Go back to Market Research
Flowlity supplies cloud software for demand forecasting, inventory planning, replenishment, and supplier collaboration. It connects to existing enterprise systems, calculates forecasts and stock recommendations, and provides interfaces for reviewing and sending supply orders. Its product also includes promotion and pricing tools, scheduled automations, and a conversational planning assistant. 1
Overall supply chain score: 5.0/10
| Dimension | Score |
|---|---|
| Supply chain depth | 4.6/10 |
| Decision and optimization substance | 4.8/10 |
| Product and architecture integrity | 5.2/10 |
| Technical transparency | 6.2/10 |
| Vendor seriousness | 4.4/10 |
Flowlity has a substantive inventory-planning core and unusually informative public user documentation. Demand and supplier lead-time uncertainty both enter its stock calculations. The manuals identify real operational functions, limitations, and configuration responsibilities. This evidence supports a more specific diagnosis than the broad AI language used to sell the product.
The principal weakness is the decision model. Probabilities mostly feed service targets, buffers, and replenishment rules whose economic choices remain with the planner. The software automates calculations and some recurring changes, but public evidence does not establish a joint financial optimization of purchasing, stock allocation, capacity, and uncertain future supply.
The marketing also overstates what the published evidence supports. An undefined 95.4% forecast-reliability claim, a projected inventory reduction promoted as an achieved result, and an autopilot narrative surrounding manual reviews all reduce confidence in the performance story. The documentation is often more candid than the sales copy.
Flowlity versus Lokad
Both companies use probabilistic forecasting for supply chain decisions. The consequential difference is what happens after the forecast.
Flowlity translates uncertainty into stock protections that users govern through service targets and inventory policies. Lokad’s inventory optimization service ranks recurring decisions by their expected financial contribution, with Supply Chain Scientists maintaining the economic logic. The contrast concerns the operating model, rather than whether a forecast has a confidence interval.
For example, increasing protection for an unreliable supplier can reduce shortages. It can also tie up capital in a low-margin product while preventing a more valuable purchase elsewhere. Choosing a buffer separately for each item does not resolve that competition. The decision needs the costs and uncertain consequences of the alternatives considered together.
Flowlity exposes useful controls for changing its policies. Lokad’s approach, explained in Introduction to Supply Chain, makes the economic assumptions themselves part of the maintained decision process. In either case, the meaningful test is the recurring financial result after software, implementation, and maintenance labor—not the sophistication of the interface.
The planning engine: probabilities inside a buffer system
Demand and lead times are both modeled
Flowlity’s Stock Min manual explicitly describes probability distributions for demand and resupply time. It sizes protection from the difference between an expected scenario and a higher-risk scenario over replenishment, adjusted by a user-selected buffer level. Inputs include demand history, forecast behavior, supplier history, and time-dependent lead-time uncertainty. The exact formula is withheld. 2
This is a material improvement over assuming that supply arrives on one fixed date. The remaining questions concern how the distributions are combined, how they are calibrated, and whether shared disturbances are retained. Demand surges and delivery delays may arise together. Independently plausible distributions can still underestimate the joint shortage risk.
The forecast documentation describes a model trained across a site’s products, optional product-level expert models, and a weekly forecast run. User forecasts take precedence for periods where they are entered; otherwise planning uses the selected model or external forecast, subject to a floor from known future demand. 4 This makes the human contribution explicit, but leaves the probability model’s architecture, training objective, and calibration results undisclosed.
The service target carries an economic decision
The buffer settings allow service targets from 50% to 99.9%, an additional stock floor, and supplier-delay protection through earlier ordering or additional stock. The customer success manager sets the site’s default target. A user-defined maximum stock level generates an alert; it does not constrain order quantities. 3
These settings embody consequential financial choices. A service percentage does not reveal whether another unit of protection earns more than it costs. Nor does a days-of-coverage floor explain the opportunity cost of capital or warehouse space consumed by that floor.
The uncertainty calculation and the economic decision should therefore be evaluated separately. A well-calibrated high-demand quantile can support a poor purchasing policy if the target is chosen through habit or a broad product classification. The missing evidence is a method that selects protection and orders from their financial consequences across the portfolio.
Constraints depend on which calculation runs
The supplier-constraints manual distinguishes the scheduled planning process from the interactive Optimize action. It says the latter does not yet incorporate supplier minimum orders, full-truck quantities, supplier order frequency, or grouped-product quantity constraints. The scheduled agent exposes which constraints are activated for each site. An emergency-supplier option selects a supplier that can prevent a shortage, choosing the least expensive when several qualify. 5
That is narrower than the AI page’s assurance that all real-world constraints are respected at any detail level. 29 A planner can obtain a recommendation from one calculation path that does not obey the same rules as another. The distinction belongs in an assessment of executable decisions, not merely in a feature checklist.
Capacity is another boundary. The manual describes grouping products by a constrained resource, comparing load against limits, and raising alerts. Resolving an overload involves moving or reducing planned orders in the Planning interface. 6 This documents capacity visibility and manual correction, rather than establishing automatic selection of the economically preferable feasible production plan.
The emergency-supplier rule also leaves an economic question open: preventing every avoidable shortage can cost more than accepting some shortages. The price difference, lost margin, customer consequences, and probability of the ordinary supplier arriving late must determine whether expediting is worthwhile.
Simulations expose policy trade-offs, with limits
The Tactical module lets users change buffer levels for selected products and inspect projected inventory, supply, coverage, and zero-stock days before applying the changes. Its documentation warns that zero-stock-day estimates in Lite simulation can lack precision and points users to a fuller simulation mode. 30
These projections can inform policy changes, but they also make predictions about many future operational states. The digital twin label used in Flowlity’s marketing does not establish their accuracy. Validation should cover projected shortages, receipts, inventories, and resulting costs at the decision horizon, including situations where the policy changes behavior.
The November 2026 roadmap places a richer Scenario Studio in active development and budget-constrained purchasing later on the roadmap. 26 These distinctions limit the present claim of comprehensive optimization: a system cannot jointly allocate scarce purchasing capital through a capability still described as future work.
Forecast accuracy and customer results
A 95.4% claim without a reproducible meaning
The inventory page reports 95.4% consumption-forecast reliability at Saint-Gobain, evaluated against actual sales at three months. It does not identify the formula, product-location granularity, aggregation weights, evaluation dates, or stockout treatment. 21 Such a high percentage, presented without these definitions, is a red flag. It cannot establish accurate predictions for individual replenishment decisions.
Flowlity’s metrics manual makes the aggregation issue explicit. It provides aggregate error calculated after totaling forecasts and demand, weighted absolute errors, and an unweighted average. It acknowledges that the aggregate measure permits opposite errors to cancel. The dashboard uses monthly measurements at several forecast horizons. 7
An aggregate can look excellent while individual items are badly wrong. If one item is overpredicted by 100 units and another underpredicted by 100, their combined forecast is correct while the replenishment decisions can create both excess and shortage. The manual’s clarity helps inspect the product; the public headline still needs to say which measure it uses.
An April 2025 release note also announced replacing product-by-customer forecasting with product-level forecasting subsequently divided among customers. It justified the change partly through higher accuracy at the aggregated level. 24 The relevant comparison is the quality of the resulting customer-level decisions. Better accuracy before splitting does not demonstrate better allocation afterward.
Customer evidence has materially different scopes
| Case | Published result and scope | Assessment |
|---|---|---|
| La Redoute | The October 2026 page reports 50% less packaging inventory measured as storage space, and €37,000–€78,000 annual cost reduction after six months, for 20 packaging references across two sites. 18 | A concrete, limited replenishment application. Storage space is not the same measure as inventory value; the case does not provide a full project-cost reconciliation. |
| Saint-Gobain Sekurit | The October 2026 case reports a 15% SKU-level forecast improvement, 9.25% lower inventory, and availability increasing from 95.8% to 97.2%. It includes named customer testimony. 19 | These operational figures support deployment. The page says end-to-end cost measurement is still being developed, which limits its accompanying claim of proven ROI. |
| Camif | The vendor reports about one FTE of capacity freed, expressed as 1,760 hours annually, with 44% growth and two additional warehouses handled without increasing the planning team. It estimates €40,000 additional annual revenue from avoided stockouts. 20 | Stronger productivity evidence than a generic time-saving claim. Protected revenue is not profit, and freed capacity is not necessarily payroll removed. |
| Danone | The September 2026 headline says inventory fell 40%. Its FAQ describes 17.28% reduction in a six-month simulation and a projected 28–40% reduction over a year, on an initial 27-product pilot. It separately claims subsequent supplier synchronization reduced Danone inventory about 12.40%, and suppliers’ finished goods 30–60%. 17 | The later claims do not turn the earlier 40% projection into a documented achieved result. The page needs to reconcile the periods, baselines, and populations behind its headline. |
The older applications matter historically: they show an operating product before the recent agent branding. They cannot demonstrate the incremental contribution of capabilities introduced in 2025 or 2026.
The economic test also differs by outcome. Inventory released is a balance-sheet change; reduced carrying cost is a recurring expense benefit; additional sales contribute only their margin after associated costs. Combining all three into an undifferentiated ROI claim can count the same improvement more than once.
Productivity: background automation and a persistent review process
The current agent manual documents weekly anomaly correction, forecasting, and product matching, plus supply planning that usually runs daily. Some corrections can be accepted automatically; other configurations wait for review, and changes to anomaly settings go through the customer success manager. 9 This is substantive background computation.
The separate Automations module schedules bulk policy changes against saved views. Those views define the affected products dynamically, so editing a shared filter changes the scope of future runs. 10 This can remove recurring edits, but creates a maintenance obligation: a display filter has become part of executable planning logic.
The prescribed forecast-review process remains labor-intensive. The guide recommends 10–20 products per planner per review, initially weekly and later monthly, with a 45-minute meeting to diagnose items, incorporate commercial input, and record decisions. It assigns final forecast ownership to the planning team and monthly KPI oversight to management. 8
The problem is not the existence of human judgment. It is repeatedly spending expert time on individual SKU forecasts instead of correcting a general defect in the decision process. A recurring shortage caused by supplier correlation will not be resolved by polishing another item’s central forecast.
This is where Holimization changes the assessment. A productive system should help discover why its economic model failed, revise that model, and test the revision across the affected decisions. Flowlity supports input corrections and policy changes. Public evidence is thinner on the labor and validation needed when the policy family itself is inadequate.
MCP changes the interface, not the underlying division of labor
The April 2026 Co-planner announcement presents a connector for investigating SKUs, creating promotions, and operating workflows through an external assistant. It explicitly leaves judgment with the planner. 13 This is a comparatively inexpensive interface extension, not evidence of a new replenishment method or autonomous economic reasoning.
The current manual narrows the operational scope: the connector can read planning data and make selected changes such as creating promotions and views, but cannot validate or modify orders, edit forecast values manually, or change settings. 11 The accompanying skills prepare reviews; users decide and act in the application. 12
Promoting this workflow as supply chain autopilot preserves the existing review model beneath a conversational interface. The engineering accomplishment that would change productivity is a reliable decision process requiring much less recurring human attention. No published comparison establishes that improvement for Co-planner.
Configuration maintenance and economic accountability
A Flowlity implementation leaves a substantial set of choices to maintain: service targets, stock floors, delay policies, product similarities, expert models, supplier constraints, and the views controlling automations. Some choices interact. A fixed floor can neutralize a lower service target; a data feed can overwrite an application-level constraint change.
The cost grows with exceptions, not simply with the number of screens. A customer must establish who recognizes an obsolete policy, who changes it, and how the resulting orders are checked. A supplier changing delivery behavior should trigger more than a permanently larger buffer if a different supplier, purchase cadence, or commercial agreement would be cheaper.
The published commercial model charges Enterprise customers a subscription plus an implementation fee; Lite uses a monthly subscription without an implementation fee. Pricing depends on scope such as SKUs, users, or inventory value. 28 Forecast recomputation and some agent operations also consume credits. 4 9 Full ownership cost includes those charges, customer review work, vendor configuration support, and changes to upstream feeds.
The agent marketing promises that users will have “no more excuses” for missing inventory and service targets. 27 That is an unfortunate allocation of responsibility when users must choose important policy settings and validate recommendations. It makes adoption sound sufficient while leaving the financial consequences with the customer.
The public evidence does not show a standard outcome guarantee or a shared measurement protocol that separates model shortcomings from poor inputs or configuration. The customer cases should disclose the baseline, intervention dates, full recurring costs, and who owns remediation when the promised economics fail to appear.
Technology history and roadmap integrity
| Period | Evidence | What changed |
|---|---|---|
| 2019–2021 | Flowlity’s PR agency identifies its 2019 launch and describes the product as AI-based forecasting and stock optimization in its 2021 client reference. 22 | AI was already the commercial label for a specialized prediction and replenishment application, before the LLM era. |
| 2023 | The article dated September 2023 describes decoupling points, separate consumption and lead-time forecasts, and probabilistic buffers. 23 | The technical center remained inventory protection around local forecasts, with an explicit uncertainty argument. |
| April–September 2025 | Release notes record customer-axis forecast aggregation, the Supply Chain Agents module, and scheduled Automations. 24 | The agent label included existing categories of background forecasting and planning work; actual policy automation was also added. |
| April–May 2026 | Release notes document pricing/promotions and a pricing research agent, followed by Forecast Studio. 25 | These are identifiable product deliveries, including model controls and commercial calculations, rather than an entirely aspirational expansion. |
| November 2026 | The roadmap distinguishes shipped features from development and later work, including Scenario Studio, returns forecasting, and budget-constrained purchasing. 26 | The disclosed future scope is narrower and more useful than an unrestricted claim of end-to-end autonomy. |
Flowlity’s history shows a consistent focus on replenishment. Its weak point is inflation of the language surrounding that core. A scheduled forecasting job, a probabilistic buffer, and an LLM using tools are different technologies with different failure modes.
The release notes also expose a concrete delivery gap. In April 2025, Forecast Studio was expected within six months; its launch is recorded in May 2026. 24 25 Publishing that sequence makes follow-through assessable. It is more informative than awards or claims that the product is uniquely AI-native.
The current doctrine still needs a sharper distinction between forecasting improvement, policy simulation, and financial optimization. The product has advanced, but a sequence of interface and automation releases does not establish that those three problems have become one solved problem.
Bill of software materials
An employer description on investor Fortino’s job board identifies the backend environment. A separate Flowlity data-engineering posting identifies its data-development skills. These sources allow a more concrete inventory than the generic microservices description on the website. 15 16
| Components | Attributed role |
|---|---|
| Node.js, TypeScript, NestJS, PostgreSQL | Application backend; the job identifies PostgreSQL as the source of truth |
| Vue | Application interface |
| RabbitMQ | Listed backend messaging component |
| dbt | Data models, tests, documentation, and lineage |
| Docker, Kubernetes, Terraform | Packaging, orchestration, and infrastructure management |
| GitHub Actions; Elastic/Kibana | Delivery tooling and observability |
| SQL and Python | Required data-engineering skills in Flowlity’s own posting |
| Microsoft Azure | Vendor-disclosed hosting, with customer databases in France 14 |
The backend role also calls for retries, idempotency, schema versioning, observability, and controlled rollout of algorithms. These responsibilities link ingestion, data modeling, and algorithm deployment. The component roles are identifiable; their combined maintenance burden is not quantified.
The recurring cost lies in keeping those layers synchronized. A schema change must preserve data-model tests, ingestion behavior, planning calculations, and downstream exports. Separate customer databases improve isolation while requiring repeatable migrations and monitoring across customers.
Flowlity’s documentation exposes the application’s operations much more clearly than its algorithm internals. Users can inspect and change policies, but the central probability-to-buffer calculation remains proprietary. Exporting forecasts and orders does not make that decision model portable.
Security: meaningful boundaries, incomplete assurances
Flowlity states that each customer has an isolated database, that communications and stored data are encrypted, and that supplier access for support requires authorization. Its integration page describes OAuth 2.0, role-based permissions, and audit trails for assistant access. 14 These mechanisms are more relevant than its certification language.
The MCP manual also distinguishes data destinations. The in-app assistant uses an Azure-hosted model under Flowlity’s agreements; an external assistant receives authorized results under the customer’s separate agreement with that provider. OAuth avoids sharing the user’s password with the assistant, and the documented tool set restricts the available mutations. 11
Authorized access still includes sensitive data and economically consequential promotion creation. The public material does not describe the complete defense against malicious instructions embedded in imported content, nor how the consequences of an unintended promotion are bounded and reversed.
The April launch article argues that deterministic database queries make the numbers real and that only response wording varies. 13 This assurance is too strong. An assistant can select the wrong period, products, or tool parameters and obtain perfectly real numbers answering the wrong question. It can also interpret correct results incorrectly. Correct query execution is one layer of correctness, not a guarantee for the decision.
Supply chain score
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 4/10 |
| Decision end-state | 6/10 |
| Conceptual sharpness on supply chain | 5/10 |
| Freedom from obsolete doctrinal centerpieces | 4/10 |
| Robustness against KPI theater | 4/10 |
Average: 4.6/10
Economic framing — 4/10. Inventory, availability, and commercial costs are recognized. Service targets and buffer settings still carry choices that are not derived from a disclosed portfolio-level financial objective. 2 3 30
Decision end-state — 6/10. The product generates dated order quantities and supports operational transmission. Planner validation and different calculation paths limit the claim that it autonomously produces consistently executable decisions. 5 9 10
Conceptual sharpness on supply chain — 5/10. Demand and lead-time uncertainty receive explicit treatment. The public design is less clear about shared risks, competition for resources, and the distinction between a protective stock level and an economically preferred action. 2 3 5
Freedom from obsolete doctrinal centerpieces — 4/10. Probabilistic buffers improve on fixed averages. The operating model still relies heavily on service targets, stock corridors, product categories, and recurring forecast review. 2 3 8
Robustness against KPI theater — 4/10. The metrics manual explains aggregation and error cancellation. The undefined high-accuracy headline and inconsistent treatment of simulated customer benefits undermine that discipline. 7 17 21
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 6/10 |
| Distinctive optimization or ML substance | 4/10 |
| Real-world constraint handling | 5/10 |
| Decision automation and labor productivity | 5/10 |
| Resilience under real operational complexity | 4/10 |
Average: 4.8/10
Probabilistic modeling depth — 6/10. Both demand and resupply distributions feed time-varying stock protection. The undisclosed formula, calibration, and dependence structure prevent a stronger assessment of their joint adequacy. 2 3
Distinctive optimization or ML substance — 4/10. The documentation identifies cross-product learning, expert models, and operational calculations. It does not provide a reproducible comparison demonstrating distinctive economic optimization. 2 4
Real-world constraint handling — 5/10. Product and supplier rules are concretely documented. Interactive optimization omits several grouped constraints, and capacity handling is described through alerts and planner changes. 5 6
Decision automation and labor productivity — 5/10. Background planning and scheduled policy updates reduce work, and Camif reports a meaningful capacity gain. Co-planner retains the review process, while complete recurring labor remains unmeasured. 10 13 20
Resilience under real operational complexity — 4/10. Users can revise inputs, buffer policies, and simulations. The public evidence is weaker on reformulating the decision problem after failures expose missing dependencies or economically unsuitable rules, and on testing those revisions across the portfolio. 3 25 30
Product and architecture integrity
| Criterion | Score |
|---|---|
| Architectural coherence | 6/10 |
| System-boundary clarity | 7/10 |
| Security seriousness | 5/10 |
| Software parsimony and ownership cost | 4/10 |
| Compatibility with programmatic and agent-assisted operations | 4/10 |
Average: 5.2/10
Architectural coherence — 6/10. Forecasts, policies, orders, and configuration share an identifiable application model. Different computational paths and expanding modules require clearer guarantees about consistent decisions. 2 5 25
System-boundary clarity — 7/10. Flowlity is explicitly an overlay on existing enterprise systems. Current documentation explains assistant permissions, planner responsibilities, and the scope of planned capabilities. 11 14 26
Security seriousness — 5/10. Tenant database separation, authentication, permissions, and data-destination disclosures are substantive. The agent attack model and the overconfident correctness assurance leave important gaps. 11 13 14
Software parsimony and ownership cost — 4/10. The stack has attributable roles, but infrastructure maintenance combines with numerous policy settings, paid recomputations, and recurring customer review work. Total ownership cost is not reconciled publicly with benefits. 3 4 9 15 16 28
Compatibility with programmatic and agent-assisted operations — 4/10. APIs and scheduled calculations support integration. The prominent MCP use cases largely reproduce manual investigations, and the connector cannot carry the ordering process through execution. 10 11 12
Technical transparency
| Criterion | Score |
|---|---|
| Public technical documentation | 7/10 |
| Inspectability without vendor mediation | 7/10 |
| Portability and lock-in visibility | 5/10 |
| Implementation-method transparency | 7/10 |
| Evidence density behind technical claims | 5/10 |
Average: 6.2/10
Public technical documentation — 7/10. Public manuals cover formulas, settings, limitations, permissions, and releases. The central probabilistic and optimization algorithms remain insufficiently disclosed. 2 5 11 25
Inspectability without vendor mediation — 7/10. A reader can establish substantial operating detail without a sales demonstration. The key missing layer is reproducible inspection of the engine and its performance claims. 2 4 5
Portability and lock-in visibility — 5/10. Data and operational outputs have export paths. Proprietary calculations, configuration conventions, and supplier-supported settings would still require replacement during migration. 5 14 30
Implementation-method transparency — 7/10. The manuals and cases describe configuration, testing, and operational review concretely. They expose continuing work, though they do not quantify its full cost. 3 8 17
Evidence density behind technical claims — 5/10. Named cases, current manuals, and detailed release notes provide substantial evidence. Algorithm superiority, high forecast reliability, and fully autonomous operation remain much less substantiated. 17 20 25 29
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 3/10 |
| Technology roadmap integrity | 6/10 |
| Conceptual sharpness | 4/10 |
| Economic accountability | 4/10 |
| Defensibility in an agentic-software world | 5/10 |
Average: 4.4/10
Technical seriousness of public communication — 3/10. The public claims about accuracy, universal constraint handling, autopilot, and deterministic tool correctness exceed the more limited behavior documented elsewhere. 13 21 27 29
Technology roadmap integrity — 6/10. Dated release notes and a staged roadmap show identifiable delivery and make delays visible. Recent agent terminology must still be separated from underlying computational advances. 24 25 26
Conceptual sharpness — 4/10. The product understands inventory uncertainty better than its slogans suggest. It blurs probability-based protection, manual scenario selection, and autonomous financial optimization. 2 27 30
Economic accountability — 4/10. Some cases quantify operating gains and labor capacity. Simulated and achieved benefits are not consistently separated, and responsibility for disappointing economics is not clearly allocated. 17 18 19 20 27
Defensibility in an agentic-software world — 5/10. A maintained replenishment engine and customer integrations can retain value. Conversational access and SKU-review assistance provide much less durable differentiation than demonstrably better decisions with lower total labor. 5 11 12 20
Sources
Sources were checked on November 4, 2026. Customer metrics are vendor-reported unless otherwise stated; publication dates do not establish when the underlying deployment occurred.
[1] AI-Powered Supply Chain forecasting and planning
AI-Powered Supply Chain forecasting and planning. Undated; accessed November 4, 2026.
[2] How Flowlity calculates Stock Min
How Flowlity calculates Stock Min. Undated; accessed November 4, 2026.
Safety stock settings. Undated; accessed November 4, 2026.
[4] Demand forecasting explained
Demand forecasting explained. Undated; accessed November 4, 2026.
[5] Supplier constraints and order management
Supplier constraints and order management. Undated; accessed November 4, 2026.
Capacity. Undated; accessed November 4, 2026.
Metrics glossary. Undated; accessed November 4, 2026.
[8] Run a forecast accuracy review
Run a forecast accuracy review. Undated; accessed November 4, 2026.
Agents. Undated; accessed November 4, 2026.
Automations. Undated; accessed November 4, 2026.
[11] Co-planner in your AI tools
Co-planner in your AI tools. Undated; accessed November 4, 2026.
Co-planner skills. Undated; accessed November 4, 2026.
[13] Flowlity Co-Planner: Conversational AI for Supply Chain
Flowlity Co-Planner: Conversational AI for Supply Chain. Published or updated 2026-04-13; accessed November 4, 2026.
Integration and security. Undated; accessed November 4, 2026.
[15] Senior Software Engineer — Backend
Senior Software Engineer — Backend. Undated; accessed November 4, 2026.
Data Engineer. Undated; accessed November 4, 2026.
[17] Danone automated replenishment case
Danone automated replenishment case. Published or updated 2026-09-29; accessed November 4, 2026.
[18] La Redoute packaging inventory case
La Redoute packaging inventory case. Published or updated 2026-10-14; accessed November 4, 2026.
[19] Saint-Gobain Sekurit forecasting case
Saint-Gobain Sekurit forecasting case. Published or updated 2026-10-19; accessed November 4, 2026.
Camif productivity case. Published or updated 2026-10-16; accessed November 4, 2026.
Inventory optimization. Undated; accessed November 4, 2026.
[22] Citron Plume: Flowlity client reference
Citron Plume: Flowlity client reference. Published or updated 2021; accessed November 4, 2026.
[23] How to build a resilient Supply Chain
How to build a resilient Supply Chain. Published or updated 2023-09-29; accessed November 4, 2026.
2025 release notes. Published or updated 2025; accessed November 4, 2026.
2026 release notes. Published or updated 2026; accessed November 4, 2026.
Roadmap. Published or updated 2026-11-01; accessed November 4, 2026.
AI Agents Suite. Undated; accessed November 4, 2026.
Flowlity FAQ. Undated; accessed November 4, 2026.
Artificial Intelligence. Undated; accessed November 4, 2026.
Tactical. Undated; accessed November 4, 2026.