Home »
Salescast » Here
Getting started with Salescast
Testing and going into production with Lokad
does not require upfront payments (free trial). It is a very lightweight process, however a couple of hours of work need to be invested to properly setup the data export. 90% of the overall effort is not specific of Salescast, and relates to the data retrieval.
Get Started
For a mid-sized company, the typical process is:
- Management issues instructions to IT and supervises the proof of concept (POC): 1h.
- IT prepares the data import: 3h (1 person).
- Management validates the forecasts: 1h.
- IT finalizes data access for production: 1/2 days (1 person)
Naturally, the Lokad team is always here to support you (
contact@lokad.com).
N.B.: Yes, IT has to be involved, importing data from Excel will not work. We have an extensive experience supporting this unfortunate conclusion. A lot of competing solutions support importing data from Excel, but we believe that, despite the best intentions, the process is not in the best interest of our clients. For more information, please read the article Excel as data repository will hurt your companyHistorical data needed by Salescast
Our forecasting technology is strictly statistical, and this section is intended to give you a better understanding of what data is actually needed by Lokad to produce the best possible forecasts.
There are basically
3 types of data that support the forecasting logic:
- Order history (required).
- Item descriptors, also named tags (optional but advised).
- Events, such as promotions (not advised for initial deployment).
We use the word
item to refer to SKU / Product / Barcode, i.e. the individual item being stored and sold, the terms varying from one field to another. For each data type, we provide more details below.
Before digging into the specifics, let's emphasize that there are
data typically NOT needed by Lokad. Our forecasting engine is natively designed to leverage
calendar-based patterns such seasonality, Mother's day, religion events such as Easter or Ramadan, etc. Hence, those information do not need to be transfered to Lokad, as the forecasting engine will auto-detect calendar-related sales patterns.
Sales order data
As a basic principle, the more data we can analyze, the higher will be the accuracy of your forecast. This applies to order history and aggregation level. The ideal scenario is therefore for us to receive from you
daily sales data on an item level, for at least 2, ideally 3 years of history. Lets explain this in more detail:
Concerning
order history, the longer history is - the better. Having longer history is about the easiest way to improve forecast accuracy. As a rule of thumb, with only 12 months of history, no yearly seasonality analysis is possible;
2 years of history is good, 3 years (and above) is excellent.
Note: With Lokad, we don't need all products to have a long history. When we say "2 years of history", we only mean that at least some products have 2 years of history. Based on the analysis of those longer-lived products, Lokad is able to apply seasonality to more recent products that have less than 12 months of sales history.Then, it must be noted that data aggregation is a lossy process that discards valuable information. In practice it means that even if we want to produce weekly forecasts, it's better to have the daily sales as input; better in the sense that it will result in more accurate forecasts.
Also, it's better if data is available at the
lowest level such as SKU (rather than product family). In the end, your inventory needs to be optimized at the SKU level. Unlike classical forecasting system, Lokad is very capable at dealing with intermittent demand.
In practice, the order history is available in its fullest detail in the ERP, eCommerce, or accounting package already used by your company. This is where we are going to get the data from.
Tags (product descriptors)
Item descriptors - "tags" in the Lokad terminology - represent additional data that is leveraged by Lokad to deliver better forecasts. In practice, a
product group, family, sub-family are ideal candidates for tags, Basically it's hierarchical information that is used to differentiate an item from all other items sold by your company. For example, if we consider an item that has less than 3 months of history, it is nearly impossible to establish the yearly seasonality of the item through a direct correlation with sales of other items: there isn't enough data to support the correlation in the first place. Yet,
with tags, Lokad is able to learn which items share the same yearly seasonality, and apply a seasonality pattern even to a product that has been on the market for less than 3 months.
In practice, much like the transaction history for orders,
your current business application already contains a structured description of the items being sold. As a result, there is no need for a tedious manual reentry of this data to feed Salescast with tags. Salescast will be able to import item descriptions already available in your systems. In our experience, identifying 3 or 4 tags is enough to yield substantial accuracy improvements.
Events (ex: promotions)
Events are more subtle, because in our experience, very few companies actually keep track of them in a
sufficiently precise way so that the information is directly exploitable for statistical purposes. In general, we advise to exclude events from iteration with Lokad, in order to focus on the most immediate benefits first.
Your IT team or Lokad, your call
Finally, there are
two ways of exposing data to Salescast. The most frequent option:
your IT staff takes care of creating a replicated database, where your historical data is formatted according to our guidelines. In this case, there are no setup charges, as Salescast will be able to natively import the data exposed by your company. Usually, this is the most
cost-efficient solution. We suggest to forward our
technical instructions to your IT department to get started with the data export.
The other option is:
Lokad staff takes care of creating a custom data adapter within Salescast so that we are able import data from your system without requesting any work from your IT staff, except granting a remote access to the data. In this case, we typically agree on a package fee that ranges from $3000 to $8000 depending on the complexity of your setup.