Go back to Market Research
John Galt Solutions develops Atlas, a supply chain planning platform covering demand, inventory, supply, production scheduling, distribution, and sales and operations planning. Manufacturers and distributors use it to calculate forecasts, inventory requirements, production plans, and replenishment decisions. The company also sells ForecastX, a forecasting application within Microsoft Excel. 1
Overall supply chain score: 5.0/10.
| Dimension | Score |
|---|---|
| Supply chain depth | 5.4/10 |
| Decision and optimization substance | 5.8/10 |
| Product and architecture integrity | 4.2/10 |
| Technical transparency | 4.6/10 |
| Vendor seriousness | 4.8/10 |
Atlas combines forecasting and inventory calculations with configurable decision workflows. Its published methods extend to uncertain lead times, network risk pooling, purchase-order economics, and sequential decisions. Its practical limits lie in the opacity of those models and the continuing dependence on customer-defined targets, objective weights, and rules. Naming Q-learning or a digital twin does not establish that their recommendations outperform a simpler policy under the actual economics of a supply chain. 9 10 15
The automation record is more substantial than the generic planner-assistant positioning suggests. Reddy Ice documents automatic next-day replenishment and delivery planning, supported by operational data and reported improvements in stockouts and delivery productivity. The newer LATICRETE agentic deployment describes information retrieval and exception prioritization, with autonomous execution still a subsequent step. These are different levels of productivity, and the newer AI label should not obscure the distinction. 4 12
The commercial narrative also outruns its measurements. John Galt promotes 95% forecast accuracy without defining the resolution or metric on the cited page, and presents configuration and adoption as a path to rapid returns without a corresponding account of recurring ownership costs. The public security material provides little basis for judging containment or recovery of production planning data. 17 24 27
John Galt Solutions versus Lokad
John Galt organizes Atlas around planning applications, reusable workflows, and customer configuration. Lokad concentrates on executable decision logic that ranks purchasing, production, allocation, and replenishment actions by expected economic contribution under uncertainty.
Atlas’ AI guide discusses probabilistic lead times, demand distributions, network relationships, and Markov decision processes. Its procurement examples include order minimums, bulk discounts, and supplier selection. These overlap with concerns central to Lokad. 9
The sharper distinction is how the objective is expressed. Atlas’ multi-objective interface lets users weight profitability, service, sustainability, and other goals. Lokad’s design perspective seeks a consistent economic valuation of competing actions, including the costs of shortages, waste, delay, and scarce resources. An objective weight can express a management preference while hiding how much money the company should spend to improve a particular measure. 10
Lokad’s inventory optimization approach also treats recurring decisions as computations to execute and improve. Atlas can reach that end-state, as Reddy Ice demonstrates. Its assistant and workflow offerings also support a different model in which planners continue investigating exceptions and making selections. The relevant productivity comparison is the remaining work per decision cycle, including the specialists who maintain the logic. Both vendors’ engineering and service costs belong in that comparison. 4 12
Technology history: delivery, then broader promises
John Galt dates its founding to 1996 and retains ForecastX alongside Atlas. The forecasting product’s public documentation is concrete: Procast selects a method for each series by minimizing a chosen error statistic; safety-stock settings expose lead time and service level. This is a recognizable statistical-software foundation. 29 2 3
The Reddy Ice relationship provides an eleven-year trail that is more informative than a list of product awards.
| Date | Published commitment or result | What it establishes |
|---|---|---|
| October 2015 | Reddy Ice selects Atlas to combine weather, seasonality, and events with demand and supply planning | A defined deployment problem and an initial automation objective |
| May 2018 | John Galt presents machine learning and IoT work; store replenishment reportedly needs almost no manual intervention | A narrow operational automation claim, alongside a broader promise of unprecedented forecasting performance |
| June 2022 | A customer presentation describes demand and replenishment automation across more than 82,000 locations | Continued use and a materially larger stated operational perimeter |
| April 2025 | A further customer presentation again describes automated replenishment, production, and delivery planning | Continuity of the deployment beyond the initial technology announcement |
These statements come from the vendor’s dated releases and customer presentations. The progression supports delivered automation in this application; the public material does not isolate how much each algorithm contributed or publish the complete ongoing cost. 30 5 6 7
The subsequent roadmap introduces different mechanisms. Intelligent Workflows arrived in September 2023 as a no-code canvas for rules, tasks, and processes. The May 2024 probabilistic-planning announcement named Q-learning and uncertainty across purchasing, manufacturing, demand, supply, and costs. August 2025 brought weighted multi-objective optimization to parts of MEIO and rough-cut capacity planning. September 2025 added generative explanations of MEIO and ensemble forecasts. 15 8 10 11
These releases describe more than a change of vocabulary. They also leave distinct commitments open: calibration of the probabilistic model, the economics behind objective weights, and the faithfulness of generated explanations. A plausible narrative about why stock moved is not a reproducible account of the computation that moved it.
The July 2026 LATICRETE announcement is especially revealing. Its headline presents an end-to-end agentic deployment; the customer describes easier access to insights and a goal of progressing toward agentic workflows. The stated current uses are finding exceptions, directing attention, and identifying risks. John Galt itself places recommendations and autonomous execution in the next stages of adoption. The deployed capability is an assistant that supports planners, not the end-to-end autonomy suggested by the broader product positioning. 12 13
The historical judgment is therefore specific. John Galt has delivered unattended operational calculations, but its latest autonomy story is broader than the named agent deployment proves. The Reddy Ice result should be credited to the forecasting, data integration, and execution work that produced it, rather than retroactively to generative AI.
Probabilistic planning: the right problem, an insufficiently exposed model
John Galt’s AI guide describes network risk pooling, statistical demand distributions, dynamic programming for inventory, and probabilistic lead-time ranges. It also names Markov decision processes and Q-learning for sequential planning. An illustrative furniture-manufacturer case traces a launch risk to a small, externally sourced component and describes buying additional stock before production. This is a useful example of dependencies across a bill of materials rather than treating every SKU independently. 9
The missing substance is the specification of the model. A sequential decision method needs a state representation, feasible actions, consequences, and an objective. The public material does not expose how Atlas models correlated supplier delays, distinguishes lost demand from observed shipments, values service failures, or calibrates the distributions used in its experiments. Naming the learning algorithm does not supply those definitions.
The same issue applies to the digital twin. Atlas promises a virtual representation of suppliers, production, inventory, capacity, and transportation, with scenario experiments before real-world action. Its page says the model should be continuously validated, but gives no forecast-error measures, calibration results, or tests of predicted interventions. 14
A twin that forecasts the consequences of a supplier loss or production change is a high-dimensional conditional forecast. Keeping its data synchronized proves that it reflects recorded operations; it does not prove that an alternate scenario will unfold as simulated. Errors in lead times, substitution, capacity response, or demand can propagate through several decisions. A model that confidently recommends the wrong intervention is worse than a dashboard that merely reports a disruption.
This is also a maintenance problem. Holimization asks whether unexpected operational interactions can be incorporated into the decision logic at an acceptable cost. Atlas’ visual workflows allow rules and processes to change. Public documentation gives much less visibility into how a customer changes the underlying uncertainty model or economic objective when the original formulation proves inadequate. 15
Multi-objective optimization and manufacturing boundaries
The August 2025 release is explicit about its mechanism: users assign weights to competing goals, with hard constraints respected and soft-constraint impacts reduced. Examples include favoring high-margin production, freshness, supplier reliability, and carbon footprint. The release scopes the enhancement to key parts of MEIO and rough-cut capacity planning. 10
This improves on optimizing one inventory measure in isolation. It also transfers a difficult valuation problem to the person choosing weights. What financial loss corresponds to a unit of missed service? How should a carbon preference compare with a product margin? How do penalties change when volume, product mix, or lead times change? An interface that makes weights easy to adjust does not make those answers correct.
The distinction between rough-cut planning and detailed scheduling also matters. The same release promotes fewer data and configuration requirements for companies not ready for detailed scheduling. Reduced modeling effort is purchased by omitting detail. When sequence-dependent changeovers, cleaning, or resource interactions determine feasibility, those omissions can change the best decision.
John Galt’s detailed scheduling offer includes PlanetTogether. The February 2024 partnership expansion and joint brochure describe production sequencing, material and resource constraints, and clean-outs based on time, quantity, cycles, products, and attributes. These are concrete manufacturing capabilities. 23 22
The customer consequently needs agreement between the planning and scheduling models: the same orders, resources, available quantities, calendars, and priorities must survive the interface. A partnership can provide this integration, but the joint brochure does not expose how competing objectives are reconciled or how decisions return after scheduling changes. The engineering boundary remains even when both products are sold as part of one planning proposition.
Automation that changes the work
Reddy Ice’s case explains its previous operating problem. Delivery plans were static for months, spreadsheets required manual updates, some visits arrived too late, and others were unnecessary. Atlas combined driver-reported inventory and delivery data, weather, point-of-sale information, and customer orders. An IoT pilot added icebox readings every 30 minutes. The delivery plan was then recalculated daily for the following day. 4
The vendor reports a 50% reduction in customer stockouts, more than a 20% improvement in delivery productivity, and a 20% increase in sales. The useful mechanism is the alignment between a short decision horizon, new operational observations, and a changeable delivery schedule. This is stronger evidence than a forecast-accuracy percentage detached from the action it informs. The case does not disclose a controlled comparison or complete net costs. 4
The benefit also has a clear boundary. Rapid sensing can change tomorrow’s delivery; it cannot retroactively change a long-lead-time procurement commitment. Calling both tasks demand sensing obscures the difference. For each application, the relevant test is whether a new observation arrives before the costly decision becomes irreversible.
Galt AI’s current product page describes investigation, reasoning, recommendations, and workflow execution. Those verbs cover a wide range of labor outcomes. At LATICRETE, the disclosed result is less navigation and faster access to relevant information, while planners still analyze trade-offs and decisions. No before-and-after measure of total planner hours or autonomous decision volume is published in that announcement. 13 12
John Galt’s 2025 explainability release reinforces this operating model: conversational explanations make inventory and ensemble-forecast recommendations easier for planners to trust and act on. The economic question is whether those reviews become exceptional or simply more comfortable. Adding an explanation to every SKU-level intervention can preserve the largest source of recurring cost. 11
Accuracy claims and evidence of returns
John Galt’s summary of the Nucleus Anatomy of a Decision report promotes a distributor’s forecast accuracy increasing from 70% to 95%, an average 17-percentage-point accuracy improvement, and an 85% team-productivity increase. The page supplies no definition of the accuracy statistic, product/location/time granularity, forecast horizon, or denominator for the productivity figure. Those missing definitions make the headline figures unsuitable for evaluating replenishment decisions or remaining labor. 24
The aggregation problem is quantitatively large. Recalculating the M5 organizers’ top 50 point-forecast submissions using 100 × (1 − WAPE) produces 22.1–25.1% at SKU/store/day resolution, but 95.8–98.2% after pooling products and stores by day. The same forecasts produce both ranges. These are diagnostic calculations on the final 28 days, not M5’s official WRMSSE score. 25
Aggregation allows over- and under-forecasting errors to cancel before their magnitude is measured. It cannot make surplus stock in one location satisfy a shortage elsewhere. John Galt’s 95% claim therefore needs the missing decision-level measurement; it earns no forecasting-performance credit as published.
ForecastX’s own documentation distinguishes top-down forecasting at more stable group levels from the greater item specificity of bottom-up methods. That is a useful acknowledgment of the problem. A forecast’s aggregated appearance and its usefulness for individual replenishments are different properties. 28
Accountability requires similar precision about money. Reddy Ice identifies operational outcomes, but the reviewed materials do not reconcile them with software fees, integration, data collection, model maintenance, or remaining human work. The broad rapid-ROI language in product releases has no corresponding public baseline, evaluation period, or failure criterion. Improved adoption and customer satisfaction cannot fill that economic gap. 4 10
Configuration maintenance and total ownership cost
John Galt describes implementation as configuration with customer participation, joint quality testing, production setup, and a handover to a trained customer team. Post-implementation work includes forecast-lag analysis, training, process-design reviews, and, around months three and four, review of safety stock and inventory policies. The service also supports embedding S&OP into the organization. 17 18
This is a concrete ownership model: the software arrives with logic that is configured, reviewed, and maintained by people. Intelligent Workflows lets users assemble tasks and tune rules on a visual canvas. A later article describes decision histories and traceability, which can help diagnose changes. It does not document version-controlled promotion, automatic regression testing, or rollback of the combined planning model. 15 16
The effort is not removed by calling the interface no-code. A rule that changes an allocation still needs a defined purpose, a test against historical cases, an owner, and a way to detect unintended effects. Multiple weights and thresholds can create conflicting incentives even when each screen looks simple. Staff turnover and vendor releases add further maintenance work.
Atlas’ flexibility can be valuable when corrections become reusable logic. It becomes expensive when corrections accumulate as exceptions that planners must remember. Subscription cost alone captures neither outcome. The buyer needs the initial integration effort, recurring configuration work, vendor support, and routine decision review measured separately.
Bill of software materials and security
Two employer-authored engineering vacancies expose parts of the operating estate. The systems role spans customer provisioning, server sizing, database maintenance, identity migration, and internal business applications. The data-engineering role is explicitly responsible for bidirectional flows between customer ERPs and John Galt software. 20 21
| Evidence scope | Disclosed components or responsibilities | Ownership consequence |
|---|---|---|
| Customer-hosting work in the systems vacancy | OVHcloud, Windows Server, Microsoft SQL Server, SQL Agent and T-SQL; regional deployment and customer placement on servers | Capacity planning, patching, backup, and isolation must be operated as part of the service |
| Integration-engineering vacancy | SQL Server staging and transformations, Python, REST APIs, JSON/XML, Kubernetes jobs; Talend experience preferred | Data mapping and synchronization require maintained code and orchestration |
| Monitoring responsibilities | Zabbix and server-management tooling | Additional infrastructure to configure, secure, and keep operational |
| Atlas Schedule | PlanetTogether scheduling integration | A separate product boundary for resources, orders, priorities, and schedule changes |
| ForecastX | Microsoft Excel and the forecasting add-in | A distinct desktop execution and workbook environment |
These customer-hosting and integration responsibilities identify operating costs behind the composable microservices positioning: server placement, capacity, databases, integration jobs, and monitoring all require ongoing engineering. 19 20 21
SQL staging, transformations, and repeated data exchange can introduce delays or divergent interpretations of inventory and orders. The integration page offers either customer-led or vendor-led implementation and emphasizes cleansing, correction, enrichment, and automation. These steps require maintained semantics, not just compatible file formats. 26
Security disclosure is weak. The systems vacancy mentions permissions, backups, certificate management, and migration from legacy authentication to single sign-on. It does not explain enforced tenant separation, the privileges of integration jobs, or recovery of planning state after a compromised account. The privacy policy describes organizational and technical processes for personal information, with no comparable production-system threat model. 20 27
The integration jobs’ runtime boundary is also undocumented: which identities can read or alter customer data, how secrets and network access are restricted, and what damage remains possible after compromise.
Atlas promotes open APIs, but the developer-documentation link leads to a sign-in page. Public readers can inspect some ForecastX methods and integration descriptions, while the Atlas API contract and planning-engine behavior remain vendor-mediated. That limits independent verification of automation, permissions, and portability. 19
Supply chain score
Each dimension averages its five sub-scores. The overall score averages all 25.
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 6/10 |
| Decision end-state | 8/10 |
| Conceptual sharpness on supply chain | 7/10 |
| Freedom from obsolete doctrinal centerpieces | 3/10 |
| Robustness against KPI theater | 3/10 |
Average: 5.4/10.
Economic framing — 6/10. Purchasing examples include discounts, supplier choice, and margins. Multi-objective planning nevertheless leaves economic arbitration partly in user-selected weights and service policies. 9 10
Decision end-state — 8/10. Reddy Ice documents replenishment quantities, delivery frequency, and routing decisions produced automatically. Atlas extends beyond forecast reporting into operational action. 4
Conceptual sharpness on supply chain — 7/10. Probabilistic lead times, network risk pooling, bills of materials, and sequential decisions address substantive supply chain interactions. Their implementation and validation remain less visible than the concepts. 9
Freedom from obsolete doctrinal centerpieces — 3/10. Service targets, consensus planning, exception handling, and S&OP remain prominent. Post-implementation guidance explicitly reinforces that process model. 3 18
Robustness against KPI theater — 3/10. The promoted 95% accuracy and 85% productivity figures lack definitions needed for operational comparison. Named customer outcomes provide useful evidence, but do not repair these broad, unqualified metrics. 4 24
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 5/10 |
| Distinctive optimization or ML substance | 5/10 |
| Real-world constraint handling | 7/10 |
| Decision automation and labor productivity | 7/10 |
| Resilience under real operational complexity | 5/10 |
Average: 5.8/10.
Probabilistic modeling depth — 5/10. Atlas’ stated scope includes uncertain demand, lead times, costs, and sequential actions. The public guide does not expose calibrated joint distributions or a reproducible decision experiment. 8 9
Distinctive optimization or ML substance — 5/10. Published material identifies dynamic programming, Q-learning, and weighted objectives. These supply a technical direction; they do not establish a distinctive algorithmic advantage or quantify the contribution of each method. 9 10
Real-world constraint handling — 7/10. Procurement minimums and discounts, freshness, production capacity, and detailed cleaning constraints are documented. Integration between rough-cut decisions and the scheduling product remains consequential. 9 10 22
Decision automation and labor productivity — 7/10. Reddy Ice’s repeated deployment accounts substantiate substantial unattended planning. The newer assistant case demonstrates a narrower result and supplies no equivalent total-labor measurement. 4 6 7 12
Resilience under real operational complexity — 5/10. Editable workflows and ongoing customer support provide routes for adaptation. The cost and expressive limits of revising the numerical model, rather than adding another rule, remain unclear. 15 17 18
Product and architecture integrity
| Criterion | Score |
|---|---|
| Architectural coherence | 5/10 |
| System-boundary clarity | 5/10 |
| Security seriousness | 2/10 |
| Software parsimony and ownership cost | 4/10 |
| Compatibility with programmatic and agent-assisted operations | 5/10 |
Average: 4.2/10.
Architectural coherence — 5/10. Atlas has an understandable planning role and modular interface. The scheduling partnership and disclosed integration estate require model and data reconciliation that the common-platform description does not explain. 19 21 23
System-boundary clarity — 5/10. The vendor identifies ERP integration responsibilities and a distinct scheduling partner. Public descriptions stop before the detailed contracts governing planning state and returned decisions. 21 23 26
Security seriousness — 2/10. Recruitment identifies basic operational duties, while the privacy policy offers general process assurances. Neither exposes enforceable containment, agent permissions, or recovery of corrupted operational state. 20 27
Software parsimony and ownership cost — 4/10. Hosting, integration code, orchestration, customer workflows, and ongoing policy reviews all require maintenance. The public rapid-value proposition does not quantify this combined recurring burden. 17 18 20 21
Compatibility with programmatic and agent-assisted operations — 5/10. Automated replenishment and bidirectional integration show that Atlas can operate without repeated screen interaction. The public API documentation remains gated, and the assistant’s deployment story is primarily planner-facing. 4 12 19 21
Technical transparency
| Criterion | Score |
|---|---|
| Public technical documentation | 5/10 |
| Inspectability without vendor mediation | 4/10 |
| Portability and lock-in visibility | 3/10 |
| Implementation-method transparency | 6/10 |
| Evidence density behind technical claims | 5/10 |
Average: 4.6/10.
Public technical documentation — 5/10. ForecastX method instructions and the Atlas AI guide disclose useful mechanisms. Atlas’ probabilistic formulation, calibration, and engine interfaces remain substantially less accessible. 2 3 9
Inspectability without vendor mediation — 4/10. An outsider can examine forecasting settings and named deployment workflows. Reproducing the flagship planning results requires access beyond the public documentation. 2 4 19
Portability and lock-in visibility — 3/10. Integration options are described, but the reviewed material lacks a portable specification for customer rules, objectives, and planning-model configuration. Data exchange alone does not preserve those accumulated assets. 15 26
Implementation-method transparency — 6/10. Joint QA, customer handover, forecast-lag analysis, and scheduled policy reviews make responsibilities more tangible. Pricing and recurring effort for those activities are not published. 17 18
Evidence density behind technical claims — 5/10. The Reddy Ice chronology, technical guide, and engineering roles substantiate several capabilities. The strongest twin, autonomy, and accuracy claims still lack matched validation. 4 9 12 14 24
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 4/10 |
| Technology roadmap integrity | 5/10 |
| Conceptual sharpness | 6/10 |
| Economic accountability | 3/10 |
| Defensibility in an agentic-software world | 6/10 |
Average: 4.8/10.
Technical seriousness of public communication — 4/10. Detailed planning concepts coexist with undefined near-perfect accuracy and expansive autonomy language. Generative explanations are promoted without a disclosed test of their faithfulness to the calculation. 9 11 24
Technology roadmap integrity — 5/10. The Reddy Ice record shows continuity from deployment to repeated automation reports. The recent probabilistic, explainability, and agentic releases need comparable evidence of sustained outcomes. 30 5 7 8 12
Conceptual sharpness — 6/10. John Galt recognizes that supply uncertainty and sequential consequences matter. Its weighted-goal interface and planner-maturity narrative still leave critical economic decisions outside a clearly defined valuation framework. 9 10 18
Economic accountability — 3/10. Operational case results are not accompanied by complete net ownership costs or defined economic failure criteria. Public support commitments describe participation and training more clearly than responsibility for a missed return. 4 17 18
Defensibility in an agentic-software world — 6/10. Maintained integrations, domain-specific numerical logic, and demonstrated unattended replenishment retain value. Assistant-led exception review is more exposed to commoditization and does not establish additional decision-engine productivity. 4 12 21
Sources
Sources were accessed on November 4, 2026. Publication dates are given where available.
[1] John Galt Solutions: Atlas and ForecastX
John Galt Solutions: Atlas and ForecastX. Functional scope of Atlas and ForecastX.
[2] ForecastX Procast method documentation
ForecastX Procast method documentation. Per-series forecasting method selection and error-statistic parameters.
[3] ForecastX safety-stock documentation
ForecastX safety-stock documentation. Lead-time and service-level settings for safety-stock calculations.
Reddy Ice case study. Named deployment, data inputs, next-day replenishment, and reported operational outcomes.
[5] Reddy Ice: machine learning and IoT conference presentation
Reddy Ice: machine learning and IoT conference presentation (2018-05-14). Contemporaneous machine-learning and IoT claims for Reddy Ice.
[6] Reddy Ice: end-to-end supply chain automation
Reddy Ice: end-to-end supply chain automation (2022-06-01). Continued customer automation across more than 82,000 locations.
[7] Reddy Ice automation webinar announcement
Reddy Ice automation webinar announcement (2025-04-02). Follow-up account of Reddy Ice automation.
[8] Introduction of end-to-end probabilistic planning
Introduction of end-to-end probabilistic planning (2024-05-08). Dated introduction of Q-learning and end-to-end probabilistic planning.
[9] Your Guide to AI in Supply Chain Planning
Your Guide to AI in Supply Chain Planning. Vendor technical guide: network risk pooling, dynamic programming, lead-time ranges, purchase orders, and sequential planning.
[10] Simultaneous Multi-Objective Optimization release
Simultaneous Multi-Objective Optimization release (2025-08-12). Weights, hard and soft constraints, and the scope of the enhanced optimization.
[11] Explainable AI for MEIO and ensemble forecasting
Explainable AI for MEIO and ensemble forecasting (2025-09-16). Generative explanations of inventory and ensemble-forecast recommendations.
[12] LATICRETE deploys Galt AI
LATICRETE deploys Galt AI (2026-07-21). Named agent deployment, present assistant uses, and future automation objectives.
Galt AI product page. Current autonomy and workflow-execution positioning.
[14] Digital Supply Chain Twin
Digital Supply Chain Twin. Scenario-model scope and the stated validation approach.
[15] Atlas Intelligent Workflows release
Atlas Intelligent Workflows release (2023-09-13). No-code rules and process composition.
[16] Building Trust with Intelligent Workflows
Building Trust with Intelligent Workflows. Decision history, traceability, and planner interaction.
[17] Implementation Excellence
Implementation Excellence. Configuration, joint QA, production setup, and customer handover.
Post Implementation. Recurring training, forecast analysis, policy reviews, and S&OP support.
[19] Composable and Extensible Atlas
Composable and Extensible Atlas. Microservices, open-API positioning, and the link to gated developer documentation.
[20] Systems and Database Engineer vacancy
Systems and Database Engineer vacancy. Employer-authored vacancy: customer-hosting responsibilities and infrastructure requirements; undated.
Data Engineer vacancy. Employer-authored vacancy: ERP integration, data pipelines, and Kubernetes requirements; undated.
[22] John Galt and PlanetTogether technology partner brochure
John Galt and PlanetTogether technology partner brochure (2024). Joint product brochure, detailed scheduling functions, and manufacturing constraints.
[23] Expanded PlanetTogether partnership
Expanded PlanetTogether partnership (2024-02-27). Scope of the expanded detailed-scheduling partnership.
[24] Nucleus Research: Anatomy of a Decision, vendor summary
Nucleus Research: Anatomy of a Decision, vendor summary. Vendor-promoted accuracy, productivity, and inventory figures; the public summary does not define their measurements.
[25] M5 organizers: fixed revision of data and submissions repository
M5 organizers: fixed revision of data and submissions repository. Organizers’ data and submitted point forecasts at a fixed repository revision. The aggregation comparison uses all 30,490 series over days d_1942–d_1969, including zero-sales observations; forecasts are summed without refitting.
[26] Atlas integration framework
Atlas integration framework. Integration ownership, data transformations, and system connections.
[27] John Galt Solutions Privacy Policy
John Galt Solutions Privacy Policy (2023-06-20). Personal-information policy and general security statements.
[28] ForecastX group forecasting documentation
ForecastX group forecasting documentation. Top-down aggregation, bottom-up forecasting, and allocation behavior.
[29] John Galt Solutions company history
John Galt Solutions company history. Company founding and retrospective product timeline.
Reddy Ice selects Atlas (2015-10-08). Original vendor-issued announcement of Reddy Ice’s Atlas selection.