m recommending to buy. You can see four units here, three there, five and so on and so on. Now, the important thing is what the optimization should just intuitively be doing.

I’ll explain how we get there. We are going to go deeper and deeper. But first, does it intuitively make sense? Intuitively, what should an optimization that goes after efficiency?

If efficiency is AOG reduction by dollar spent or service level increase per dollar spent, what we should be seeing is that parts that are, A, relatively cheap and, B, highly important should be overweighted in the availability relative to parts that are expensive and not important. So if we’re looking at this list, what we should be seeing is that let’s look at the first two parts because they have the pretty similar unit price. And what we can see is that one part is a lot higher. The final service level—current service level is what I’m covering with the current stock that I own.

Final service level is where I’m getting to after the optimization. And, in that case, we can see this one we bring to 94% and this one we bring to 80%. Why? If we look at it, this we as an essentiality: no-go. And this one is a go-if.

So, bottom line, the consequence of an out of stock here is a lot more expensive. So this is why it’s relatively overstocked, and so that’s— should be the intuition. If we go down the list, what we should be seeing again. So a part that here we see 99. Actually you can tell me if it’s too small, right?

Conor Doherty: That’s better. Thank you. Yeah, at least for me, because I’m almost blind.

Fabian Hoehner: Yeah. So here we can see it’s a 99.5. What we can see is the part is relatively cheap at $8,000 and it’s a no-go part. So that intuition should stick all along: that parts that we are aggressively stocking relative to others, should be no-go and fairly cheap, while the ones that we are less aggressively stocking should be— here we can see it’s a go part that is not really expensive, but it’s only go.

So, and that’s the—sorry, that’s the underlying logic first, that should stick. Now, the question is: how do we get to, exactly four, three, five, and so on? So, intuitively, it makes sense. Great.

Second step is now how do we get to exactly that? So this is where for the first time—so this should, high level, we already see there is an concept of economic prioritization hidden in there. Now, how do we get there in more detail? And this is what we call a ranked prioritization. So a ranked investment list in that case.

And what we are doing here is we are looking at every single purchase, or in that case, investment possibility that we have, and we simulate what would be the economic consequence of investing into that part. And what you can see, if you look at the first two, three lines, you can actually see that it’s the same PN, but it’s not the same stock-keeping unit because we are buying the first part, and then the second, and then the third, and this is basically what we’re doing all over. We simulate purchase of every single individual part we could be buying. And so this list here is basically endless. So I could scroll down endlessly, and that’s, that’s what we do.

Conor Doherty: It’s limited by budget, presumably, because I see that it’s ranked again. The total investment is increasing as you scroll down. So, presumably, you could set a hard limit, like “I only have X amount of—my budget is X. It’s not infinite. Therefore, optimize up to this point and then no further.”

Fabian Hoehner: Yeah. In that case, it is not limited by budget, but it’s limited by my objective being, in that case, 98%. But you’re absolutely right. So basically the 98% results in a budget of 1.2, 1.25 million. So what we’re going to do is actually exactly that.

We’re just going to scroll down here until we hit a total investment of 1.25. I mean, in this case, we did it differently. We try to reach 98 and then 1.2 was the result. But I could have done the inverse. I could have said—

Conor Doherty: Yeah, of course.

Fabian Hoehner: It’s the same, actually just looking at it from a different perspective. Okay, cool. Now let’s go through a few examples to just show, basically, how does this economic prioritization logic work? And the a good litmus test also for members in the audience that, that are running their own optimization is always: are you able to say, if I tell you, “okay, you have $100,000, which would be the single most important item to buy today?”

And that is a very important concept. It’s not going to be for this first part, but generally to prioritize every action that you’re taking because, with that logic you can introduce whatever constraints you have, be it “my purchasing team can only handle five parts a day”, “my warehouse can only receive 10 a day”, whatever the constraint may be, or “I have a million in budget”. If you don’t have a prioritized view, it’s very hard if you have a just a yes, no. So let’s say “my objective is to be at parts A at 99%, parts B at 95, and parts C at 91.”

Extreme, but okay. In that case, it’s just yes and no, and you don’t have a ranking system, and that is incredibly important here, this concept, because what we can do is indeed prioritize, and you always need to: limited time, budget, whatever it may be. So— again, so the question for the audience would be: “can you pinpoint which is the single most important item I need today?” And again, don’t care about the one, but what are the 50 most important ones?

So, and that is indeed what we can do here, and we’re going to do so by looking at the first line. What we can see, the—we are looking at the 1338. So, we’ll see we’ll look at the 1338 for quite some time. So, bear with me here.

We see that currently we don’t have any in stock. We are looking into buying one. We see how many were requested over the last year. and then we are buying one unit which cost us $1,200, and currently we have zero service level or expected service level.

Why? We have zero in stock. So, if you have zero you’re not going to cover any uncertainty. That’s basically what that means.

My expected service level means: how much uncertainty over the future am I covering? What that means exactly, we are going to get to that in the next step. As I said, we go from high level to low first. You could just stop at: “these are the items you should be buying”, and that’s it. But, obviously, we want to go a little bit deeper and understand where this is coming from.

Conor Doherty: So, just to jump in, because that is a good point because one of the questions I received multiple iterations. Please have some drink. I received multiple iterations of the same question that like, “interested in the idea, sounds great, because some people are already familiar with Lokad, but what does it look like for, let’s say, my planning team?” Because, so you’ve just said, “look at these dashboards, here are the ranked decisions.”

Then you said “we’re going to get into much deeper analysis” you said you could stop here. So, essentially your planning team could stop at this point. You’ve already got the decisions. If you trust the system, you can execute. If you want to learn more, you can investigate why this unit, this PN, is above this one, etc. Etc. Etc.

Fabian Hoehner: Yeah.

Conor Doherty: Okay.

Fabian Hoehner: So, indeed, I mean, we could have, like, hoping that we have set up the system. We can go into how long it takes but let’s just say after 6 months, you have a system that is working and running and so on. You could just say, “I trust the system”, and we—the recommendations get exported and executed by the operational systems.

Again, we’re not a transactional system. We are system of intelligence. So we are there to run complicated simulations and then push the intelligence back to the operational system.

Conor Doherty: Oh, please go ahead.

Fabian Hoehner: I just said it. Having said that,

Conor Doherty: yes,

Fabian Hoehner: There’s—there would be very few people that if we remember the parts that we looked at before, a part at $46,000, you’re not going to just run that automatically. And this is not—

Conor Doherty: Yeah.

Fabian Hoehner: And this is not how it works. This is—these are parts where you actually need to make a quote. So, in the, in aeronautics, this is not like Amazon e-commerce.

No, you would you would request you would request the price for the part, and the price that we have here is an assumption until it’s validated. So this is a manual process, and in—or can be automated, but my point being: the rotable parts that we’re looking at here, that’s probably not something that you would fully automate. You could of course, but indeed, so—

Conor Doherty: the execution of the decision is not necessarily, but the analysis and the actual generation of the decision is automated.

Fabian Hoehner: Your question is basically: “what are people looking at?” I would say typically if it’s not worth the time, so if the parts are, I know, below $2,000 or so if we were looking at C&E, then you can and we do indeed just automate everything. So it’s updating the operational systems on a daily basis and just automatically pushing an order. If we talk about rotable investments, that’s typically something where you have experts looking at to validate lots of things.

And in that case, we do actually something very similar to what we’re doing here. So we have the recommendation, tell you “buy five units of that.” And then you have an expert who has been doing that, who is challenging that result, and in the beginning challenging it to build and design and improve the system with us and then also to investigate and actually the investigation process, the reasoning, how I get to my “buy five units”, that’s exactly what I’m doing here, what I’m showing you. How did I get to that? So we go back to if you hadn’t interrupted me, it would already be done, but yeah.

Conor Doherty: My apologies.

Fabian Hoehner: We—so we look at the 1338, and we said: “buying one unit brings us up to 47%.” So it’s covering 47% of the uncertainty. The delta from 0 to 47 is 47.

And then we put that into relation to the entire catalog. So what’s my delta on the, on the catalog? So that depends on what’s the, what’s the consumption of that part. So, obviously, a part that is, has a higher consumption will have a higher impact here and so on.

Then my gain on the overall service level, the AOG reduction in dollars. How do we get to that? So again, we custom-build this and I’m not expecting many of the people in the audience and even our clients when we start with them to have a fixed number and say, “oh, I know what an AOG costs”, but our approach is to always say “it’s better to be approximately right than exactly wrong.” So, in that case, that means, yes, it’s very hard to pinpoint: what does an AOG cost us?

So, typically what you want to see here is in that case, we could have looked at how many AOGs did I have that were event related and and where stock event was the cause, and what’s my high-level estimate of the cost. So basically we can also see I had to make an emergency purchase, something like that. What is the value of that? In this way, I can try to derive a value that is approximately right.

Again, the argument: better to have a number in dollars than to have none. Because if you don’t have one, then you’re always just somewhat flying blind with service levels where you can’t really argue “should we be at 99 or at 99.5?” I don’t know. But the moment that you put dollars on something, you can argue and you can also argue across department.

Conor Doherty: Yes.

Fabian Hoehner: And you can also change that over time. That’s fine. If one year later you say, “okay, I think our estimates are wrong here, or on this fleet type, it is, it is right, but on this fleet type, it should be more expensive.” Doesn’t matter. But the moment that you quantify things, you can actually—

Conor Doherty: You give a common language for people to discuss differences of opinion.

Fabian Hoehner: Exactly. Perfect. So, and that brings us to in the end, it’s the AOG reduction per dollar spent, or you can say the service level gain per dollar spent. It’s both going to lead us to a score which is essentially where do I have my most efficient investment route? So now I’m going to take you just to the second and third line and then we are going to move on.

So the second line being, we can see, actually, the same part. So it’s the same ID, but we’re not buying the same part. This time we buy the second unit. The first one we already bought, so we can only invest now in a second unit. And what we’re going to see is that this one is going to bring us to an—can always tell me to zoom in.

Conor Doherty: No, it’s fine. It’s fine.

Fabian Hoehner: Crouching there. So brings us up to 72% in that case, meaning we get 25% added service level. And the first one, obviously, covers the most uncertainty. What does— 50%, and the second only 25.

So what’s the logic consequence? The second part, and that’s kind of obvious, has less value to our operation. Therefore, we are going to see that all the values decrease. So my service level increase, my AOG cost is lower, and therefore my ranking score is lower.

Okay, that’s not the revolution to say, “the second part is less valuable than the first.” I can’t buy, obviously, the second before the first. So, but now we look at the third line. Same logic, exact—everything is the same.

I only gained 14% for that part. I still pay 1,200. So, my ranking score is going to get lower. The interesting part is now the fourth line where we can see just a different part number.

It’s just everything is different about that. I have a different lead time request. What you can see in here, you have a different TAT underlying. You have a different lead time potentially underlying.

The price is different. Everything is different on that part. However, what remains the same is my ranking logic, which is going to be what is my AOG reduction? We can also talk over that over my investment.

And that’s the only the core concept that we’re going to apply here. I’m making a little parenthesis. Of course, we go more complicated in reality and actually also in these calculations, we do add some factors on no-go, if, and so on. In that case, it’s included in the AOG cost.

So, to make it simple, the AOG cost is higher if you have a no-go part. That can be extremely complicated. So we have then, and this is not because we are geniuses, This is the feedback that we get from our client that, basically you look at an output, and we call this experimental optimization. You show a list to someone and say, “hey, this is what I would be doing in your place”, and then you talk to the experts and the experts are telling, “yeah, okay, five sounds reasonable, but I would have gotten 20.”

In that case, there’s something that I’m missing because the operational experts, they typically know what they’re doing. So, in that case, could be we’re looking at a seat cover. You could say “it’s not a no-go part technically. I mean, you can fly with it.”

But the operations will tell you, “yeah, but it looks awful”, and actually the cost of, you know, someone getting into a beautiful plane and seeing, like, a—“we have to put some red ribbon on there. We can’t—”Yeah. That’s a no-go. So that is extremely expensive.

Fun fact: coffee machines, no-go. So you can’t have, like, you cannot not have coffee machines. Technically, yeah, the plane flies without them, but—so these are the kind of things where we learn together and then adapt the recipe and it may also be different from airline to airline. I mean, what are priorities?

What aren’t? So what we’re doing here, we rank everything by the ranking score being: what is the service level gain per dollar spent, the AOG avoidance per dollar spent, and just go down this list. And you can see it’s steadily decreasing, decreasing, increasing. And so every part, so we can see a part that costs $46,000 and a part that costs $5,000 become comparable because the question is just okay, at what point does it make sense to buy the sixth item of a cheap one or the second one of an expensive one?

And this is exactly what we’re doing here. So, next question is, okay, how do we get to these values, or actually, looking at an operational level? You asked before, what would the operations look at? they would be looking at a, at a list like that.

That would be absolutely reasonable. But then they would also jump into what we are calling here is an item inspector. So, in that case, and that is also for us. So this is now just, in the end, a KPI dashboard, but that is explaining how we got to the values that we looked at before.

So, typically again, so that could be the clients that investigating that ourselves, but also supply chain scientists from our end. So the people that are coding and sparring partners. So, I mean, I would guess that most people are familiar with the concept of supply chain scientists. If you, if you’re joining this lecture, then you’ve seen something from Lokad before. But the bottom line, the, the people that are implementing and challenging back and forth with the client.

So we use the same evaluation screens as the client would potentially do to then judge or evaluate: is that a reasonable decision that we’re giving? So, always drilling down from top: “buy five”, to why. So first, now we looked at an economic ranking, and second, now we look at all the details. So if we are disagreeing, typically we would be going here and seeing: do we have the same just view at reality? Because, especially in the beginning of a project, the majority of cases why we are misaligned are—do you have an idea?

Conor Doherty: The economic value of decisions?

Fabian Hoehner: No, data. It’s always data. So, basically, that you’re not aligned with the reality that you’re looking at because I mean our complicated clients and aircraft— aeronautics companies typically have made lots of purchases. So seeing six different ERP systems with legacy is something fairly usual, and three Excels left and right.

So getting just the right or same interpretation of the same reality is not easy. So just the first step of, okay, do we have the same look at reality? Do we see the same units that are in that case, they’re all—let’s just look at another part, whether we see one. Yeah, do we have the same amount of units that are in a return process in a repair process, in the logistics process right now?

Yes, no, maybe? That’s pretty important to see. And then let’s say, the data is in—indeed coherent. Then we look at, okay, what is our demand projection?

How do we get to that? that is where—so we talked about economic prioritization. That was in the very beginning, I said there are two main concepts that I want people to go away with: economic prioritization. So we covered some of that. And I said uncertainty, probabilistic forecasting, and this is what we’re going to look at now.

So— and again, for some people, this is going to be pretty evident, but I’m going to go a little bit slow just to get everyone along. So what we’re seeing here is a consumption history that is extremely typical. So, by the way, we stay with our favorite 1338. Okay. So just we—

Conor Doherty: the same part on this journey. Okay.

Fabian Hoehner: We’re looking—we’re going through the journey of this part, indeed. What we can see is this is a very typical consumption history which is sparse and erratic. So, nothing, one, one, one, two, then, two, one. What that means it is extremely difficult to forecast.

And if you were looking at that, and I know in aeronautics very few people would be doing that, but if you were looking at that from an average consumption perspective to just take a rolling rolling average, that’s what it would look like. Does that help you at all? In that case, it tells you 0.17 units. Okay, great.

So, what does that give me? Pretty much nothing. The view that you want to have, and what we’re going to do, is if it’s a probabilistic view, which is going to tell us: what is the likelihood of a consumption over a given horizon? So, when I say over a given horizon, let’s start simple. So, I mean, and again, for everyone that had statistics,

Conor Doherty: Assume I’m an idiot. You can talk down to me. That’s fine.

Fabian Hoehner: This is going to be a difficult assumption to make. So that histogram here it’s just a let’s just start simple and say it would be or it was a display of reality. So, just past, just looking past. In reality, of course, we’re looking forward, we’re forecasting, we’re doing super intelligent stuff, but simplicity, let’s just say we look at the past.

So, what is a histogram doing here? We’re just saying I don’t know whether—yeah, I think one can see here the—there’s, there’s little bars here. Yeah, I think that’s with the—and so let’s just say we assume that a period that we’re looking at is between those bars. Let’s say 30 days, and I’m now randomly saying, okay, if I was let’s say the past represents the future, and we are randomly picking out a spot on that.

In that case what that would mean is, okay, how often do I hit one demand in a 30-day window? How often do I hit a zero? How often do I hit a two, a three, and so on? And this is what this says.

So it says, randomly, we are going to hit, in 25% of the cases, we’re going to hit zero. In 34% of the cases, we’re going to hit 1, 2, 3, 4, 5, and so on and so on. Now, that is the theoretic case, and we can see here, so, basically, if I had this interval here, that would be a demand of three. If the interval was bigger it would be all of those.

Now, that’s the simple theory. The practice is, and this is where we become, I said in the beginning, appreciate, embrace uncertainty, where we get into, what’s the horizon to forecast over? Because, yeah, it’s not 30 days, because, 30 days would mean a fixed number, and we’re a lot insisting on: uncertainty exists, and you can’t plan it out of the picture. I mean, it’s nice to assume that it’s always 30 days to have a deterministic value to make your planning easy, but that doesn’t mean that it’s more real. So the reality is, sometimes it takes 30 days, sometimes 50, sometimes 60, sometimes 180, or

Conor Doherty: it doesn’t come at all.

Fabian Hoehner: Exactly, a scrap. So what we are looking at here are rotable parts. Rotable parts are normally repaired. So what we are looking at are not lead times, but typically turnaround times.

So how long does it take for my part to become serviceable again? I get my part, plane comes in, unserviceable part goes out. I replace it with a serviceable part that I had on stock, and the unserviceable part now goes through a long eternal loop. So what I want to do is ideally, I have the data to pinpoint every station of that and have lots of small distributions, but the bottom line for me is, I want to understand: what are all the possible delays that I could be facing?

And that is the horizon that I want to forecast over, because I want to understand take the extremes: if a part could be repaired within one day, you wouldn’t need a stock at all, or not— you wouldn’t need extra stock. One unit would be sufficient. So every time you have plane comes in, unserviceable out, serviceable in, unserviceable repaired, a stock of one would be sufficient. Now, reality is obviously not that, but we see that the determination of this period is absolutely essential, and that is actually what we’re doing here.

This is our, in that case it’s called replenishment lead time, can be a turnaround time, doesn’t matter. It’s— it does, but just from a principle perspective here, what we want to understand is: what is the horizon that we forecast over? We can see that this is smoothed, obviously, and we can see it’s a bimodal distribution. What could be an explanation for that? Typically, so in aeronautics, so this one here, that it’s around 80, would be everything is smooth.

You have a part, send it to your to the repair workshop, gets repaired, everything is beautiful. And this here represents the fact that something is broken that needs to be fixed that takes time. This gives you a beautiful bimodal distribution. The important message here is that it’s a very different interpretation to say that in—what are the values here? Let’s just go into that in— that in the majority of the cases, it’s, let’s say, in 80% of the cases, it’s around 80, and 20% of the cases, it’s, it’s 150 is a very different interpretation than to say it’s always the average being 110 units, so—

Conor Doherty: which very rarely happens in that distribution.

Fabian Hoehner: Yeah. Yeah, I mean, in that, I mean, obviously, this is smooth and demo data and so on. But actually that could be a very real case that you see quite frequently. So either everything goes smooth, in that case it’s 80 days, or it doesn’t, and then it’s 120.

And I’m not drawing any conclusion here. I’m not saying, “oh, choose one of the scenarios.” No, I’m just saying what you want to do is you want to understand that uncertainty exists and forecast actually over all the possible futures that exist. So in that case, I said in the beginning, “okay, we just say we pick a 30-day slot.”

30 days. In that case, it’s a super small probability of less than a percent. But bottom line, what we’re going to do is we’re going to take basically make a distribution, a demand distribution, over every possible demand horizon. And so, just imagine, I mean, whether it’s done exactly like that or not doesn’t matter, but just for the visual part, just imagine now you have a hundred different distributions over all the different horizons that exist.

Those we take and condense them into one distribution, which is that one. So what does that do? That gives us the demand over all possible futures. So why these last 10 minutes?

Because this gives us the most accurate representation of demand over an uncertain future, and that is incredibly important. While, when we talk in aeronautics about, we said in the very beginning, the only thing that I care about is the extremes. So my 90-plus scenarios, in that case, it’s extremely important to understand where I am in this distribution. And if we look at some, we’re just going to click through a few ones.

Here we go. We see this is very typical that we have these zero-inflated, right skewed distributions, which means you have a long tail that you see all the time, and then you have actually no demand over horizon is oftentimes the most likely scenario. But again, just understanding exactly how likely the outliers are, that is where the importance lie, because if that part here costs us $20,000, the question is, do we want, for these extra, I don’t know let me, for these extra, I don’t know, what is— for these extra 7%, not exactly the right calculation, but for these extra 7%, do we want to invest another $40,000 or do we just want to get the first couple ones? That’s exactly what we did before.

But this is why it’s so absolutely critical to understand what’s the shape of the distribution. That’s the first step. And then if we want to go into look into more detail, or back to our part, we remember, maybe let’s go here. So, if we were to put in one part, then we would be getting an expected service level of 47%. The second one would bring us up to 72 and up to 87, and if we—

Conor Doherty: is what was in the list.

Fabian Hoehner: Very good. You—

Conor Doherty: I was paying attention.

Fabian Hoehner: Great. So if we go here and we look back, then we are going to see, yep, we were at 37 for the first, then 72, and so on and so on, and this is basically the work back to, okay, how did we get there? And this explains how we got to exactly that number. Just breathing for a second, so if you—

Conor Doherty: I think at this point, I have a couple of questions already submitted, but I think it’s a good point now to transition to this, because one of the things that I was asked over and over again when I spoke to people in advance of this was: how does this, how does everything that people have just seen, how does that fit into a pre-existing workflow? Because, obviously, any company, any clients that we have or any potential future clients will have their own software, they’ll have their own existing workflows. So is Lokad—and do you have to rip everything out to put this in place? Does it sit alongside? What, how does that work?

Fabian Hoehner: Yeah, it’d be a pretty bad question if that was the truth.

Conor Doherty: Yeah, exactly, obviously.

Fabian Hoehner: Yeah, no, I mean we sit on top, and yeah, actually, I’m just going to show you one, I mean, literally just for a couple of seconds, I’m going to show what’s, what’s in the background of this. So this is—

Conor Doherty: that’s AI, right?

Fabian Hoehner: Yeah, this is all magic. So, obviously, this is all black magic. No, so this is, this is our programming language called Envision.

That’s, in the end, Lokad is a couple of things. It’s, on the one hand, it’s a very powerful big data platform designed for inventory optimization and indeed there are many AI features. So, for example, we can, like, write this language with internal agents and so on. But what I wanted to show here is the—you asked about the data.

We adapt it to everyone. We don’t expect anyone to pre-design anything. We just want the raw extracts, and that is what part, a big part of the work is, just to manipulate the data on our end, get it right, or get the manipulation of the data right in order to make sense of that. So data coherence.

So for example giving you an example of the fair value of a part that is, that’s a quite complex thing to achieve because in aeronautics, you you have lots of parts that have been there for a long time and you may write them off write off a part over, over eight years and then in your book value of that part will be $1. But the part is still able to to be replaced, and the new purchase is, I don’t know, $50,000. So what is the right value to assume? What’s the fair value of the part?

That’s not an easy answer. So, you have two different systems. The accounting system says it’s $1 because you leave an accounting dollar in there. And then if you want to rebuy it it’s 50,000.

How do you how do you get the data on that, right? That takes time, takes discussion, and from our perspective, especially takes flexibility. So, this is why we have a— many reasons for that, but bottom line, why you have a programming language to adapt to that. So, all our clients have very different setups and we’re just sitting on top of that.

And whether they have AMOS, whether they have TRAX, whether they have SAP, most of the times a mix of everything. Yeah, that’s just part of the system of intelligence. So we build it out. Does that answer the question?

Conor Doherty: Yeah, more or less, because the reason I asked was because I paraphrase the question. The concern was more avoiding sort of duplicate transactions, because if you, if you have multiple systems, then do I have to reconcile the—between system A and Lokad and vice versa? That’s, that was basically the underlying meaning of the question.

Fabian Hoehner: Yeah. if you ask if there’s doubling of function, so again I think if we go back a step again, you have your transactional systems that are there for transactions. So—

Conor Doherty: ERP.

Fabian Hoehner: Yeah. So ERP, M—So, basically, I take one unit out of stock and install it. That’s something you want to see in a millisecond. Everywhere needs to be updated.

That’s not Lokad. We run complex simulations that can take 20 minutes. That’s fine. I mean, these you can see it’s pre-calculated.

So it takes us just a it takes no time at all because we pre-calculated it. And this one would probably flow through in a few a few seconds, but the most complicated stuff could be taking longer, but that is fine because, that’s not our job. So our job is intelligence and provide, basically, the best insights the best decision support, the best automated decisions and then push them back to the automated systems. To get there we do have quite a few dashboards of insight dashboards.

So a system of reports. So system of records, then on top of that, typically, system of reports, Tableau, what’s the extra one? Yeah, doesn’t matter anyways. So Power BI, so these kind of things, and then system of intelligence is the class that we are doing.

In order to do that, yeah, of course. Once I have the data, we build lots of dashboards that give us an explanation. So, for example these investigation dashboards that we that we look at here, obviously this is also a system of report. This dashboard is just a report.

I mean, the calculation is happening somewhere else. But so to your question, do we double things? Yeah, you may double some things, but globally, our ambition is not to become a system of reports. That is just for us internally and also then for the for the clients to validate what we’re doing, because, yeah, we don’t want to look at only code. You want to see what it, what it does.

Conor Doherty: Okay. Speaking of what it does, when we talked right at the start, I said we would look at investment and divestment decisions. It occurred to me that we have focused a lot on investment decisions.

Could you show us something in line of, like “I’ve got too much?” Again, we were talking about rotables. “I’ve got too much stock. How can I identify units that I should probably shed to free up some capital?”

Fabian Hoehner: Yeah. As if you would have known that I have that somewhere. Brilliant.

Yeah. So, basically in this case again, the dashboard is designed however we want it. So, it’s a demo dashboard. It’s just a choice that we made here.

In that case, we have divestment and investment opportunities in the same overarching dashboard. And in that case, what we’re looking at are indeed divestment opportunities. So what’s the— in the end, divestment is pretty much the same as investment, but when we look at an investment, what we do is we take the current stock that we have and we ask ourselves: what if I was to add one more unit, and another one, and a third, and a fourth, and so on, of every stock possibility that I can? And then we build our ranking score.

What we do here is for a divestment opportunity, we look at all the stock that we own and do the same. So we ask ourselves: if I was to divest one unit of stock that I own, how much service level do I lose and how much cash am I freeing up? And that’s essentially, I mean, the same logic, and I’m not going to go through through it all, but basically, what you want to see in a list like that, you want to see items that have a high unit price, high, and you want to see a low service level loss. So, in that case, we can see yeah, we lose less than a percent on that part.

And if we look at the total loss, it’s rounded. We can’t even see it. Obviously, there are some parts that you have just too much. Then the AOG increase that you see from that, and then the ranking.

So, we should probably add a few zeros, then you could see. But if we scroll down, you can see that it’s going to increase. So, but bottom line, the feeling, what you should see here is high unit price, low service level loss, and that is what, the inverted logic. And then in theory, you can even find an ideal point between the two, and what—and then you can say, “okay, I want to invest in all those and divest until you’ve hit a sweet spot.”

That is rather theoretic, because the reality is more complex in terms of what are the fair market values in that? So, again, this is not something—this is not Amazon where you go out and say, “oh, I have 10 too much of a part that costs $85,000, gone.” No, I mean, you have to sell them. You have to find someone, location transfer and so on.

So, but that is indeed dashboards that are highly valuable for clients to divest assets and to then go out and say, “okay, yeah, this makes sense. We have at least five here that, yeah, pretty much serve only for a very, very long-tail event.” So, really, if all our fleet at the same day has an issue, that’s where this would be needed. So yeah, that’s the high level idea of of divestment.

Conor Doherty: Yep. For me, again, so when we’re just differentiating between how we would allocate budget depending on sorry, how we would generate decisions based on investment or divestment decisions and how those slightly differ. A follow-up question that was submitted ahead of time. So one person in oper— in operational procurement wanted to know: “how should, or how does, a purchasing recommendation change depending on if the unit is being acquired for exchange or for outright purchase?”

Fabian Hoehner: Depends. The answer’s always—

Conor Doherty: These do actually require quite a lot of time, so I am asking you for terse answers, so I’m aware, like, if we don’t go into all the details, we can follow up, and it’s not a problem.

Fabian Hoehner: Yeah, of course.

Conor Doherty: But, like, a general answer.

Fabian Hoehner: Yeah. So the question here would be, so in that case, we here we look at, typically at at at the pool that you’re owning. So what’s the overall stock level that I have? So this is all investments that are going to increase your overall pool level.

If it’s, if it’s just for an exchange then the pool number normally doesn’t change because you’re exchanging it and you’re giving back at a certain point of time. So there, the question would be more in evaluating, so that you could do because you have too many in a—so in theory, you have enough parts, but they’re all stuck in a repair loop. So you don’t really need more in stock, but you need to bridge a time where they’re coming back. So it would essentially be a different optimization.

So what we’re looking at here is investments that go into the pool. Yeah, the—that would also go to, into repair prioritization. So, in the end, so what, the underlying question where we would go to is if I have a given amount of parts in my repair process, which of those parts should I be prioritizing to be repaired? Sometimes you can change it but if you have every company has, if we go to the overview, we saw, actually, we have currently we have 500 units going through a repair process. Not all of them are equally important.

And what we could do here, and we do this with, basically, again, prioritized list where we say, okay, these ones should be—we then have expedite actions where you can just see basically which are the priorities and if you could be shortening the lead times, because for this, for the supplier, doesn’t matter. They have 20 of your parts and they’re all repairing them and they don’t know which ones are important to you, but for us, we say, “we have five serviceable still lying there, and just because we sent it out earlier we don’t need back earlier.” So this could be something very typical when I talked about prioritization. Again, which one has the most important impact for us? So, yeah, that could be, again, it’s, in the end, just lists of what do we do to get the highest impact?

Conor Doherty: Okay. I can, I can push on? You’re good? Okay, cool.

Because you mentioned suppliers, that leads to, again, another key concrete question that was asked. I have it written down here. So essentially, this came from someone who has previously seen suppliers refuse to join a new platform because of onboarding costs or subscription fees. So, essentially, “do suppliers need to change how they work with the client in order for Lokad to be able to do all of this or generate value?”

Fabian Hoehner: Suppliers in terms of what, the suppliers of the clients?

Conor Doherty: Yeah. Yeah. So

Fabian Hoehner: No, I mean typically, so they are the—I mean, do we, do we need the data of the suppliers? The short answer: no. I mean, we just need the data of the client because they have all the data.

Can we? And yes, we do actually integrate then with some suppliers on top of that, typically for timestamps. So, you are the MRO, so you repair. I’m the airline.

I’m just sending my data to Lokad, and the only thing that I know is I’m—I send the part out, and I know when it’s coming back. If I have your data being “it’s currently in a repair process. It’s currently at the border, or whatever, it has been checked.” If I have that data and we do have some clients who have good with our clients and then could be as easy as an Excel sheet that is just uploaded to Lokad on a regular basis, then we can better update these data points.

So, to answer your question, if I’m integrated with my supplier and I have the data in my ERP system, in that case I’m just pulling the data from the ERP system, that’s fine. Done. But we are very flexible, and this is probably the advantage that we have to integrate with the third-party supplier. It’s again, I mean, we’re an IT company.

For us, it’s probably a lot easier to integrate with a third party, like any kind of data flow than it is for a big organization to, you know, get third-party data into an ERP system. That’s a transition project. For us, that’s a that’s a couple of days. So, that’s probably the big difference.

Conor Doherty: Okay. But in conclusion, again, there’s at least two categories of data. There’s the nice-to-have and then there’s the must-haves.

And having the supplier data that you just described is nice to have. Essentially, it’s nice to have. It helps, but not a dealbreaker.

Fabian Hoehner: Yeah, absolutely. So, just to be clear, yeah, good point.

Conor Doherty: I want to be concrete, right?

Fabian Hoehner: Yeah. I mean, the data question is, I mean is always good. So, what’s, what’s the kind of data we need?

I mean, basically the— anything that—so aeronautics, you have to track everything. So, there is enough data. The question is, “oh, do we have enough?” Yeah, I mean, you do.

Question is, is it nice? Is it clean? No, it’s never. But there is sufficient data, and then the work begins in getting the same interpretation of the data.

So that’s— data is just data, but how do you interpret it? That’s the question. Yeah, consumption history, that’s obviously important. Stock levels catalog, yeah, everything, but the standards.

And then, yeah, there’s much nice-to-have data. Historic stock levels per location, stuff like that. I mean yeah, then, obviously, supplier timestamps, like, it’s super difficult. Typically, you don’t see that.

So you just see the turnaround time as two values. So out and in and that’s it. If you have the intermediate timestamps, that’s really powerful, but, yeah, if you don’t then you just have a distribution of, that is large, of those. Otherwise, you have several distributions.

Conor Doherty: You, you commented earlier again with regards potentially messy data. You said “then the work begins.” But just to clarify, and then this is something that was also asked in terms of data preparation, what is required of clients? And this kind of ties back to what you said about supply chain scientist earlier, I guess.

Fabian Hoehner: Yeah. I mean, so normally, we, our attitude is that we do it together because, I mean, this, again, this is what we’re designed to do. We are big data platform.

We are quick in that. We are typically a lot quicker than our clients are in even preparing the data. So we just need raw extractions. having said that, obviously, having people that are familiar with the data, that know where the data is, is important, is of value, speeds up a project.

So I’m, I’m not going to say that it’s not great to have people that have looked into the data, know what it means before. But otherwise our approach is you go to the, and this is the beauty, you go to the decision level. So if I tell you, again, “buy four parts, do that”, the underlying issues become pretty obvious. So doing a data cleaning mission.

So, basically, say, oh, we—I’ve seen many companies that say, “oh, we are not ready. We want to clean our data first.” And then you speak to them two years later, and the answer is, “we are still cleaning our data”, because it is incredibly difficult to know where to start because there’s always a messy data, not the right categorization, so hierarchies of products. Yes.

Yeah. Okay, cool. I mean, for us, would be nice to have a better categorization. Interchangeability mapping, that’s obviously— interchangeabilities in aeronautics, one-way, two-way, super important.

So, yeah, it’s not clean. Would it be a lot better to have it clean? Yes, absolutely. But will you also, by going to the decision, see very obviously where there is data that is of value?

Yes, because let’s say, interchangeability, so a new part, can I use it in an old equipment, piece of equipment, or not? Yes, or— that will become pretty obvious if I’m giving a wrong recommendation. So obviously, if I have a part that is that is compatible in both ways, I want to have it more aggressively. That’s interesting.

So if my recommendation is very wrong to again, to a user, I say “five”, the user say, “hey, why? I mean, we don’t need the old part anymore. The new one can fix the old and the new part, so we want 20.” In that case, we are very quickly come, when we look at the inspector and so on, to the conclusion, “oh, yeah, we didn’t know that this part actually works for these two demand scenarios.”

Why? Because we didn’t have the data right. And then you know, “yeah, this is data that is worth fixing”, because we are actually looking at the yeah, at data points that are helping us. While if you just clean for the sake of cleaning, you’re never going to get anywhere and never going to stop cleaning.

Conor Doherty: Good.

Fabian Hoehner: what you smirking about?

Conor Doherty: Somebody congratulating us on— Alex just refer to himself as smart. He is pretty smart. He’s a smart boy. Yeah.

So, again, one of the questions, and I’ve kind of amalgamated various versions of this question into one general version, but it reply—it pertains to, essentially, implementation. So, specifically integration efforts, data requirements, you’ve already touched on data requirements, but integration efforts, data requirements, security, support, scalability across a large supplier base, engineering, just like the comment on that, timeline requirements, etc. Etc. Again, I realize, how long is a piece of string? It depends on each. I, I understand that, but it’s a broad stroke answer for people who are watching and might be interested.

Fabian Hoehner: Six months.

Conor Doherty: Okay. Approximately, can be faster, can be longer.

Fabian Hoehner: Yes. And yes. Okay.

Now, okay. So we typically, we just, again, depends on everything, but two months to get, to get data and to get an approx—So to get the same understanding of what you’re looking at, so high- and low-level data health, where you just, again, have the same interpretation of the data. Two months for a rough both for an optimization. So we’re actually quite quick in that and then another two months to fine-tune, but also run in parallel, because the way that we, that we operate is, again, these are—everyone is doing what we’re doing here.

I mean, the companies that we work with, they are already taking decisions on a daily basis on “what should I be buying? What should I be investing into?” And, so, basically, you just run that in parallel to their processes, and this is where you get to improvement. That is bringing me to, actually, one, just a little feature in terms of what, historic, how do you how do you compare to, to the past, or how do you evaluate?

One feature that I that I like to just highlight is on the platform, we are completely compatible with with evaluating the past. What do I mean by that? So if we, if we look at the history, I can see all the past executions that I have. So for example, I can literally, you can see, basically, what I did, did that today.

I can look at the dashboard that I executed that I executed a couple of hours ago. So, and what I can do here, and what you can actually see what I did is I was assuming that you would be asking me about budget. So, in that case, I put the budget to, sorry, to 1 million instead of 10 million. And so now if we go back to our 98%, where before, what was the investment, you remember?

Conor Doherty: 1.2—2 million.

Fabian Hoehner: Very good. So, and what we are doing here now as now we only get to 900,000, because now I put in a limitation of 1 million. And we can actually see I can play with this dashboard. I can execute it.

I can do that with any dashboard that I had in the past. I can tell you most people don’t think that this is very—that’s a very important feature, but it’s actually incredibly valuable, especially if you want to go you can go back in time to any point in time and see what we did. So oftentimes, when you have an AOG situation, companies will, I don’t, I don’t want to say panic, but you go into an investigation mode to say, “oh, how did it happen? What did we do wrong?”

And in that case, it becomes incredibly, I mean, we’re talking about lead times of six months of a part, and it becomes incredibly difficult to challenge your system and to challenge what you’re doing if you’re not able to put yourself into the spot that you were on in January 2026. But in that case I can just go into my simulation 2026, and I’ll see that I’ll potentially see that “yes, we wanted to get a third unit of that one. It was interesting to buy. However, we were constrained and we only had a budget of, I don’t know, of 10 million, and therefore that’s the reason why we didn’t buy it.”

So, but then we know, and so the consequence would be, “yeah, we should be increasing our budgets and we’ll have more money available.” Or the consequence could also be, “oh, actually, yeah, we didn’t prioritize that right.” So we should have given that a higher penalty of being out of stock, and then the consequence is that we change the algorithm, and that is something absolutely crucial. So you go from I want to say, a kind of pointless investigation to challenging the underlying system, and the consequence of that is then either you say “the algorithm did what it’s supposed to do and it’s fine.”

Yeah. Sometimes you do have extreme events and you do have stockouts. So what we’re saying here, we go to 98. So that means 2%, we will have stockouts.

That’s just that’s statistics. But you could also say “look past that. We should change the algorithm so that in the future is better”, but that’s a continuous process. So optimization is a continuous process.

It’s a continuous flow of: you try to strive for perfection, which by definition you never reach. Yeah. So that’s the and therefore, having something that you can actually look into the history, and I can actually execute new data on an old code. So I can take the calculations, the weighting, the what’s the penalty that we give for a customer-facing part, and so on.

I can execute data from today with a code that is on an algorithm that is six month old, or vice versa. Take the old data and say, “okay, what would it do if I was to put that into into today’s optimizer?” And again, very few people will ask about that, but that’s an extremely strong, strong feature that was initially created to fix bugs because that’s, it’s important, because if data changes from one day to another, you can just not have the bug anymore, and that sucks because then you know there’s something, but you can’t pinpoint it. But anyway, I digress.

Conor Doherty: Okay don’t digress. It’s good to have the concrete information on the record. We have a few more questions.

You’re, you’re good to go? Keep going. Perfect.

Fabian Hoehner: No, sorry.

Conor Doherty: All right. This one again, this one isn’t— it’s not explicitly about purchasing. It’s kind of more about other optionality.

So, it’s essentially, great. “Can Lokad decide whether we should buy another unit or move an existing unit between locations?” So, we’re kind of getting away from the topic of today. We’re not going to open up another dashboard, but a high-level answer to that.

Fabian Hoehner: Yeah. So,

Conor Doherty: Yes. Next question.

Fabian Hoehner: Okay. Done. No.

So this is where economic prioritization comes in. So, in the end, to the probably, and this is going to be very boring, but my answer is always going to be: “if it’s logic, we can do it.” So what’s the logic that we should be applying here? The question is just, so first you want to simulate: what is the probability of need?

So, in that case, so for an allocation problem, you first need to simulate the demand per location. For example, you have MBKs on 100 outstations, you need to, I mean, it’s obviously, it’s difficult. It’s more difficult to forecast when it’s even more sparse, but you need to do it. We always say “it’s not the it’s not the ease of the calculation that decides the level of simulation, but it’s the decision that you want to take.”

So, if I want to simulate the need for a part by location, I need to forecast on a location level. To now answer the question: what do we need to do? Do we need to buy more for the central stock or do we need to reallocate? It’s a question of economics in the end.

It’s just the question of what is the cost of, what’s the cost of moving the part relative to the and then, so you move the part. So that means you expose more risk on the station that you’re, that you’re moving the part from, and you, so increased risk plus logistics cost versus external investment cost. So that is the cost of capital. Yeah, that’s, in the end, fairly logic and doable to calculate.

Yeah, and for us, the bottom line is always so we can do that to the euro, dollar, a pound, whatever. But the question is always: how big, how often do we have a question, a problem with that? You can also solve that with a simple rule of three. But if it’s a relevant and high-value question, then this, we can go to exactly that level.

Conor Doherty: Okay. Some—

Fabian Hoehner: Tell me if it’s not clear.

Conor Doherty: No, no, no. Perfectly clear. I’m also aware that there are other questions to get through, and I don’t want to drag down. Again, anything that is unclear, I’ve already had a few messages from people saying they want— we’ll follow up, because this is going to take too long, because, “okay, what about this situation? The situation.”

Fabian Hoehner: Just to be clear, I just don’t want to— So, my point is I wanted to make sure that we get those two principles: uncertainty, and so, basically, different uncertainties, turnaround time and demand mixed, gives us a view of the uncertainty, and economic prioritization. With those, we can pretty much answer, 95% of all underlying questions now. So, be it you asked in the beginning, consumables and expendables. Now, we talked about the loop.

Okay, loop is more complicated, but in the end, C&E is the same. It’s just stock in and out, so you’re 100% scrap, if you want to say. And the underlying principle is somewhat the same. You can do exactly the same economic optimization.

You can you don’t have to. If the underlying problem is simple, you can also resolve it simpler. So we don’t have to do it with that kind of an optimization. But that would be the normally the strong recommendation. But again, if we just have to do quick and dirty, we can always start quick and dirty to just have something that’s better than the existing and automated, because automated pretty much always beats human.

And then go to the second stage. yeah, so just the—I do have tons and tons of other dashboards, but that I don’t want to go into, but just, basically, so allocation. Yeah, of course you can have a dashboard and then it tells “you need to allocate from one location to another.” Okay, great. again, I don’t want to go into the math here.

Conor Doherty: Yeah, that’ll be a subject of a, of a follow-up. That’s going to muddy the water.

Fabian Hoehner: High-level things that I wanted to to point out, but, yeah, we do have I would say, it depends always: what do you define as a module? So is is different lead times on turnaround times scaling up, is that a module? It depends on—or is it just purchasing in the module? So, but you can say we have 20 different modules, so we can see some ideas here.

Conor Doherty: Yeah. Actually, on that, on that note, again, this is, perhaps somebody missed an earlier section, but it’s a live question, so we’ll ask it. It’s from Saddaf, hopefully I’m pronouncing that correctly. “How—can you comment a little bit further on how Lokad handles supplier delays when calculating the optimal, the optimal purchasing quantity?”

Fabian Hoehner: Yeah. So, I mean, the—

Conor Doherty: You can show again if you want to jump back into the— into a dashboard just for anyone who might have missed it earlier.

Fabian Hoehner: Yeah, sure. I mean, the question is: what is, what is a delay versus what is just something that you need to appreciate? So a delay means that you have a, you have a value that you expect, and now it’s taking longer. So there are two, two answers to that.

So the one would be the expedite action. So, basically, you assumed a certain you assumed it would be taking 30 days, and now you’re at day 40. In that case, what we provide typically is a ranked priority list in: who do you call first? Because, again, if I have a part that is 10 days late, but I have five serviceables lying around, I don’t care whether that’s late.

It’s really not important to me. On the other hand, if I have one that is not even late, it’s, we are at day 25, but I’m out of stock and I know it’s super urgent I probably want to pick up the phone and just tell them, “hey, can you really— what, whatever you can do to get the part in.” So, that’s the half of the answer on: how do we deal with delays? The question is, can you do something?

And then give users the possibility to prioritize, right? And this is always: what’s the impact of something being wrong? That’s the first half of the answer. And the second half is what we looked at here would be—

Conor Doherty: what’s the, what’s the definition of delayed?

Fabian Hoehner: Yeah. Yeah. I mean, in the end, we would rather say, I mean, it’s just there are different lead times, and, yeah, nice to have that in some contract it says “60 days.” If in reality it takes always 120, what you want to do is you want to appreciate that in your purchasing recommendation.

That’s exactly what we did here. So, you don’t say “it’s always 80 or it’s always 120” or doesn’t matter, different numbers here. But you say “there are all these possibilities”, and then, again, the demand distribution that we see here is a demand distribution over all possible lead times. So that’s exactly why we are doing all of this.

So to embrace the uncertainty that is uncontrollable, and that’s— turnaround times and supplier lead times are a big one in that. And, to be very clear, I mean, we didn’t go into that, but this is, this is also forecasted. I always spoke about made it—I’m making it simple. So, basically, we just say we look at the past and the past represents the future.

We are obviously aware that the past doesn’t always represent the future. So, we saw, obviously, we were there during COVID with our clients and after COVID, so, yeah, decrease and then ramp up and incredible increases in turnaround times. We could also see, basically, if a turnaround time was going this way, we wouldn’t say “the past represents the future.” No, no.

Obviously, then we forecast turnaround times and then forecast demand over forecasted turnaround time distributions. And I’m just one—anticipating one more question. Again, I said we’re looking at the past year. Obviously, we can look at checks that are coming up. So if I have a— Yeah, feel free if you have a question.

Conor Doherty: No, no, that is actually going to be part of a later one. So please keep—

Fabian Hoehner: Okay, so, basically the typical question would be: so how do you forecast future demand? Do you just look at the past? It depends. So if you have an operation of a thousand planes and so big MRO, and you have lots of statistical mass, then actually looking at the past is actually a pretty good representation of the future.

However, if you have more information, and that’s always the question, what’s the information level that you have? For example you know you have your row of D-checks coming in. then, of course, we are going to take the probabilistic BOM, the bill of materials, which is in itself uncertain, you know, in theory. So some parts you know that you’re going to replace, then you have a 100% probability of replacing a consumable part. But others you just know I’m opening it up—

Conor Doherty: and then you see corrosion and you weren’t expecting that.

Fabian Hoehner: Yeah. Or you know that based on past type C-checks, you replace the part in 30% of the times. That’s a probabilistic bill of materials where you say “these are the 100 parts that I’m going to need with their probabilities.” And if I know that this event is going to happen in two months then I can plan that out.

That’s exactly what we do. Now, the reality is, of course, more complicated because when you say “we have a C-check in two months”, how often is it really going to be in two months? There’s also uncertainty in that. So, again, when I’m saying I’m introduced two main uncertainties, in reality, we can have even more, and that is just always going to just increase my uncertainty, increase the distributions, but in the end, it goes, it boils down to the same question: how much coverage am I getting if I’m investing into one more part, and is it worth it? So am I willing to spend, I don’t know what this one costs, 10,000 bucks to get another what is it, 0 point, or another 4%, or whatever?

Conor Doherty: Clear to me. Again, also, also, again, it’s good to point it out, is why having the dashboards here is so critical. A follow-up, it’s, it’s from Metan, forgive me if I’m mispronouncing it, Met—it’s actually about, excuse me, a very key point that had been raised before which is, like, generating, let’s say, the greatest set of investment, divestment or allocation decisions is amazing, but if there’s a lack of adherence, meaning, if people are just not executing that’s actually a problem.

So this question is: how do you actually know— this is a public question—“how do you actually know whether the company, let’s say a client, actually bought the suggested part?” So basically you were talking about traceability earlier. Can you, can we track, oh, “X percentage of the time these decisions were followed?”

Fabian Hoehner: Oh, okay. Yeah. So

Conor Doherty: Simple audit, audit, auditability, excuse me.

Fabian Hoehner: So how would you track that, Conor?

Conor Doherty: With a computer.

Fabian Hoehner: Yeah. Okay. Great. So, bottom line, we see the execution of the if I’m getting, on a daily basis I’m getting the transaction histories on everything.

So that includes the purchase. So I literally see whether someone followed my recommendation, and I mean, obviously, I’m, I’m simplifying a lot here, but just, let’s just say on Monday, I recommended five. So Sunday night ran through. Monday morning, the operator sees, “okay, five recommended.”

And then on, then the Tuesday run, I see five. Then I can say “I have 100% adherence.” I’m simplifying a lot, and in reality, it’s, it’s more complicated than that, because you can have, obviously, in these kind of questions, long delays before you make a proposal and so on. But, bottom line, yes, we do follow that, and for something that is automated, so if we talk consumables, you can track that really well.

So there, it’s a lot easier. You have just pure mass. That’s easier. Here, so on the entire rotable side that we talked about, I would argue that the, just the human feedback is probably the most important. So, literally, I mean, this, these lists, we look at them with our clients, and again, this is not— we are very good at math, statistics, but when I say we, I mean, it’s—

Conor Doherty: yeah, you’re absolute genius.

Fabian Hoehner: No, I mean, it’s, it’s a very strong, platform, and we have smart, smart engineers and that is our expertise. I mean, we’ve been in the aeronautic sectors for the last 13 years or so. We have built up quite some expertise, but we never claim to know aeronautics better than our clients. I mean, when is a retrofit coming in? I mean, again, when we are with aeronautics companies, and these are very passionate people, when they see an airplane, they know immediately which plane, what are the parts in there, and so on, and we just, we are more the statistics side.

So what does that mean? We always heavily rely on the expertise of and the feedback of the users, and then we also design the dashboards 100% for them. If they tell us, “yeah I want this to look different. I want different colors. I want whatever.”

Yeah, we listen, and we design it for that purpose, and that’s I mean, rebuilding that, I mean, that, yeah, just redesigning that, that’s half a day for us. Not even. I mean yeah, if you want, I can talk about AI, but that’s a different topic.

Conor Doherty: No, I will push on. So this is more just about a comment. Sorry, hold on. Yeah.

Yeah. It’s basically more about the—how much of the manual purchasing process can actually be removed if, if working with Lokad, because I don’t want to misrepresent this. You did touch on this earlier when you talked about rotables. Again, depending on the price that you’re paying, you’re obviously still going to have an expert in the loop to validate a $90,000 purchase, but how much of that process can actually be— how much of the time invested in that process can actually be liberated? And then same question when you’re talking about C&E.

Fabian Hoehner: Yeah, it’s—I’m going to resume my answer from before.

Conor Doherty: Yeah, I’m just reading.

Fabian Hoehner: Yeah. No, sure. But it depends on the complexity of what you’re doing. And I would argue that C&E you can really automate to a very large degree.

And when I say, I mean, beyond—Yeah, I know that much is already automated with the reorder points. We have a min-max. We can automate that, but with a lot smarter system. So something like that that can potentially, for example, just reset reorder points on, if we want, on a daily basis, so that it executes to the decision that we want.

We do that for some clients. And then I would argue there’s huge time savings in that as well. But when you, again, when you purchase for millions of dollars parts I would argue, on that point, it’s less about the time savings, that are definitely there and quite huge, but it’s about being better. I mean, that’s really, like, this is the name of the game. Like, you avoid, you avoid a few AOGs, and then costs become irrelevant, I—to certain degree.

Conor Doherty: I think I have covered, because it’s now 5:45. We’ve been going for about 75-ish minutes. I know there was a hard out at around six, so I think, oh, no, there’s one last, excuse me, there is one last question. I think anyone who’s watched the demo will understand it, but just to be concrete the question is: how is Lokad different from other procurement platforms that we’ll find in the space, like, for example, Aeroxchange, because one person is actively comparing two categories and wants to understand the distinction between decision optimization and things like processing RFQs and quotation comparison and PO processing, etc.?

Fabian Hoehner: so, I mean, there’s, there’s different aspects that no, I’m just saying there’s different aspects that Aeroxchange does, and so but high level, I just want—we are not a, per se, a procurement platform. Again, a system of intelligence. So the difference—

Conor Doherty: decision-making.

Fabian Hoehner: Yeah, so, again you have your transactional systems. We sit on top. We are the brain, and we push back the best possible decisions that statistics can provide and business intelligence that then should be executed. Aeroxchange is also has more procurement features of automating the procurement function itself, the communication.

That is not what we do. Again, we can integrate with different suppliers, so then match things and have communication in there, but again, that’s, yeah, we don’t list prices or stuff like that. That’s, that’s just not our, our business model. So I wouldn’t call us a procurement platform. Again, I would say decision, decision automation, decision support and system of intelligence as, as summary.

Conor Doherty: Oh, all right. One last, it’s, again, more comment. People appear to be more comfortable, understandably, I guess, DMing than publicly commenting. And I’m just reading verbatim: “nice, but I’m—nice, but I’m more interested in repairs. Can this approach be applied to prioritize repairs and expediting?”

Fabian Hoehner: Yes. That was easy, right? Yeah,

Conor Doherty: Humor the audience a bit with that one.

Fabian Hoehner: No, I mean, in more seriousness, it’s, it’s, again, the same logic. So now, instead of, instead of having instead of buying a part, I’m asking, even, so I’m asking whether I should repair my unserviceable. Let’s say you have 10 unserviceables and the first zero in stock. And then instead of buying, you’re just making it serviceable.

Making it serviceable in that case is a repair that has a repair loop, which is a turnaround time, which costs something. There’s— whether it’s internal or external, both have a cost, but let’s make it external repair because it’s simpler. Let’s say it costs 50,000 to make the part serviceable. The question is, first, how do I prioritize?

Because I have a budget of repair budget that I want to, what I want to send out, or even just physically capable. I mean, I’m capable to process 50 parts a day. Which ones are the most urgent? And then even which ones are just economically, feasible to repair or not?

So, for example after COVID, we had, that was a big one. So, for some of our clients, cash was obviously becoming quite an issue. You, like, there was simply no cash. So the question is what do you do?

because you still have all your leases and so on, but you don’t get any more cash inflow. So the first thing that you could do was actually exactly that: just hold repairs, because there’s a big cash outflow just repairing unserviceable parts. So you can just cannibalize your existing stock and leave them unserviceable, which for that time was very reasonable to just leave stuff unrepaired, because, obviously, if you don’t have demand, so during COVID, 20% of the fleet is flying, that’s going to reduce quite significantly your demand for serviceable parts. You just stop, but you can’t, and this is the, again, the problem, you can’t just say, “I don’t repair anymore.”

You need a prioritized list. So you need to prioritize between some parts that are crucial, you still need to repair them because if you don’t have them, this is really going to break your neck. While others, yeah, you have 10 on stock and you have five unserviceables, you know what? You’re going to be fine to take a little bit more risk and leave a couple more unserviceable and, or just repair one.

So, the same logic. Yes. And then turnaround times become even more important. So, if you then have extra timestamps probability of scrap, very important in that. Yeah, but same underlying logic, and yeah, also show dashboards for that, but that’s leading down a different path.

Conor Doherty: All right. My closing thought is if you were to summarize, again, like, full circle, if we were to summarize the key points that people should take away from this, if they’ve only, if they’re only tuning in now or if they’re just skipping straight to the Q&A section, what, how would you summarize approach to purchasing, investment, divestment decisions in aerospace? And again, what are the key takeaways for people?

Fabian Hoehner: Yeah, first, they should be ashamed for not listening to everything. Secondly,

Conor Doherty: Pretty good. Third one for—

Fabian Hoehner: So I’m going to repeat the: appreciate uncertainty. That means uncertainty exists in many forms, be it turnaround times, be it demand, be it probabilistic bill of materials. So it exists everywhere, and don’t plan it out of the picture by simplifying. So just appreciate it, and there’s technology to do it better.

And then, secondly, prioritize with economic consequences. In aeronautics, typically, the consequence is service level gain or AOG avoidance per dollar spent. These are core concepts to take away. And, yeah, what do we do at Lokad? We have a powerful platform to do so, and we have brilliant supply chain scientists that design this together with our clients, and yeah that’s Lokad in a nutshell for aeronautics.

Conor Doherty: All right. Fabi, I don’t have any other questions. Thank you very much for joining me today. It’s a pleasure to have you in the studio, and I hope everyone else enjoyed listening to your voice as much as I did.

Fabian Hoehner: Too kind.

Conor Doherty: You’re a credit to humanity, sir.

Fabian Hoehner: Thanks.

Conor Doherty: No problem. And thank you all for watching. Thank you for your comments, your questions.

A special thank you to all the people I spoke to over the last month or so. I mean, we promoted this event for about a month. I connected with a lot of people, asked a lot of people, “what did you want to see?” And, as you can see, I took the feedback and I tried to adapt what we saw today, or what we showed today to the desires of the audience.

Now, if there are things we didn’t cover that you are actually interested in, feel free to connect with Fab and me directly on LinkedIn. You can see us clearly. We like to talk. We’re lovely.

Click on the profile, shoot us a message. Or, if you’re already convinced and you want to book a call and learn a little bit more, there should be a link in the chat. Or, failing that, you can shoot us an email directly at contact@lokad.com. Now, as Fabi showed a little bit earlier, there are other modules that we can talk about in aerospace.

We didn’t cover them today, but we will be back at a future point to cover those. But until that day, there’s nothing left to say except yeah, get back to work. Oh, okay. Please.