Go back to Market Research
ThroughPut supplies cloud software for analyzing industrial bottlenecks and recommending changes to inventory, replenishment, production capacity, and logistics. Its applications use existing operational data to identify delays, prioritize corrective actions, and estimate financial opportunities. The company also offers services that deliver analysis through customers’ existing tools. (1, 18)
Overall supply chain score: 4.0/10.
| Dimension | Score |
|---|---|
| Supply chain depth | 4.4/10 |
| Decision and optimization substance | 4.4/10 |
| Product and architecture integrity | 4.0/10 |
| Technical transparency | 3.8/10 |
| Vendor seriousness | 3.4/10 |
ThroughPut’s focus on material flow, downtime, and the financial cost of bottlenecks addresses real operational problems. Its patents also expose considerably more substance than the broad AI language suggests: statistical process-control rules, data preparation, opportunity ranking, and estimates of avoidable losses. These mechanisms can accelerate diagnosis. They do not establish a joint economic optimizer that chooses among uncertain supply chain actions. (2, 3)
The larger weakness is the distance between those mechanisms and the autonomy and return claims. Automated purchasing and rebalancing are advertised, but execution controls and measured operating labor are poorly documented. A historical customer case further illustrates the evidential problem: projected benefits in an older PDF reappear as achieved improvements in a later article, without a new observation period. The resulting proposition is more convincing as operational analysis than as a proven autonomous decision system. (1, 12, 13)
ThroughPut versus Lokad
Consider an industrial business choosing whether to expedite a spare part, transfer one between plants, repair an existing unit, or accept downtime. ThroughPut’s replenishment release exposes batch and stock-level controls. Those controls can make recommendations executable, but do not explain how the software values the competing actions when delivery times, failure rates, and repair outcomes are uncertain. (4)
Lokad’s inventory optimization offering organizes recurring decisions around their expected financial consequences under uncertainty. Supply Chain Scientists construct and maintain the decision logic, including costs and the interpretation of operational data. The practical comparison is whether ThroughPut’s bottleneck analysis and customer settings produce better decisions at a lower lifetime engineering and operating cost than a maintained economic model.
The perspective in Introduction to Supply Chain also challenges treating a bottleneck as a sufficient objective. Removing a delay can increase output that has little value, consume scarce capital, or move the shortage elsewhere. The task is to choose the most valuable feasible action for the business, with the associated risks, rather than maximize a local flow measure. ThroughPut’s emphasis on profit is appropriate; the numerical connection between profit and its recommendations remains the unresolved part.
What the patents actually disclose
The bottleneck-detection patent, US11815882B2, claims priority from September 2018. It separates corrupt records from unusual observations: outliers can be excluded when calculating reference statistics yet retained for detecting abnormal behavior. It then applies statistical-control rules, ranks operational problems, and presents recommendations and history. This is a useful distinction; erasing every exceptional observation would erase much of the phenomenon the application is supposed to diagnose. (2)
The follow-on patent, US12118491B2, granted in October 2024, describes a rapid operational analysis application. A user selects an achievable control parameter, such as a threshold derived from standard deviations. The system sums observations exceeding that threshold to estimate reducible defects or costs. Its example accepts shift times, operating hours, and a unit defect cost. (3)
The central assumption is substantial. Observing that some periods perform better does not establish that every worse period can be improved to the chosen threshold, at an acceptable cost. Product mix, wear, operator availability, and demand may all change together. A sum of excess defects is an opportunity estimate until a feasible intervention, its expense, and its consequences have been evaluated.
These patents document statistical diagnosis and configurable opportunity accounting. Their methods do not supply a validated distribution of future demand and supply, nor a financial ranking of complete purchase, production, and allocation policies. They explain part of the product’s engineering ancestry; current replenishment and pricing capabilities require their own evidence. (2, 3, 4)
Forecasts, constraints, and uncertainty
The demand-sensing page promises to put planning on autopilot using time-stamped sales, purchasing, manufacturing, and distribution records. It advertises a 10% improvement in forecast accuracy and roughly 30% lower inventory without defining the error metric, baseline, item-location resolution, or evaluation interval. “Demand sensing” identifies no reproducible statistical method here. (5)
The launch article, dated January 2021, describes historical sales patterns enriched by holidays, promotions, and external events. Those are familiar forecasting inputs. They can be useful, but their freshness does not eliminate sparse demand, censored sales, long replenishment delays, or uncertain supplier performance. The article supplies no benchmark demonstrating an incremental advantage at the horizon at which orders become irrevocable. (14)
The December 2024 replenishment release is more concrete. Users can set order batches, override recommended safety stock, impose maximum replenishment levels, and stop replenishment after discontinuation. Recommendations use dynamic min-max levels and EOQ/MOQ adjustments. This establishes practical constraints, while leaving the optimization formulation and interactions among them unspecified. A floor on one item and a ceiling on another can consume the same scarce budget without identifying its best use. (4)
Lead-time realism receives explicit attention in the Air Force work. The December 2025 Phase III announcement describes predictive replenishment with realistic lead times and feedback between local operations and enterprise supply. That is relevant to spare parts. A revised lead-time estimate still differs from a calibrated distribution of repair and delivery outcomes coupled to failure demand and downtime costs. The announcement does not disclose that joint model. (17)
Capacity planning recognizes scheduled and unscheduled maintenance, product allocation across machines, and replay of historical data. Scenario analysis is proposed for uncertain demand and resources. The public material does not explain how scenarios receive probabilities, how economic risk changes the chosen schedule, or how the model is revised when a new operational failure exposes a missing constraint. (7)
Digital twins and unexplained success percentages
The site labels its bottleneck offering a supply chain digital twin. The corresponding page describes value-stream mapping, multivariate analysis, scenario comparison, and financial calculators. It does not document a validated dynamic model whose forecasts of interacting process states are checked against later observations. When a twin predicts the effect of interventions, prediction error in those interacting states becomes central; a diagram of the network and a financial calculator cannot establish that predictive fidelity. (6)
The homepage says its AI can set allocation and pricing without underpricing or overpricing more than 97% of the time, and can auto-order, cancel, and rebalance inventory. The 97% is a pricing claim, not SKU demand-forecast accuracy. Its denominator, reference price, tolerance, and counterfactual profit are undefined. A price inside a chosen band can still be economically poor; a frequency of accepted prices does not establish maximum margin. (1)
History: industrial AI, autopilot, and the status of results
The public record follows a consistent bottleneck-management theme, with expanding promises around it.
| Period | Public development | What it establishes |
|---|---|---|
| 2018–2019 | Initial patent priority and promotion of ELI as industrial AI | A statistical-control and operational-analysis foundation, accompanied by claims of automated bottleneck elimination. |
| April 2020 | ELITE launched on Azure Marketplace | A SaaS dashboard and report offering for quality analysis and estimated financial savings. |
| 2021 | Demand-sensing module and autopilot messaging | Forecasting and collaborative planning added to the operational-analysis proposition. |
| 2022 | $6 million angel round; Air Force Phase II and Church Brothers selection | Financing and identifiable engagements, with the Air Force award funding investigation of capability gaps. |
| December 2024 | EOQ/MOQ and inventory-setting release | Specific replenishment controls become publicly identifiable. |
| 2025 | Catalyst services, Aankhen tariff partnership, and further Air Force work | Additional delivery routes and application scope; these announcements do not quantify a common economic result. |
Sources: 2, 23, 11, 14, 15, 16, 21, 4, 18, 19, 17.
The early AI presentation was especially expansive. A 2019 conference announcement associated 95% of unresolved production issues with bottlenecks and portrayed ELI as preventing costly operational problems. Such a broad denominator has little diagnostic value unless the classification and measurement are exposed. The underlying statistical-control approach is narrower and more intelligible than the promise of an artificial operations manager. (23, 12)
The sharper historical concern is how results are presented. A Japanese automaker PDF, whose creation metadata dates to December 2019, describes checking casting-process timing to determine readiness for robotics. Its impact page projects 15%–30% labor-productivity improvement, 10%–20% fewer defects, and $500,000–$1.1 million in estimated annual savings per factory from full automation. (12)
The November 2025 retelling presents the same labor and defect ranges as improvements seen after deployment, while retaining estimated annual savings. It adds no measurement dates, comparison protocol, or account of the subsequent robotics investment. The presentation upgrades prospective benefits to operational results without supplying the evidence needed for that upgrade. It also leaves the contribution of the diagnostic software entangled with that of physically automating production. (13)
This is more consequential than a fashionable label. A vendor asking customers to trust financial recommendations must preserve the distinction between an opportunity calculation, an implemented intervention, and its measured net return.
Customer evidence and economic accountability
Church Brothers’ October 2022 selection concerned demand forecasting and profitability analysis for fresh produce. The current testimonial describes generating a weekly baseline extending a year and a half, followed by adjustments for information absent from the data. This substantiates a forecasting workflow with human input. It does not quantify realized waste reduction, margin improvement, or autonomous purchasing. (21, 1)
The Air Force sequence supplies useful follow-through. The October 2022 announcement specified a $749,971 Phase II award to investigate supply-chain capability gaps. In December 2025, ThroughPut announced a second Phase III contract for aircraft availability at Davis-Monthan. Continued procurement is evidence of an ongoing engagement. The disclosed material gives no measured availability improvement, avoided downtime, or total cost against a baseline. (16, 17)
The website encourages physical validation of improvements, a sound principle. Its public commercial material does not define the economic result that would count as failure, the customer obligations required to reach it, or responsibility for remediation. The privacy policy refers to a separate subscription agreement; those references do not expose its economic commitments. Advertised returns consequently lack published acceptance criteria that would make the economic proposition testable. (1, 22)
Productivity and the cost of maintaining decisions
ThroughPut has a plausible productivity proposition in replacing repeated data collection and statistical reporting. The automaker case concerns exactly that work. Eliminating manual analysis, however, leaves the work of validating causes, choosing interventions, and implementing changes. The published cases do not count that remaining labor or show how it scales with the number of recurring decisions. (12, 13)
The Catalyst program makes the service component explicit. Launched in June 2025, it assigns data preparation and synchronization to ThroughPut, runs the vendor’s analysis behind the scenes, and delivers reports and recommendations through familiar tools. This can reduce a customer’s initial software project. It also makes the recurring service workload part of the total cost; retaining existing tools does not remove the work of keeping their data interpretations aligned. (18)
Replenishment settings are another maintenance obligation. Item-location floors, ceilings, batches, and end-of-life dates encode business decisions that must change as suppliers, products, and costs change. An override can address today’s exception while preserving tomorrow’s error. The release documentation does not expose versioned rule changes, regression tests, or a replay-and-rollback process for those settings. (4)
The July 2025 Aankhen partnership identifies a further boundary: fiChains is to supply SKU-level financial costs, while ThroughPut uses them for inventory, pricing, and profitability recommendations. Aankhen’s current site confirms the joint offering and its own procurement-logistics-finance scope. The integration must reconcile cost timing, SKU identity, and changes to tariffs. Public descriptions do not show how those semantics are tested or who repairs discrepancies between the two products. (19, 20)
Improvement productivity ultimately requires a feedback loop from a bad operational result to a changed model, followed by a test that the revision improves decisions elsewhere. Historical charts and faster recommendations can help that work. They do not substitute for it. This is the distinction developed in Holimization: the formulation itself is a maintained artifact, not a permanently correct problem handed to a solver.
Software materials and security
ThroughPut’s engineering jobs and older implementation descriptions expose a small portion of the architecture:
| Component | Role and evidence | Scope |
|---|---|---|
| React and JavaScript | Client-facing application and dashboard | Explicit duties in currently accessible, undated developer postings. |
| Python and Django | Application backend | Explicit full-stack developer responsibilities. |
| Third-party APIs | Dashboard and backend integration | Developer duties include building integrations; the front-end role describes multi-tenant applications. |
| Microsoft Azure | ELITE hosting and distribution | Documented in the April 2020 launch; current detailed cloud topology is undisclosed. |
| Embedded Power BI template | Dashboard cloned for an uploaded dataset | Implementation described in the rapid-analysis patent, rather than a current inventory of every product module. |
The data-scientist role also assigns production-model work jointly with engineering, DevOps, and QA, plus client data troubleshooting. Those responsibilities make the maintenance work tangible. The public record does not identify the production database, queue, orchestration layer, or their replacement history. (10)
Software parsimony remains an ownership question: ingestion, analytical copies, generated dashboards, and partner cost feeds require synchronization, monitoring, and upgrades. The available architecture descriptions do not quantify those costs or demonstrate redundant heavyweight systems. The clearer cost exposure lies in the recurring data and configuration work.
Security is poorly specified. A multi-tenant application needs enforced separation of data and credentials, restricted connector privileges, controlled production changes, and recoverable state. The jobs and public privacy policy describe neither those boundaries nor a tested recovery design. This is particularly material when the product is promoted as able to create or cancel orders. An automatic decision needs an equally explicit limit on what a compromised account or connector can do. (9, 22, 1)
Supply chain score
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 6/10 |
| Decision end-state | 5/10 |
| Conceptual sharpness on supply chain | 5/10 |
| Freedom from obsolete doctrinal centerpieces | 4/10 |
| Robustness against KPI theater | 2/10 |
Average: 4.4/10.
Economic framing — 6/10. ThroughPut connects process losses to costs and directs attention toward working capital and downtime. Its opportunity accounting nevertheless relies on achievable-performance assumptions rather than a complete financial comparison of feasible actions. (3)
Decision end-state — 5/10. Replenishment and product-allocation recommendations address concrete operational actions. Automatic execution is advertised without sufficient documentation of approvals, transaction reconciliation, or sustained decision quality. (4, 1)
Conceptual sharpness on supply chain — 5/10. The emphasis on flow and realistic lead times is useful. The missing connection is between identifying a constraint and valuing the alternative responses under demand, delivery, and failure uncertainty. (7, 17)
Freedom from obsolete doctrinal centerpieces — 4/10. Bottleneck diagnosis gives the product focus, but an excessive emphasis on control limits and throughput can substitute local stability for system-wide economic value. The scenario and allocation machinery remains only partly exposed. (2, 7)
Robustness against KPI theater — 2/10. Forecast improvement, pricing success, and operational benefit percentages lack essential definitions. The automaker retelling further blurs projected and achieved results, weakening the financial interpretation of the figures. (5, 1, 12, 13)
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 3/10 |
| Distinctive optimization or ML substance | 5/10 |
| Real-world constraint handling | 5/10 |
| Decision automation and labor productivity | 4/10 |
| Resilience under real operational complexity | 5/10 |
Average: 4.4/10.
Probabilistic modeling depth — 3/10. Control-chart statistics and revised lead times recognize variability. They do not establish calibrated joint projections of future demand and supply that drive risk-adjusted action choices. (2, 17)
Distinctive optimization or ML substance — 5/10. The patents expose a real operational-analysis method and a useful treatment of exceptional records. The claimed financial and autonomous scope exceeds the numerical methods and comparative validation publicly demonstrated. (2, 3)
Real-world constraint handling — 5/10. Published replenishment settings cover practical ordering restrictions, and capacity planning includes resource allocation. Their joint treatment and the economic effects of conflicting settings are unspecified. (4, 7)
Decision automation and labor productivity — 4/10. Automating statistical analysis can remove repetitive work. Available customer and service descriptions do not measure the full remaining effort needed to approve, implement, and maintain operational decisions. (12, 18)
Resilience under real operational complexity — 5/10. Historical analysis and operational feedback can reveal failures and shifting constraints. The missing step is an inspectable process for revising the decision formulation and retesting it before future recommendations are changed. (7, 17)
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 | 4/10 |
Average: 4.0/10.
Architectural coherence — 5/10. Operational analysis forms a recognizable common core, supported by an identifiable web application. The relationship between historical analytical implementations and the current pricing and replenishment modules remains incompletely documented. (3, 8, 4)
System-boundary clarity — 5/10. Catalyst and the fiChains partnership identify significant service and financial-data boundaries. Production writeback, authoritative state, and reconciliation across those boundaries receive little public specification. (18, 19)
Security seriousness — 2/10. The public material identifies a multi-tenant application but supplies no inspectable isolation, privilege, or recovery design. This leaves the consequences of a compromised connector particularly uncertain for advertised automatic ordering. (9, 22, 1)
Software parsimony and ownership cost — 4/10. The visible application components do not establish infrastructure bloat. Data engineering, partner reconciliation, and customer rule maintenance nevertheless impose recurring work whose scope and cost are not quantified. (8, 18, 19)
Compatibility with programmatic and agent-assisted operations — 4/10. APIs and backend integration support programmatic data exchange. Public operation contracts and controls for unattended changes are too sparse to establish a dependable autonomous operating model. (8, 9)
Technical transparency
| Criterion | Score |
|---|---|
| Public technical documentation | 5/10 |
| Inspectability without vendor mediation | 3/10 |
| Portability and lock-in visibility | 3/10 |
| Implementation-method transparency | 5/10 |
| Evidence density behind technical claims | 3/10 |
Average: 3.8/10.
Public technical documentation — 5/10. The patent disclosures provide algorithms and examples, and product releases describe selected operational controls. Current deployment documentation, interface contracts, and model-validation records are much thinner. (2, 3, 4)
Inspectability without vendor mediation — 3/10. A reader can inspect historical methods and some application roles, but cannot reproduce the current forecasting, replenishment, or pricing calculation from the public material. Most product assessment still depends on vendor interaction. (2, 5, 4)
Portability and lock-in visibility — 3/10. Existing-data ingestion reduces the need to replace transactional systems. The portability of configured rules, analytical models, and execution history is not established by those input connectors. (6, 18)
Implementation-method transparency — 5/10. The patent examples and Catalyst steps show how data preparation becomes analysis and a report. The transition from that implementation to continuous autonomous execution lacks comparable detail. (3, 18)
Evidence density behind technical claims — 3/10. Named engagements and continuing Air Force work establish concrete application contexts. Quantified numerical comparisons, measured operating labor, and economically attributable outcomes remain sparse. (21, 16, 17)
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 3/10 |
| Technology roadmap integrity | 3/10 |
| Conceptual sharpness | 5/10 |
| Economic accountability | 2/10 |
| Defensibility in an agentic-software world | 4/10 |
Average: 3.4/10.
Technical seriousness of public communication — 3/10. Inspectable statistical methods sit alongside expansive autopilot, digital-twin, and success-rate language. The latter claims lack the definitions and validation needed to make them technically useful. (2, 5, 6)
Technology roadmap integrity — 3/10. The record contains actual replenishment releases and continued engagements. Its credibility is weakened by the automaker case changing the status of benefits without supplying a corresponding measurement record. (4, 12, 13)
Conceptual sharpness — 5/10. Material flow, downtime, and financial consequences are a coherent focus. Statistical stability and a bottleneck ranking remain insufficient substitutes for an explicit economic policy under uncertainty. (2, 3)
Economic accountability — 2/10. Projected savings and named projects are not tied to public net-return acceptance criteria or a defined remediation commitment. Buyers cannot determine from the published proposition what economic failure would oblige the vendor to correct. (12, 13, 22)
Defensibility in an agentic-software world — 4/10. Industrial data interpretation and operational experience retain value. The documented offer still concentrates on accelerating analysis and supplying recommendations, with little measured evidence of eliminating the recurring work of running the decision process. (18, 10)
Sources
[1] Platform and automation claims
Source. Accessed November 4, 2026.
[2] Industrial bottleneck detection patent US11815882B2
Source, 2023-11-14. Accessed November 4, 2026.
[3] Rapid operational analysis patent US12118491B2
Source, 2024-10-15. Accessed November 4, 2026.
[4] EOQ, MOQ, and replenishment settings
Source, 2024-12-05. Accessed November 4, 2026.
Source. Accessed November 4, 2026.
[6] Digital twin and bottleneck detection
Source. Accessed November 4, 2026.
Source. Accessed November 4, 2026.
[8] Full-stack developer responsibilities
Source. Accessed November 4, 2026.
[9] Front-end developer responsibilities
Source. Accessed November 4, 2026.
[10] Supply chain data scientist responsibilities
Source. Accessed November 4, 2026.
[11] ELITE on Azure Marketplace
Source, 2020-04-28. Accessed November 4, 2026.
[12] Japanese automaker case-study PDF
Source, 2019-12-19 (PDF creation metadata). Accessed November 4, 2026.
[13] Japanese automaker case-study retelling
Source, 2025-11-17. Accessed November 4, 2026.
[14] Demand sensing launch article
Source, 2021-01-29 (page date; subsequently edited). Accessed November 4, 2026.
[15] Angel funding announcement
Source, 2022-04-21. Accessed November 4, 2026.
[16] Air Force Phase II announcement
Source, 2022-10-31. Accessed November 4, 2026.
[17] Air Force Phase III announcement
Source, 2025-12-09. Accessed November 4, 2026.
[18] Catalyst services program
Source, 2025-06-05 (release dateline). Accessed November 4, 2026.
[19] Aankhen tariff partnership
Source, 2025-07-14 (release dateline). Accessed November 4, 2026.
[20] Aankhen fiChains platform
Source. Accessed November 4, 2026.
[21] Church Brothers selection
Source, 2022-10-25. Accessed November 4, 2026.
[22] Privacy policy and linked contract framework
Source. Accessed November 4, 2026.
[23] Air Cargo Conference announcement
Source, 2019-09-19. Accessed November 4, 2026.