Go back to Market Research
Ganacos is a cloud planning platform for demand forecasting, S&OP, production and distribution planning, scheduling, replenishment, and financial planning. Customers adapt shared data models and formulas, compare scenarios, and prepare orders for their ERP systems. The French vendor serves manufacturers, retailers, and other organizations replacing spreadsheet-based planning. (1, 4)
Overall supply chain score: 4.4/10
| Dimension | Score |
|---|---|
| Supply chain depth | 4.2/10 |
| Decision and optimization substance | 4.0/10 |
| Product and architecture integrity | 4.6/10 |
| Technical transparency | 4.2/10 |
| Vendor seriousness | 4.8/10 |
Ganacos organizes planning around a shared forecast, configurable business rules, and human arbitration. Its replenishment application calculates requirements and safety stock; its scheduling interface lets users resolve competing priorities; its financial application values alternative plans. The missing connection is an explicit calculation that chooses actions by their expected economic consequences under uncertain demand, delivery dates, and production outcomes. Connecting departments to the same model does not supply that decision logic.
The platform’s strongest engineering proposition is the ability to revise that model. Keeping software aligned with changing operations is a consequential cost. Ganacos nevertheless supplies little public material on testing, deploying, and recovering changes. Customer testimony describes continued work with consultants and knowledgeable administrators. Model flexibility therefore comes with a continuing maintenance obligation. (7, 8, 12)
Ganacos versus Lokad
Ganacos helps an organization construct and operate its planning process. Lokad’s quantitative supply chain approach centers on recurring decisions: calculate which purchases, production orders, and allocations have the highest expected economic value, then improve that calculation as operational evidence reveals mistakes.
Consider a manufacturer deciding whether to order a shared ingredient before a supplier’s uncertain delivery window. The purchase affects several finished products, consumes cash, and may expire before use. Ganacos can represent the material relationships, compare capacity and inventory projections, and display financial consequences. Its documented order calculations use forecast requirements, replenishment parameters, and approval workflows. They do not establish a joint probabilistic assessment of the ingredient’s arrival, competing demands, waste, and forgone alternatives. (3, 4, 6)
Lokad’s inventory optimization documentation describes probabilistic demand and lead times combined with customer-specific economic priorities. Supply Chain Scientists maintain the data interpretations, numerical logic, monitoring, and production operation. The objective is to make a better purchase calculation reusable across the assortment, with limited recurring review. Ganacos distributes more of the modeling and arbitration work among customer teams, its consultants, and integrators.
Both approaches require maintenance. The useful comparison is the cost of incorporating a newly discovered interaction, validating the resulting decisions, and keeping them running. Condor could reduce Ganacos’s authoring burden. It has not established that the resulting models make economically better choices or require fewer operating interventions. Lokad’s service cost and delivered results belong in the same assessment; its design perspective, developed in Introduction to Supply Chain, is a proposition to test through actual decisions.
From forecasts to orders
Demand Planning cleans historical observations, selects a forecasting model for each item, and combines the statistical result with commercial input. Product changes and promotions enter the process, while users review deviations and reconcile views across organizational levels. The published description ends with a consensus demand plan. It does not specify predictive distributions or a calibrated representation of future supplier performance. (2)
That distinction matters when the forecast feeds purchasing. Ganacos describes net-requirements calculations, order proposals respecting minimum quantities and packaging multiples, and approval before transmission to ERP. Inventory policy uses ABC/XYZ classes and coverage measures. These mechanisms can replace repeated spreadsheet calculations, but their parameters determine how much inventory the company buys. A classification by turnover and variability cannot, on its own, price the damage from a shortage against expiration, lost margin, and scarce cash. (3)
Production Planning propagates requirements through bills of materials and warehouses, compares resource loads with capacity, and exposes bottlenecks. Scheduling adds routings, changeovers, and interactive sequencing. These are concrete operational relationships. The published workflow still leaves priority resolution and rescheduling to users, without disclosing an objective function or search method for the complete economic problem. (4, 5)
Finance sees budgets, margins, cash projections, and scenario effects from demand or input-cost changes. That makes the consequences of a selected plan more visible. Choosing the plan remains a separate task: knowing the margin of three scenarios does not determine which of the many feasible purchase and production combinations is preferable. (6)
Accuracy depends on the level at which it is measured
The December 2024 executive interview describes Ganacos AI as selecting statistical methods across product hierarchies and aggregation levels. The 2026 S&OP white paper continues that approach. Neither supplies a reproducible evaluation showing the error at the SKU, location, and decision horizon after the selected result has been distributed back down. (13, 17)
Selecting a useful aggregation level is a modeling technique. Reporting its smoother aggregate error as purchasing accuracy would be misleading. A family can be forecast accurately while its constituent items are consistently overstocked and unavailable. Evaluation must preserve the information available at each forecast date, include the effects of disaggregation and commercial overrides, and compare the resulting decisions with a simple baseline.
The missing validation includes a loss function, held-out periods, treatment of intermittent demand, and results at the resolution where orders are placed. A model selected for forecast error may also be wrong for the economic decision: small errors on costly, scarce materials can matter more than large errors on inexpensive items.
Condor: model building, operating work, and permission boundaries
Condor has a more consequential role than answering questions about inventory. Its product page describes creating and changing tables, attributes, formulas, and integration mappings. It can also explain inherited models and trace a displayed value through its calculation chain. These capabilities address the effort required to understand and revise customer software. (8)
The distinction between present functionality and future functionality is explicit. Condor’s recommendation of actions and simulation of their consequences is still under development. The planning examples inspect coverage and identify threatened items. The same page offers access through an MCP server for external assistants, including ChatGPT and Claude. (8)
An assistant that reproduces an exception-review queue leaves the organization dependent on that queue. The stronger opportunity is to turn recurring explanations and corrections into tested calculations that work across all relevant items. Editing a formula is progress toward this outcome, but successful text generation is only the start: the revised model must preserve units, dependencies, financial meaning, and production behavior. Ganacos has not published measurements of maintained applications per specialist, regression effort, or recurring planning labor after Condor adoption.
The 2026 award submission describes user-scoped access, source traceability, and confirmation before modifications. Those controls identify who may act and make changes visible. They do not establish that a proposed change is safe. An authorized user can approve a faulty formula or export mapping. Inspectable isolation, restricted destinations, reproducible tests, and recovery of production state are needed to contain that failure. (18)
Data residency and external assistants
Ganacos says planning data remains in France, passes through no provider outside the EU, and is separated between customers. It offers a subprocessor list on request and promises usable data return. The current product is SaaS; deployment on customer infrastructure is reserved for defense organizations. (9)
The external-assistant option creates an unresolved processing boundary. An assistant answering questions about planning data must receive enough information to construct its answer. The MCP description does not explain which results leave Ganacos, where the language model processes them, what it retains, or which deployment conditions preserve the residency promise. Keeping the database in France does not establish that every downstream computation stays there.
This is also a security issue beyond residency. Model-building permissions and integration permissions can give an assistant considerable influence over production behavior. The public material does not document a runtime threat model, enforced network restrictions, or a recovery exercise covering mistaken or compromised agent actions. Corporate ownership and an ongoing ISO certification process do not answer those engineering questions. (9)
Configuration maintenance and total ownership cost
Ganacos directly acknowledges that configuring tables, formulas, flows, and rules consumes much of a planning project. Its six applications supply a reusable foundation, which customers, Ganacos staff, or partners adapt. The FAQ puts preconfiguration at 70% and distinguishes a first operational scope in weeks from a larger deployment. These are sensible ways to reduce repeated implementation work. (1, 10)
The continuing cost lies in the customer-specific part. A new site can change allocation rules, units, lead-time interpretations, capacity relationships, and integrations simultaneously. A readable formula helps locate the affected logic; it does not prove that the new result is coherent. Model owners need representative test cases and a way to compare decisions before releasing changes.
Customer accounts make this workload tangible. Showroomprivé reports a first module in 20 days and appreciates a dedicated consultant for subsequent extensions. A global supply chain manager describes the need for a knowledgeable administrator and warns that flexibility can create complexity. Terrena’s scheduling testimony emphasizes joint specification work and the difficulty caused by changes of interlocutor. These are recurring ownership tasks, even when the application is pleasant to use. (12)
Ganacos’s own FAQ says adding sites or changing rules need not require a provider. That is an authoring capability, not a measured staffing outcome. Ownership cost includes customer modeling time, consultant support, regression work, and the delay between an operational discovery and a reliable correction. The reviewed public material does not supply a complete comparison of these costs before and after deployment. (10)
The published audit trail and plan versioning help explain who changed information. Full application maintenance requires more: versioned structure and formulas, controlled promotion, repeatable testing, and recovery of a compatible model and dataset. Public product descriptions do not make that lifecycle inspectable. Consequently, the buyer cannot yet estimate how expensive a major correction will be after several years of local customization.
Software materials and operating cost
Ganacos has a recognizable calculation core rather than a public catalogue of loosely connected planning acquisitions. The available sources identify the following components and boundaries.
| Component | Documented role and scope |
|---|---|
| Chulengo and multidimensional calculation | Current proprietary time-series language and OLAP-style model engine. (7) |
| Java, MySQL, AngularJS, JavaScript/TypeScript, JUnit | Ganacos’s undated employer profile identifies these as its application stack and testing tooling. (14) |
| Java/Spring Boot, Angular, Scala, PostgreSQL, Ebean | A developer’s account attributes these technologies to Ganacos work during April–July 2022. (15) |
| French SaaS hosting | Current hosting commitment; provider and detailed service topology are not publicly identified. (9) |
| OpenID authentication | Current engine description connects corporate identity to application access and data-scope permissions. (7) |
| Condor and MCP | Current model-building and assistant interface; the model provider and inference deployment are not specified publicly. (8) |
The database descriptions concern different periods; the current persistence layer and any migration remain unspecified. Model working memory, backups, and calculation state duplicated across users or scenarios also determine operating cost. Those quantities need to be measured on the customer’s application.
Ganacos describes rapid recalculation over hundreds of millions of values. Its published account does not specify the populated dimensions, hardware, concurrent workload, or worst-case dependency propagation. A responsive model during a meeting can still require expensive memory or slow substantially as more relationships are added. Customer reports of occasional slowness reinforce the need to measure an actual application rather than a cell-count headline. (7, 12)
The proprietary language can reduce repeated formulas and give the vendor control over execution. Its public semantics, resource limits, and security boundaries remain thinly documented. Exporting data also leaves a second exit cost: recreating the logic and interfaces elsewhere. Ganacos’s data-return commitment does not explain how another engine would execute a Chulengo model.
Technology history: a consistent planning product, an unproven agent transition
Ganacos describes itself as director-owned, profitable, and self-funded. Its current chronology places the company’s creation in Nantes in 2016, followed by an in-memory engine, S&OP deployment, additional planning applications, and initial AI agents in 2025. This is the vendor’s retrospective account of its development. (11)
| Period | Public evidence | What the subsequent record establishes |
|---|---|---|
| 2016–2022 | The current chronology traces the calculation core through demand, supply, scheduling, and master production planning. (11) | Today’s applications retain that configurable planning structure. |
| December 2024 | An executive interview identifies Ganacos AI with statistical model and aggregation-level selection. It also describes continuing support as customer needs evolve. (13) | Current forecasting materials retain automatic model selection. No published comparison establishes its incremental economic return. |
| October 2025 | Symrise Pet Food and Bretagne Supply Chain organize a visit and webinar around the deployed S&OP process. (16) | Customer adoption concerns process coordination and planning; the event does not supply measured gains from autonomous decisions. |
| 2026 | The S&OP white paper reports broader named deployments, while Condor’s award submission describes model construction and conversational assistance. (17, 18) | The agent extends the authoring interface. Economic action recommendation remains a stated development item. |
This record supports continuity in the underlying product. The in-memory architecture remains an identifiable implementation choice, and the earlier AI description specifies a narrow forecasting function. Its unresolved promise concerns the value of automatic forecasting relative to simpler methods and the work required to maintain it.
The more relevant concern is whether the new AI positioning changes the economics of the established operating model. The 2026 white paper still builds decisions around category, demand, supply, reconciliation, and executive reviews. Its Symrise example describes about 30 plants and 100 users sharing planning information; Carrefour’s example describes a six-month European S&OP deployment. These are meaningful operating deployments, but their breadth does not measure labor eliminated or incremental profit after software and support costs. (17)
Condor should therefore be judged against a concrete transition: how much faster can the organization turn a faulty replenishment rule into a tested replacement, and how many subsequent decisions can run without inspection? The published record has not answered that question.
Economic accountability
Ganacos’s company page makes adoption the central failure test: a rejected planning tool sends users back to spreadsheets. Adoption matters commercially, but an extensively used system can still recommend poor purchases or consume more skilled labor than its financial benefits justify. A successful interface rollout is an insufficient definition of economic success. (11)
The FAQ identifies gains in consolidation time, forecast accuracy, and anticipation. It does not define a net-return baseline, an evaluation period, or responsibility for correcting an economically disappointing deployment. The available customer accounts support useful coordination and quicker initial implementation; they do not establish a general economic return. (10, 12)
Accountability needs to follow control. Ganacos owns the calculation platform; customers and integrators may own specific formulas, priorities, and overrides. A credible evaluation would retain those versions, attribute resulting orders, and agree who diagnoses failures and maintains corrections. Publicly assigning a project contact helps implementation, but leaves this economic division of responsibility unspecified.
Supply chain score
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 4/10 |
| Decision end-state | 4/10 |
| Conceptual sharpness on supply chain | 5/10 |
| Freedom from obsolete doctrinal centerpieces | 2/10 |
| Robustness against KPI theater | 6/10 |
Average: 4.2/10.
Economic framing — 4/10. Financial visibility spans operating plans and budgets, but the documented purchasing process does not choose quantities through an expected monetary objective. A decision can satisfy its inventory policy while destroying value elsewhere. (3, 6)
Decision end-state — 4/10. Orders can reach ERP after approval. The output is actionable, but recurring arbitration remains part of the operating model, and the method for selecting economically preferable actions is unspecified. (4)
Conceptual sharpness on supply chain — 5/10. The applications address material flows, resource capacities, product transitions, and shop-floor sequencing. Uncertain supply and interactions among competing uses of inventory receive much less technical treatment. (3, 4, 5)
Freedom from obsolete doctrinal centerpieces — 2/10. Consensus forecasting, policy classes, and monthly reviews govern the design. These mechanisms retain a substantial human coordination workload and do not resolve the economic limitations of point-plan replenishment. (17, 3)
Robustness against KPI theater — 6/10. Forecast-error tracking provides a basis for challenging predictions. The published evaluation remains insufficiently specified to connect model selection with SKU-level decisions and economic consequences. (2, 13)
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 2/10 |
| Distinctive optimization or ML substance | 4/10 |
| Real-world constraint handling | 6/10 |
| Decision automation and labor productivity | 4/10 |
| Resilience under real operational complexity | 4/10 |
Average: 4.0/10.
Probabilistic modeling depth — 2/10. The public record supplies neither calibrated demand distributions nor a joint treatment of delivery uncertainty. Scenario comparison leaves the likelihood and consequences of the unchosen futures unresolved. (2, 3)
Distinctive optimization or ML substance — 4/10. A proprietary calculation language and automated model selection are identifiable engineering contributions. The underlying forecasting methods, objective functions, and evaluation results remain too opaque for a stronger numerical judgment. (7, 13)
Real-world constraint handling — 6/10. Material relationships, order-size restrictions, resource loads, and changeovers cover practical constraints. Public material does not explain how the engine resolves economically conflicting constraints across the full network. (3, 4, 5)
Decision automation and labor productivity — 4/10. Shared calculations and order proposals remove clerical work. Named users report time savings, but comparable workload, remaining staff effort, and improvements attributable to Condor are not quantified. (12, 8)
Resilience under real operational complexity — 4/10. Editable models allow a customer to incorporate new discoveries. The missing evidence concerns validation: how an economically significant correction becomes reliable production behavior without accumulating further manual work. (1, 10)
Product and architecture integrity
| Criterion | Score |
|---|---|
| Architectural coherence | 6/10 |
| System-boundary clarity | 5/10 |
| Security seriousness | 3/10 |
| Software parsimony and ownership cost | 4/10 |
| Compatibility with programmatic and agent-assisted operations | 5/10 |
Average: 4.6/10.
Architectural coherence — 6/10. The applications share a recognizable modeling and calculation core. Public information stops short of a deployment architecture that exposes persistence, compute isolation, and model lifecycle. (7, 14, 15)
System-boundary clarity — 5/10. Planning calculations and ERP execution have an intelligible separation. External-assistant processing introduces an unresolved boundary into the otherwise categorical residency narrative. (4, 8, 9)
Security seriousness — 3/10. Scoped permissions and approval of agent changes provide useful controls. Published material does not show containment of unsafe authorized changes, compromised assistant inputs, or recovery of damaged production logic. (7, 8, 18)
Software parsimony and ownership cost — 4/10. Reusable application foundations reduce initial construction. Customer-specific logic, specialist support, and incomplete lifecycle documentation leave substantial ongoing costs difficult to assess. (1, 10, 12)
Compatibility with programmatic and agent-assisted operations — 5/10. Formula-based models, interfaces, and agent-assisted editing create useful automation surfaces. Reproducible model deployment, automated testing, and unattended economic decision operation remain poorly documented. (7, 8)
Technical transparency
| Criterion | Score |
|---|---|
| Public technical documentation | 4/10 |
| Inspectability without vendor mediation | 4/10 |
| Portability and lock-in visibility | 3/10 |
| Implementation-method transparency | 6/10 |
| Evidence density behind technical claims | 4/10 |
Average: 4.2/10.
Public technical documentation — 4/10. Ganacos explains its workflow and calculation abstractions with concrete examples. It does not expose a sufficiently detailed language reference, numerical-method description, or operational manual. (7, 2)
Inspectability without vendor mediation — 4/10. Public screens and Condor examples make several interactions understandable. Reproducing a forecast, testing a model correction, or inspecting permission enforcement requires access beyond the public material. (8, 18)
Portability and lock-in visibility — 3/10. Usable data return is promised. The cost of transferring formulas, interfaces, and execution semantics out of the proprietary engine remains unspecified. (9)
Implementation-method transparency — 6/10. The distinction between reusable foundations and customer configuration is clear, with named customer accounts of initial delivery and later adaptation. Ongoing testing and release responsibilities need comparable detail. (1, 12, 19)
Evidence density behind technical claims — 4/10. The current product descriptions, executive interview, and developer evidence establish concrete mechanisms. Claims of responsiveness, forecasting quality, and agent productivity lack reproducible measurements. (13, 15, 8)
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 5/10 |
| Technology roadmap integrity | 6/10 |
| Conceptual sharpness | 4/10 |
| Economic accountability | 3/10 |
| Defensibility in an agentic-software world | 6/10 |
Average: 4.8/10.
Technical seriousness of public communication — 5/10. Ganacos identifies the calculation core and marks a consequential agent capability as unfinished. Broad speed, sovereignty, and low-maintenance assertions require more precise engineering boundaries. (7, 8, 9, 10)
Technology roadmap integrity — 6/10. The available history shows sustained development of a configurable planning product and continuing customer use. The agent transition has yet to produce documented economic or maintenance outcomes. (11, 16, 18)
Conceptual sharpness — 4/10. The vendor recognizes configuration as a major source of cost. Its governing supply chain argument still relies heavily on coordination and shared forecasts, leaving decision economics underdeveloped. (10, 17)
Economic accountability — 3/10. Adoption and customer satisfaction have identifiable owners. Public commitments do not define economic failure, net returns, or responsibility for remediating underperforming decision logic. (11, 12)
Defensibility in an agentic-software world — 6/10. An editable numerical model gives an agent something reusable to improve, and Condor addresses that opportunity. Its defensibility depends on maintaining coherent, tested applications at lower total cost; conversational planning alone will not provide it. (7, 8)
The overall score is the arithmetic mean of these 25 sub-scores: 4.36/10, rounded to 4.4/10. These are editorial judgments of the documented technology and operating model.
Sources
All sources were accessed on November 4, 2026.
The Ganacos Suite. Undated. Preconfigured applications and customer-specific modeling responsibilities.
Demand Planning. Undated. Statistical model selection, history cleansing, and collaborative demand planning.
[3] Inventory and Supply Management
Inventory and Supply Management. Undated. Net requirements, order approval, ABC/XYZ safety stock, and inventory scenarios.
MPS / DRP. Undated. Production requirements, warehouse propagation, and the ERP return path.
Production Scheduling. Undated. Resource sequencing, changeovers, and manual rescheduling.
FP&A. Undated. Financial valuation, reporting, and scenario comparison.
The Ganacos engine. Undated. Chulengo, the multidimensional engine, interfaces, and authentication.
Condor AI agent. Undated. Model-building actions, MCP access, and action recommendation still under development.
Data sovereignty. Undated. Planning-data residency, hosting scope, customer separation, and data return.
[10] Frequently asked questions
Frequently asked questions. Undated. Implementation scope, ongoing model changes, and the stated sources of ROI.
About Ganacos. Undated. Ownership statements, retrospective chronology, and adoption as the stated failure test.
[12] Customer stories and testimonials
Customer stories and testimonials. Undated. Attributed customer accounts of implementation, support, complexity, and performance.
[13] Ganacos executive interview, Supply Chain Magazine no. 74
Ganacos executive interview, Supply Chain Magazine no. 74. Published 2024-12. Contemporary executive statements about forecasting methods and continuing customer support; page 45.
[14] Ganacos employer engineering profile
Ganacos employer engineering profile. Undated. Employer-published application technologies.
[15] Developer account of Ganacos work, April–July 2022
Developer account of Ganacos work, April–July 2022. Undated. First-person technical account explicitly scoped to April–July 2022 work.
[16] Symrise Pet Food S&OP visit and webinar
Symrise Pet Food S&OP visit and webinar. Published 2025-10-07. Organizer’s announcement of the October 7, 2025 customer event.
Ganacos S&OP white paper. Published 2026. Publicly offered white paper, copyright 2026; forecasting method on page 4 and named deployments on pages 11–18.
Condor award submission. Published 2026. Vendor’s 2026 award submission describing Condor’s scope and approval behavior.
[19] Publications and public S&OP download
Publications and public S&OP download. Undated. Public landing page linking the white paper and Condor demonstration.