Go back to Market Research
INFORM develops operational planning software for manufacturing, distribution, logistics, aviation, and financial services. Its ADD*ONE suite combines demand planning, inventory replenishment, and series-production planning. It connects to enterprise resource planning systems to calculate forecasts, purchasing recommendations, and production plans. Customers can deploy ADD*ONE on premises, through INFORM’s SaaS offering, or in their own cloud. 1
Overall supply chain score: 5.0/10
| Dimension | Score |
|---|---|
| Supply chain depth | 5.8/10 |
| Decision and optimization substance | 5.2/10 |
| Product and architecture integrity | 5.2/10 |
| Technical transparency | 4.0/10 |
| Vendor seriousness | 4.6/10 |
INFORM has operational substance: its software makes replenishment decisions, coordinates production constraints, and has customers describing reduced inventory and administrative work. The weakness is the distance between these identifiable capabilities and claims of comprehensive, AI-driven optimization. Public material exposes neither the numerical objectives nor enough uncertainty modeling to establish that the decisions maximize economic returns across the supply chain.
The operating model remains a combination of forecasts, planning parameters, automated standard cases, and human exception handling. That can remove substantial work. Its limits appear when an exception requires changing what the software understands about costs, supply, or available actions. The public documentation explains how planners use the system much more readily than how they can repair and test its underlying decision logic.
INFORM’s attribution of AI failures to organizations rather than technology makes this limitation particularly consequential. Integrating operational facts, representing imperfect data, and making failures diagnosable belong to the product’s engineering obligations. Treating them principally as organizational problems weakens accountability for the promised returns. 19
INFORM versus Lokad
The comparison concerns how operational decisions are produced and improved. INFORM packages industry-specific applications around established planning processes. ADD*ONE supplies ordering and production calculations within a system that planners configure, supervise, and supplement. The overlap with Lokad concerns purchasing, inventory, and production decisions.
Lokad’s approach to inventory optimization starts with the financial consequences of possible actions under uncertain futures. The service produces prioritized purchases, allocations, or production decisions from uncertainty in demand, supply, and lead times. Supply Chain Scientists maintain the executable decision logic and data preparation. The broader design perspective appears in Introduction to Supply Chain: improve economic decisions continuously rather than make the organization agree on a forecast.
This distinction matters when a supposedly optimal plan fails. A delayed supplier, a defective batch, or a costly minimum order can reveal that the model omitted a consequence or misinterpreted data. The productive response is to revise the executable logic and test the correction across the relevant decisions. Holimization describes this continuing work on the problem itself.
INFORM’s public evidence is stronger on established operational features and named customer outcomes than on inspectable, revisable economic models. The comparison must include the cost of both vendors’ specialists, customer review, and subsequent model changes, together with the financial improvement these produce.
What ADD*ONE actually decides
Purchasing and production have operational outputs
The Inventory Optimizer identifies an item, quantity, timing, and supplier for replenishment. Its published scope includes minimum orders, delivery times, inventory targets, and transfers across warehouses. These are consequential actions, with opportunities to automate their execution. 2
Series-production planning extends the decision space: batch sizing across stages, setup groupings, material availability, machine capacity, minimum runs, and lot-size restrictions. INFORM also describes cost-based subcontracting and make-or-buy alternatives. The implementation includes a customer-specific feasibility concept. 3
These mechanisms establish more than a demand dashboard. They also expose the central question: which costs and uncertainties determine the selection? Combining capacity and material restrictions is valuable, but it does not reveal how lost sales, delayed supply, expiry, working capital, and alternative uses of scarce resources enter the same calculation.
INFORM’s operations-research explanation describes mathematical models of objectives and restrictions followed by a search for good or optimal decisions. It supplies no ADD*ONE formulation, solver identification, or reproducible comparison. 5 Consequently, neither mathematical optimality nor the company’s research heritage establishes the economic adequacy of its supply chain model.
Demand intervals do not establish joint uncertainty
The Demand AI Forecasting Service offers demand predictions and risk intervals, evaluated through a proposed six-month backtest using three years of sales data. Its public page does not specify forecast origins, tuning separation, calibration, or a comparison at the actual order deadline. 4
The missing supply side matters. A demand distribution is insufficient when replenishment dates, available quantities, or production yields are also uncertain. A plan can look inexpensive because its input assumptions guarantee that material arrives in time. Neither entering a lead time nor testing an alternative scenario demonstrates a learned distribution of future supply, its dependence on demand, or its integration into decision selection.
The same issue arises across products and stages. Safety stocks and a common planning system do not establish a joint treatment of substitution, shared suppliers, capacity competition, and correlated shortages. INFORM does not publish the representation or calibration needed to evaluate that claim.
Its series-production page also promises to meet deadlines despite restricted capacity and to avoid delivery delays. 3 No optimizer can ensure this when insufficient supply or capacity makes the commitments mutually incompatible. The economically meaningful task is then to choose which commitments to sacrifice and at what expected cost. A feasibility promise conceals that trade-off.
Forecast consensus remains part of the design
ADD*ONE Demand Planning supports item, customer, region, and product-group views and company-wide agreement on demand. The S&OP offering connects departments and uses scenarios to reconcile plans. 24 25
Aggregation can help people inspect a business, but a smooth regional forecast does not validate the quantities required at an individual item and location. The relevant forecast must be evaluated at the decision’s resolution and horizon. A consensus process adds value only when its additional information improves downstream decisions enough to pay for the coordination work.
INFORM’s published percentages require the same discipline. The demand page’s “up to 90%” concerns reduced planning effort; the inventory page’s 97.5% concerns a customer’s product availability. Neither is a forecast-accuracy result. 24 26 The more material weakness is that the forecast method and evaluation remain insufficiently specified to connect predictive improvement with incremental financial returns.
Customer results, productivity, and configuration maintenance
Four cases illuminate different parts of the proposition.
| Customer | Published result and scope | Economic interpretation |
|---|---|---|
| hagebau connect | Inventory fell 30% within six months, with ADD*ONE and other inventory-reduction measures. The deployment connects to Microsoft Business Central and replaces manual procurement steps. 7 | A named inventory improvement; attribution is shared, and released cash is not recurring profit. |
| A. Raymond GmbH & Co KG | Across 25,000 items and 13,000 resources, the case reports roughly 10% lower average inventory value, 10% less setup time, and over 60% shorter manual-planning lead time. 8 | The setup result reaches a production cost driver. Shorter planning time does not identify total staffing savings. |
| medi | ADD*ONE has been used since 2016. The case reports inventory savings in the millions, routine-order automation, and one full-time position that no longer needed filling. 9 | The unfilled position is a specific labor outcome. The inventory claim still lacks a net-return calculation. |
| Schäferbarthold | Sourcing can involve up to 30 suppliers for one item. ADD*ONE supports supplier and order selection; delivery readiness reportedly rose by over 10%. 10 | Supplier choice is materially richer than single-source replenishment. The metric’s baseline and cost of the improvement remain unspecified. |
A. Raymond additionally reports that call-off predictability almost doubled, without defining the measure. 8 That cannot be translated into a particular forecast error or monetary gain. The cases nevertheless contain recognizable operational changes. Their limitation is the missing reconciliation of gains with subscriptions, implementation, customer labor, and continuing support.
The medi account exposes maintenance that disappears from headline automation rates. Staff create new selections as requirements change. INFORM and medi jointly developed a supplier portal, used since summer 2023, that returns supplier confirmations to ADD*ONE. 9 This is delivered follow-through and ongoing application engineering. It also means that the cost of useful automation includes the people who adapt the process.
INFORM’s July 2026 automation article makes the dependency explicit: replenishment times, service levels, warehouse cost rates, purchase prices, exception thresholds, priorities, and escalation rules must be maintained. It treats standard cases automatically and routes unresolved deviations to planners. 6
That configuration is part of the decision software. A supplier deterioration can require revising lead-time interpretation, the available purchasing alternatives, and the cost of waiting. Increasing a buffer or adding a warning may preserve the old model while making it more expensive. The public materials do not explain versioning, historical replay, regression testing, or rollback for these interacting rules.
The post-2022 productivity question is therefore more demanding than whether planners can process a list faster. Does a corrected failure become reusable logic, reducing the future queue, or does the organization keep employing people to recognize and resolve it? Existing automation at medi supplies a credible baseline. A new AI feature must eliminate additional work at comparable economic performance.
The separate forecasting service explicitly leaves decisions and overrides with people and preserves their existing process. 4 Its potential contribution is narrower: a better input and less forecast maintenance. Claims about autonomous supply chain decisions require a different demonstration.
Technology history: operational continuity and expanding AI claims
INFORM began in 1969 with operations research and subsequently built distinct products for production, logistics, aviation, workforce management, and risk management. 11 The relevant history is the delivery of particular capabilities within those products.
| Date | Public claim or delivery | What the record establishes |
|---|---|---|
| 2017 | An aviation customer forum discussed machine learning for delay prediction and other airport tasks. 27 | Identified use cases and customer discussion; no quantified deployed-model result in the account. |
| December 2018 | A logistics article promoted cloud AI routing and promised 100% punctual delivery. 12 | The described mechanism was constrained transport planning. The absolute service promise outran what uncertain operations permit. |
| June 2019 | A machine-learning article explained supervised and unsupervised learning, including clustering manufacturing-tool data. 13 | Narrow technical tasks were identifiable. Their scope did not establish general enterprise intelligence. |
| 2024 retrospective | INFORM dated SYNCROTESS’s first installation to 1984 and its SaaS version to 2014; it said core algorithmic concepts remained in use. 14 | Long-lived operational software and a delivered deployment change, with later ML additions asserted rather than benchmarked. |
| September 2025 | A neural-network article described fresh-food forecasting using calendar, origin, price, and promotion information. 16 | A plausible prediction task; architecture and comparative measured performance remained undisclosed. |
| March–May 2026 | PSA Antwerp’s Europa Terminal began deploying the Integrated Terminal Scheduler for straddle carriers and automated stacking cranes. Quay-crane integration was scheduled for later in 2026. 18 | Actual phased deployment, with the remaining integration distinguished from the completed work. |
| Current preview | An aviation webinar previews an MCP server for GS WorkforcePlus’s FSI API. 17 | An interface preview, with no demonstrated change to the scheduling engine. |
The record shows engineering continuity and identifiable deployments. It also shows the AI label covering substantially different mechanisms: routing, clustering, forecasting, and interfaces. Their individual usefulness cannot validate a common claim of intelligence.
An exploratory container-management webinar questions blockchain’s benefits and makes no product-delivery promise. 15 The stronger historical concern is the gap between broad benefit claims and weakly disclosed methods.
The MCP preview warrants skepticism about novelty. Wrapping an existing API can simplify integration, but this does not improve scheduling quality or eliminate the maintenance of planning logic. Its scope is GS WorkforcePlus; ADD*ONE’s capabilities must be assessed independently. 17 No measured incremental productivity accompanies the preview.
The current customer record provides a better test of commitment than the breadth of AI terminology. The custom VMI work at medi and the phased terminal deployment show continuing engineering. Their economics and the costs of later revisions remain much less visible than their functional delivery.
Software bill of materials and system boundaries
| Component or boundary | Attributable role | Assessment |
|---|---|---|
| ERP connectors; on-premises, SaaS, and customer-cloud deployment | ADD*ONE’s published integration and hosting options. 1 | Several operating models transfer different upgrade, infrastructure, and incident responsibilities to the customer. |
| AWS in the EU; HTTPS interface and maintained Python client | Demand AI Forecasting Service, operated by INFORM. 4 | A separate forecasting service with a defined data-transfer boundary. |
| Java/Kotlin and web applications | An undated INFORM Portugal employer vacancy identifies customer projects, including business intelligence. 21 | Attributable to that development team, with no ADD*ONE engine identification. |
| Spring/Spring Boot, JPA/SQL, Angular, Maven, JUnit, Git | Skills listed in the same vacancy, several with alternatives. 21 | Evidence of its engineering environment, not a deployment inventory for every INFORM product. |
| Core ADD*ONE database, optimizer, and scheduling runtime | Not identified in the reviewed product documentation. | Their resource demands, patch burden, and architectural dependencies cannot be independently evaluated. |
The forecasting service receives data snapshots over an outbound connection. Its Python client does not imply practitioner-code execution inside the service. 4
For ADD*ONE, the economically important unknown is what must be operated together for a complete decision: master-data synchronization, forecast refresh, optimization, exception handling, order return, and any supplier portal. Errors can arise at each semantic boundary even when the network connection works. Buyers need a clear owner for reconciling supplier confirmations, available stock, and orders already released.
INFORM prices the inventory solution according to functions, item count, concurrent users, and integration requirements. 2 These drivers expose only part of ownership cost. The lifetime expense also includes adapting customer rules, maintaining connectors through ERP changes, testing upgrades, and revising custom components. Without that accounting, automation claims cannot establish inexpensive scale.
Security requires inspectable engineering
INFORM publishes a coordinated vulnerability-disclosure policy: acknowledgment generally within five business days, initial feedback within ten business days after acknowledgment, and further updates when information is verified. It normally accepts reports without a separate NDA. 22
An October 22, 2026 notice says the company assessed recent PHP vulnerabilities, applied required updates, and secured affected public-facing assets, with no customer action required. It does not identify affected product versions or the exact remediations. 23 This is evidence of public incident communication; the absence of version details limits independent verification.
The ADD*ONE materials emphasize certification and assessments. 1 Those do not reveal least-privilege connector permissions, tenant boundaries, recovery from corrupted state, or the containment of an administrator’s mistake. Under Lokad’s security principles, those mechanisms determine the exposure.
Demand AI excludes external AI providers from customer-data processing. 4 That defines one processing boundary. It leaves the details of isolation, privileged access, and recovery undisclosed. The public evidence does not justify treating certification as a substitute for these design properties.
Who carries responsibility when the model fails?
INFORM’s July 30, 2026 article on AI project failures opens by dismissing weak models as the explanation. Its example is a planner rejecting a recommendation because stock had been found defective but the system did not yet reflect that fact. The article calls for clear responsibilities, workflow integration, and human contextual knowledge; it also recognizes that requiring approval for every action creates a bottleneck. 19
The example undermines the proposed separation. Stock quality and data freshness affect the decision problem. A system that cannot represent them, detect the missing update, or support a reusable correction has an engineering limitation. The customer may control the quality inspection, but the vendor controls substantial parts of how observations become recommendations.
Assigning successful decisions to AI and unsuccessful decisions to organizational readiness creates an asymmetric proposition. The remedy is a bounded operating agreement: which facts the customer must provide, how the system detects failures, who diagnoses them, what corrective work the vendor performs, and how returns are measured after all costs.
INFORM’s July automation article similarly says automation is no longer a technical challenge and presents employee trust as the main hurdle. 6 Routine order processing may be technically mature; automating the expensive exceptions without repeatedly transferring their interpretation to people is a much harder claim.
Its August autonomy article distinguishes routine automation from decisions needing human control. 20 That distinction can be sensible governance. It does not settle how much labor the engine could remove or whether it models the consequences sufficiently well to act without recurring repair.
ADD*ONE’s advertised satisfaction guarantee is a presales exercise using the prospect’s data. 1 This can expose problems before purchase. The public description does not define a post-deployment financial target, total-cost boundary, or responsibility for remediation when expected gains fail to materialize.
Supply chain score
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 6/10 |
| Decision end-state | 8/10 |
| Conceptual sharpness on supply chain | 6/10 |
| Freedom from obsolete doctrinal centerpieces | 4/10 |
| Robustness against KPI theater | 5/10 |
Average: 5.8/10.
Economic framing — 6/10. Purchasing and production include costs and competing resource uses. The undisclosed financial objective prevents assessing whether the combined decisions price shortage, excess, delay, and capital consistently. 2 3
Decision end-state — 8/10. ADD*ONE reaches executable replenishment and production choices, with operational customer evidence. The financial optimality of those actions remains unestablished. 2 9
Conceptual sharpness on supply chain — 6/10. Supplier alternatives and multi-stage production are material dependencies. Their treatment as a jointly uncertain economic system remains unsubstantiated. 3 10
Freedom from obsolete doctrinal centerpieces — 4/10. Forecast alignment, service targets, and exception supervision remain organizing principles. Public material does not establish that their parameters are consequences of financial optimization rather than externally imposed proxies. 6 25
Robustness against KPI theater — 5/10. Named operational outcomes give the claims substance. Undefined predictability measures and uncosted percentage improvements prevent translating the headline figures into net economic performance. 7 8
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 3/10 |
| 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 | 4/10 |
Average: 5.2/10.
Probabilistic modeling depth — 3/10. Risk-adjusted decisions require calibrated probabilities for both demand and supply. The missing joint model prevents assessing the quantities and timing selected. 4
Distinctive optimization or ML substance — 5/10. INFORM describes recognizable prediction and optimization tasks. The published account does not expose formulations, reproducible workloads, or controlled comparisons sufficient to determine the distinctive numerical contribution. 5 16
Real-world constraint handling — 7/10. Production includes interacting resource and batch restrictions, with a substantial customer application. Public details are insufficient to assess which important interactions are omitted or simplified. 3 8
Decision automation and labor productivity — 7/10. medi’s unfilled position is a concrete productivity outcome. The remaining labor across exceptions, configuration, and custom extensions prevents generalizing that result into an unattended operating model. 9
Resilience under real operational complexity — 4/10. Cases show supplier collaboration and adaptation. They do not explain how a model failure becomes a versioned, replayable change to objectives, data meaning, or available actions. Operational flexibility remains dependent on expert intervention. 9 19
Product and architecture integrity
| Criterion | Score |
|---|---|
| Architectural coherence | 6/10 |
| System-boundary clarity | 7/10 |
| Security seriousness | 4/10 |
| Software parsimony and ownership cost | 4/10 |
| Compatibility with programmatic and agent-assisted operations | 5/10 |
Average: 5.2/10.
Architectural coherence — 6/10. ADD*ONE presents connected planning functions and delivered integration work. The underlying numerical and storage architecture remains too opaque to inspect cross-module consistency or duplication. 1 9
System-boundary clarity — 7/10. ADD*ONE’s ERP connections and deployment choices have identifiable roles. Responsibility for reconciling customer data and custom extensions is less precisely documented. 1
Security seriousness — 4/10. The published reporting policy and incident notice do not establish containment, least privilege, or independent recovery. Those deployed controls remain largely undocumented. 22 23
Software parsimony and ownership cost — 4/10. The customer record establishes continuing configuration and extension work. Neither a component-level operating-cost model nor the cost of maintaining those decisions is published. 2 9
Compatibility with programmatic and agent-assisted operations — 5/10. Operational integration supports automated action. The separate MCP preview supplies no measured advance in numerical decisions or lifecycle engineering. 2 17
Technical transparency
| Criterion | Score |
|---|---|
| Public technical documentation | 4/10 |
| Inspectability without vendor mediation | 3/10 |
| Portability and lock-in visibility | 4/10 |
| Implementation-method transparency | 5/10 |
| Evidence density behind technical claims | 4/10 |
Average: 4.0/10.
Public technical documentation — 4/10. Functional pages describe operational inputs and outputs. Mathematical definitions, deployment topology, and executable examples remain largely absent. 2 3 5
Inspectability without vendor mediation — 3/10. Evaluation depends heavily on demonstrations and customer-data exercises. A buyer cannot independently reproduce the decision calculation from the public material. 1
Portability and lock-in visibility — 4/10. Hosting flexibility and external-system connections reduce some infrastructure constraints. Portability of configured decision logic, historical runs, and custom extensions remains undocumented. 1 9
Implementation-method transparency — 5/10. Customer cases reveal integration, organizational changes, and vendor collaboration. Public documentation does not explain the full validation and change-control method for maintaining numerical behavior. 7 9
Evidence density behind technical claims — 4/10. Named cases establish production use. Comparative decision experiments and disclosed numerical methods do not match the breadth of the optimization claims. 5 8
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 4/10 |
| Technology roadmap integrity | 6/10 |
| Conceptual sharpness | 5/10 |
| Economic accountability | 3/10 |
| Defensibility in an agentic-software world | 5/10 |
Average: 4.6/10.
Technical seriousness of public communication — 4/10. Recognizable operational explanations coexist with absolute delivery promises and sweeping AI claims. Those promises obscure infeasibility and the need to choose among costly failures. 3 12
Technology roadmap integrity — 6/10. Long-lived products, a delivered supplier portal, and a phased terminal deployment provide continuity beyond announcements. The contribution of successive AI labels is less clearly measured. 9 14 18
Conceptual sharpness — 5/10. INFORM understands important operational constraints and the distinction between routine and consequential decisions. Its organizational explanation of model failures separates technical responsibility from operational context too conveniently. 19 20
Economic accountability — 3/10. Public results do not include net-TCO reconciliation or a defined failure/remediation agreement. The explicit tendency to attribute AI shortfalls to organizations makes this missing boundary especially important. 1 19
Defensibility in an agentic-software world — 5/10. Operational integrations and established decision engines retain value. Defensibility requires making model revision and exception elimination more productive; assistant access and familiar planning processes do not establish that advance. 9 17
Sources
All sources were accessed on November 4, 2026. Undated pages describe the offering available at that date.
INFORM source, undated. Product scope, hosting options, integration, and presales evaluation.
[2] ADD*ONE Inventory Optimizer
INFORM source, undated. Purchasing decisions, operational inputs, and pricing drivers.
[3] Simultaneous Production Planning
INFORM source, undated. Production restrictions, feasible-plan claims, and cost-based alternatives.
[4] Demand AI Forecasting Service
INFORM source, undated. Forecast protocol, human decision boundary, API, hosting, and data use.
INFORM source, undated. Vendor account of its mathematical optimization approach.
[6] How Process Automation Frees Up Time and Strengthens Your Inventory Management
INFORM source, published 2026-07-20. Automation prerequisites, maintained parameters, and unresolved exceptions.
[7] hagebau connect customer case
INFORM source, undated. Inventory result with shared attribution and ERP integration.
INFORM source, undated. Production scope and separately defined operational results.
[9] medi: Inventory Optimization customer case
INFORM source, undated. Longstanding deployment, labor outcome, recurring configuration, and supplier portal.
[10] Schäferbarthold customer case
INFORM source, undated. Supplier selection and delivery-readiness claim.
INFORM source, undated. Origins and development of distinct product families.
[12] Santa Claus Sends You Tips for a Smart Factory
INFORM source, published 2018-12-11. Early AI routing promotion and absolute punctuality promise.
[13] Machine Learning: A Look Behind the Hype
INFORM source, published 2019-06-17. Contemporary account of narrow machine-learning tasks.
INFORM source, published 2024-11-27. Historical deployment dates and continuity of algorithmic concepts.
[15] Blockchain: Container Management’s Best Friend?
INFORM source, undated. Exploratory webinar, with explicitly questioned benefits.
[16] What Is a Neural Network?
INFORM source, published 2025-09-11. Fresh-food prediction example and limits of the explanation.
[17] Aviation WorkforcePlus Product Exchange
INFORM source, undated. Preview of a separate aviation product’s MCP interface.
[18] PSA Antwerp Europa Terminal deployment
INFORM source, published 2026-05-27. Completed initial phases and planned later crane integration.
[19] Why AI Projects Fail Because of Organizations—Not Technology
INFORM source, published 2026-07-30. Organizational attribution of failures and the defective-stock example.
[20] AI Agents in Practice: How Much Autonomy Makes Sense?
INFORM source, published 2026-08-20. Vendor’s distinction between routine automation and human oversight.
[21] Software Developer Java/Kotlin employer vacancy
INFORM source, undated. Employer document identifying a Portugal team and its development environment.
[22] Coordinated Vulnerability Disclosure policy
INFORM source, undated. Reporting, response timing, and disclosure conditions.
[23] Security Information on Current PHP Vulnerabilities
INFORM source, published 2026-10-22. Vendor’s assessment and remediation statement.
INFORM source, undated. Forecast hierarchies, planning process, and labor-reduction claims.
INFORM source, undated. Cross-functional planning and scenario process.
[26] Inventory Management Software
INFORM source, undated. Inventory approach and customer availability metrics.
[27] Airline Forum 2017 a Success!
INFORM source, published 2017-11-08. Early discussion of airport prediction applications.