Go back to Market Research
Sunstice, formerly FuturMaster, provides software for demand forecasting, inventory and distribution planning, production planning and scheduling, integrated business planning, and revenue growth management. Manufacturers and distributors use the platform to coordinate commercial forecasts, production capacity, material requirements, replenishment, pricing, and promotions. FuturMaster adopted the Sunstice name in January 2026, following its February 2025 acquisition of scheduling vendor PlaniSense. (1, 2, 30)
Overall supply chain score: 4.2/10.
| Dimension | Score |
|---|---|
| Supply chain depth | 4.6/10 |
| Decision and optimization substance | 4.4/10 |
| Product and architecture integrity | 3.8/10 |
| Technical transparency | 4.0/10 |
| Vendor seriousness | 4.2/10 |
Sunstice offers a broad planning suite with concrete manufacturing constraints, a history of published optimization work, and named customer results. Its central weakness is the distance between those capabilities and its promise of continuous performance under permanent uncertainty. The disclosed operating model still revolves around forecasts, inventory policies, scenarios, and human arbitration. Public material does not establish an economic optimizer that jointly represents uncertain demand, supply arrivals, and production outcomes when selecting actions.
The implementation deserves closer scrutiny than the unified-platform presentation suggests. Current recruitment describes a monolith-to-microservices migration, a frontend rewrite, and substantial technical debt reduction. Its Helios agents, announced in June 2026, target analytical work around planning decisions while preserving human decision responsibility. That can reduce preparation time; the missing result is a measured reduction in total planning labor, including configuration, review, implementation partners, and continued maintenance. (13, 14, 16)
Sunstice versus Lokad
Sunstice organizes supply chain work around a coordinated plan. Its demand suite produces forecasts; its supply suite translates requirements and policies into distribution, production, and sourcing plans; its IBP suite structures reviews across sales, operations, and finance. Revenue growth management adds pricing, promotional, and commercial profitability analysis. This is an extensive environment for planners and managers to prepare and arbitrate decisions. (3, 4, 5, 6)
Lokad’s starting point is the economic value of the operational decisions themselves: what to buy, make, move, allocate, or discount, given uncertain outcomes and competing uses of scarce resources. Its inventory optimization approach and the broader perspective developed in Introduction to Supply Chain prioritize expected financial consequences over agreement on a forecast or a target service level. Routine decisions should emerge from a maintained numerical recipe, with practitioners improving that recipe instead of repeatedly reviewing its individual outputs.
Consider a confectionery promotion competing with regular products for a constrained production line. Sunstice supplies mechanisms to forecast the promotion, check capacity, schedule production, and compare scenarios. The harder economic question is how much capacity the promotion deserves after accounting for its uncertain uplift, lost regular sales, ingredient arrival risk, expiry, and the future value of inventory. A feasible schedule and a consensus forecast do not answer that question. The disclosed Sunstice material does not show these uncertainties joined in a common financial valuation of competing actions.
This difference also changes the productivity target. Sunstice’s Helios announcement concentrates on removing analytical bottlenecks before people decide. Lokad’s operating objective is to remove recurring intervention from ordinary decisions while retaining human responsibility for the model and its economic assumptions. Both approaches require maintenance; the relevant comparison is total labor per useful decision and realized economic performance, including the people maintaining the automation. (13, 14)
What the software actually calculates
Demand forecasts, inventory policies, and manufacturing constraints
Sunstice’s demand technology describes data cleansing, product segmentation, model selection, and the use of external variables such as weather and promotions. These are identifiable forecasting functions. The documentation does not disclose predictive distributions, calibration results, or how forecast uncertainty flows into financially ranked supply decisions. Correcting outliers also needs care: a surge, shortage, or failed promotion can contain decision-relevant information rather than merely corrupt the series. (3, 10)
The supply suite is more specific about operational constraints. Distribution planning uses inventory policies, service objectives, replenishment frequencies, transport capacities, and allocation priorities. Production planning represents bills of material, alternate routings, yields, and co-products or by-products. Sourcing incorporates supplier capacities, lead times, prices, minimum quantities, and preferred or backup strategies. Scheduling provides automatic scheduling and interactive planner control. This breadth earns credit for handling industrial detail. (4)
The distinction is between representing a lead time and forecasting uncertain lead times. Likewise, a yield parameter is not a distribution of future yields. The reviewed descriptions do not explain how demand, supplier delays, and production variability are combined, including their dependencies, before committing scarce capacity or cash. Inventory targets and service policies remain prominent intermediary objectives.
Sunstice also markets demand sensing. Its concrete foundation includes external-event data: a December 2023 partnership announcement names PredictHQ’s API and 19 event categories. That makes the information input identifiable. It does not establish incremental decision value. The decisive test is whether information available at the actual order date improves decisions over the relevant replenishment horizon, after accounting for forecast revision costs and lead times. A closer-to-consumption signal can arrive too late to change supply. (29)
Published optimization research: useful substance and visible limits
A 2015 paper co-authored by FuturMaster researchers provides unusually concrete historical evidence. It models perishable distribution across two echelons, including lot sizes, transport capacity, freshness requirements, inventory costs, waste, and shortages. It presents a mixed-integer formulation and a multi-start local-search method incorporated into an APS. This is evidence of implemented operations research, beyond attaching an optimization label to a planning interface. (7)
It also exposes the difficulty of the approach. Demand requirements and inventory-policy thresholds are inputs to a deterministic model. In a customer-derived network with 11 warehouses and 145 distribution centers, the mixed-integer formulation exhausted memory above 20 products on the hardware used in the 2015 study. The authors developed a heuristic to handle the computational burden of this formulation.
The underlying concern remains relevant. Expanding a difficult deterministic formulation into many demand, delay, and production scenarios can make it computationally unusable before it becomes economically faithful. Restricting the problem to a manageable perimeter can remove the competition for resources that matters most. More computing power does not repair a missing decision, an arbitrary shortage penalty, or an objective that treats a policy target as business value.
This is the issue described in Lokad’s Holimization essay: the model itself needs revision as economically relevant mechanisms are discovered. A useful implementation must accommodate changes to objectives, possible actions, and constraints. Sunstice’s public documentation does not show how customers perform and validate such revisions beyond the configuration options anticipated by the product.
Overnight global plans and daytime local adjustments
The late-2025 Local Optimization release is more informative than the generic continuous-planning message. It states that Global Optimization builds the complete multilevel plan outside working hours; Local Optimization recalculates a selected perimeter during the day using the same rules and configuration. The current technology page presents these as complementary capabilities. (8, 9)
Reusing rules avoids one source of inconsistency. It does not establish that a local change preserves the best network-wide decision. An urgent order can consume stock or capacity whose opportunity cost lies outside the selected perimeter. Sunstice says that effects on critical resources are highlighted, leaving arbitration with the planner. The important tradeoff is therefore response speed against the completeness of the economic recalculation. Its assertion of no divergence addresses shared configuration more clearly than shared economic optimality.
The elastic-engine page describes cloud scaling, monitoring, and scenario publication. Those mechanisms can shorten processing and improve operational reliability. They do not establish that an arbitrary expansion in products, constraints, or scenarios remains tractable, or that the cost of repeated global calculations is negligible. (31)
A digital twin needs predictive validation
Sunstice’s Data Model describes a graph of supply chain entities and relationships; its Dynamic Digital Twin uses the Network Insight Graph to represent flows, constraints, and scenarios. This can make dependencies easier to inspect. The claim that scenarios respect operational reality requires more than keeping the graph synchronized with current master data. (11, 12)
A scenario predicts the consequences of an intervention: changing a source, routing, production decision, or commercial policy. Its usefulness depends on the accuracy of those consequences and the uncertainty around them. The reviewed pages do not provide validation of intervention forecasts, error propagation across the network, or evidence that model errors are priced into decisions. Graph connectivity and current-state consistency cannot substitute for those tests.
Technology history and roadmap delivery
The history contains substantive engineering alongside repeated inflation of what the technology accomplishes.
| Period | Public commitment or delivery | Assessment |
|---|---|---|
| 2015 | Published perishable-distribution model and local-search implementation in an APS. | Concrete algorithmic work, with disclosed formulation and computational difficulties. (7) |
| 2016 | Executive announced a full-web cloud/SaaS transition, forecasting work using big data and machine learning, and an Ortems scheduling partnership. | Identifiable development direction; the end-to-end proposition already depended on combining products. (26) |
| 2017–2018 | Promotion-focused machine learning, customer pilots, and a new HTML5 interface were announced. Early laboratory improvements of 15–20% were described as revolutionary. | A narrow forecasting experiment was promoted beyond its demonstrated scope; the interview distinguishes laboratory results from planned customer testing. (27) |
| 2018 | A FuturMaster consultant compared AI learning with the brain and described its capacity for improvement as unlimited. | This broad intelligence claim was unsupported by the described forecasting and clustering applications. (28) |
| 2021–2022 | AWS selected in 2021; migrations began in 2022 after earlier privately hosted SaaS. | A documented hosting transition, including provisioning improvements, rather than evidence of a new decision methodology. (15) |
| 2023 | Forecast at Scale and the PredictHQ partnership added large-data forecasting and event inputs. | The data source and product integration are specified; economic superiority is asserted without a reproducible decision benchmark. (29) |
| 2025 | PlaniSense acquired; Local Optimization introduced. | Scheduling capability was brought in through acquisition; global and local calculations now explicitly run at different cadences. (8, 30) |
| 2026 | Sunstice rebrand, Structured Agility framework, and Helios agents. | The rhetoric rejects slow sequential planning, while the disclosed agents prepare analyses and scenarios for human decision processes. (2, 13, 14) |
The early AI language matters because it sets a precedent for presenting bounded statistical techniques as a much broader transformation. The 2018 article’s examples do not support unlimited improvement or human-like understanding. Today’s forecasting documentation describes more concrete functions, but the progression from machine learning to permanent uncertainty and agentic planning still needs evidence of changed decision economics and labor requirements.
The scheduling history also deserves attention. In 2016, management explicitly identified Ortems as FuturMaster’s scheduling component. In 2025, the PlaniSense acquisition was presented as completing the end-to-end offering. PlaniSense itself included demand, distribution, supply, and scheduling modules, creating functional overlap with FuturMaster. The current unified presentation does not explain which engines and data models were consolidated, which were retained, or the migration work imposed on existing customers. (26, 30)
Current recruitment makes the unfinished work visible. The scheduling product-manager role describes a need to balance customer project commitments against a more coherent product roadmap. The software-architect role identifies architectural and frontend migrations. These are concrete delivery obligations against which the unified-platform promise should be judged. (16, 22)
Software materials, security, and ownership cost
Attributable components and their scope
| Evidence and period | Components or architecture | Implication |
|---|---|---|
| 2015 research implementation | C++, Visual Studio 2010, and CPLEX 12.6. | Historical solver and heuristic evidence; the current production solver inventory is undisclosed. (7) |
| AWS migration beginning in 2021–2022 | Managed Oracle requirements; AWS Lambda, Fargate, and EC2. | Several cloud execution mechanisms and an established database dependency were part of the hosting transition. (15) |
| FuturMaster full-stack recruitment, approximately 2025 | Java 17, Spring Boot, React, TypeScript/JavaScript, PostgreSQL, MongoDB, and container-based AWS deployment. | The advertised engineering stack includes two further database technologies and ongoing frontend modernization. (17) |
| Current Sunstice architect recruitment | Java, Spring Boot, React, vanilla JavaScript, Docker; monolith decomposition and migration to microservices. | Legacy and replacement structures coexist within the modernization work described by the employer. (16) |
| Current integration documentation | REST/JSON, file exchange, scheduled/API/event-triggered pipelines, validation, logs, and retries. | Identifiable integration mechanisms, but no public executable API specification or complete export contract in these pages. (18) |
The database evidence spans different periods and product scopes. It does not identify which customer deployment uses Oracle, PostgreSQL, or MongoDB today. That unresolved boundary matters: database migrations, parallel persistence models, and acquired applications create upgrade and operating costs that a single platform name conceals. Sunstice’s own architect listing explicitly names technical debt reduction and complex-module refactoring.
The AWS case reports that environment creation fell to about two hours and that problems could be resolved faster after the migration. These are provider operations improvements. They do not measure planner productivity or show that customer subscriptions pass through lower infrastructure costs. (15)
Security mechanisms remain thinly documented
The October 2025 ISO 27001 announcement emphasizes audits, policies, training, and coordinated processes across operations, support, legal, and HR. Certification demonstrates completion of that management-system assessment; it supplies no direct evidence of an effective technical barrier against data theft, compromised dependencies, or misuse of administrative access. (19)
The integration page identifies encryption, access controls, logging, validation, and retries. These are relevant controls, but the reviewed material does not specify tenant isolation, the scope of support privileges, dependency containment, or recovery from a compromised deployment. Helios adds configurable agents and automated investigation without a comparable explanation of agent permissions, data boundaries, and restrictions on consequential actions. Security credit remains low because the architecture of containment is unclear. (13, 18)
Configuration is continuing software work
Sunstice’s senior-consultant vacancy describes functional design, implementation, testing, hypercare, and continuing optimization. It explicitly seeks people who will spend less time configuring software and more time solving business problems. The aspiration is sensible; it also confirms that configuration and adoption consume professional-services effort. (20)
A separate partner-enablement role targets approximately half of delivery activity being performed by partners over the next two to three years. It includes certification, delivery-quality gates, escalation, and paid enablement. This is a concrete expansion of the implementation ecosystem. Delivery hours transferred to partners remain part of the customer’s ownership cost. (21)
The recurring maintenance includes changing SKU and customer segmentation, supply policies, allocation priorities, forecast treatments, integration mappings, and new agent behavior. Those settings determine production decisions and need versioning, testing, ownership, and correction when business conditions change. A no-code interface does not eliminate that work. The public material does not quantify the labor needed to maintain a representative deployment after go-live, or disclose how customers migrate their decision logic across major product changes.
Productivity, customer outcomes, and economic responsibility
Helios targets diagnosis, root-cause investigation, forecast analysis, and scenario comparison. The accompanying article describes the preparatory work within demand reviews, supply reviews, consensus building, and executive approval. It explicitly leaves business decisions with people. This is a bounded form of automation: potentially useful assistance within a planning process, with no disclosed measurement of how many routine decisions cease to require human handling. (13, 14)
Sunstice needs to measure the entire workflow. Faster investigation can be offset by more scenarios to inspect, more explanations to validate, and more agent configuration to maintain. Useful measures would include the proportion of ordinary decisions executed without intervention, exceptions per planner, total customer/vendor/partner hours, and sustained economic results at comparable operational scale.
The customer evidence is specific enough to examine, although it stops short of a full return calculation:
| Customer | Reported result | Economic reading |
|---|---|---|
| Haribo | Forecast accuracy improved 22% over five years; service reached 99%; inventory at risk fell from 8% to below 2%. | Reduced at-risk inventory is relevant. The forecast metric, evaluation horizon, and complete cost of the transformation are unspecified. The 99% figure is service, not forecast accuracy. (23) |
| Eat Happy | German shop revenue increased 2%, and chiller sales increased 7%, in a business with products lasting 24–48 hours. | Sales increases need corresponding waste, markdown, margin, and labor figures to establish profit improvement. (24) |
| Guerlain | 20% capacity released and 160 tanks scheduled daily in fragrance production. | This addresses a concrete manufacturing bottleneck. The account does not provide an incremental profit calculation or the cost of maintaining the scheduling implementation. (25) |
Haribo’s described workflow remains an 18-month IBP cycle with commercial review, forecast validation, supply alignment, and leadership approval. Its results support the value of better coordination in that deployment. They do not demonstrate that the newly announced agents delivered those gains, or that the same process achieves the labor productivity now possible through routine decision automation. (23)
The accountability boundary is equally important. Sunstice says people retain responsibility for decisions, while its consulting recruitment says success should be measured by business outcomes. These statements need a practical bridge: a baseline, a ledger of costs and benefits, and a method for distinguishing model defects from data and execution failures. The reviewed customer accounts do not expose that accounting or contractual allocation of responsibility. An organization can complete the prescribed reviews and still make systematically poor economic decisions; successful process adoption is not a substitute for returns. (14, 20)
Supply chain score
Each of the 25 criteria has equal weight. Dimension scores are the averages of their five criteria; the overall score is their mean. The assessment credits demonstrated mechanisms and results, with uncertainty lowering the score where central claims remain unsubstantiated.
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 5/10 |
| Decision end-state | 6/10 |
| Conceptual sharpness on supply chain | 5/10 |
| Freedom from obsolete doctrinal centerpieces | 3/10 |
| Robustness against KPI theater | 4/10 |
Average: 4.6/10.
Economic framing — 5/10. RGM addresses prices, promotions, commercial costs, and profitability; the historical distribution model includes transport, stock, waste, and shortages. The unresolved issue is a shared economic objective across these domains rather than separate financial analyses and policy-constrained supply plans. (6, 7)
Decision end-state — 6/10. The suite produces replenishment, sourcing, production, and scheduling outputs, giving it a stronger operational endpoint than reporting alone. Public evidence is less clear about unattended release, decision acceptance rates, and economic ranking when outputs compete for resources. (4)
Conceptual sharpness on supply chain — 5/10. Shelf life, production routings, sourcing options, and transport constraints receive concrete treatment. The broader permanent-uncertainty argument lacks an equally concrete account of joint predictive uncertainty and its consequences for action selection. (4, 7)
Freedom from obsolete doctrinal centerpieces — 3/10. Consensus forecasting, service and stock policies, and demand-to-executive review cycles remain central. Faster scenario preparation improves the operation of that framework without showing that it is the right economic decision architecture. (3, 5)
Robustness against KPI theater — 4/10. At-risk stock, sales, and released capacity are more informative than an accuracy headline alone. The case studies still omit enough cost, baseline, and metric detail to prevent assessment of net returns. (23, 24, 25)
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 2/10 |
| Distinctive optimization or ML substance | 6/10 |
| Real-world constraint handling | 6/10 |
| Decision automation and labor productivity | 4/10 |
| Resilience under real operational complexity | 4/10 |
Average: 4.4/10.
Probabilistic modeling depth — 2/10. Demand variability, external drivers, lead times, and scenarios appear in the product descriptions. A calibrated joint model of uncertain demand, deliveries, and production feeding economic decisions is not disclosed. (4, 10)
Distinctive optimization or ML substance — 6/10. The published perishable-distribution heuristic provides inspectable algorithmic substance and implementation evidence. Current forecasting and local optimization functions are identifiable, but the public bridge from the historical algorithms to today’s complete production engine remains incomplete. (7, 8, 10)
Real-world constraint handling — 6/10. The manufacturing and distribution perimeter includes important nontrivial details, reinforced by the Guerlain scheduling case. The score is limited by the absence of a demonstrated common uncertainty treatment and financial tradeoff across the full suite. (4, 25)
Decision automation and labor productivity — 4/10. Automatic scheduling, data pipelines, and Helios analysis can remove tasks. The disclosed agent workflow preserves human arbitration, and no complete before-and-after labor account establishes a substantial change in decisions handled per practitioner. (13, 14, 18)
Resilience under real operational complexity — 4/10. Local recalculation and cloud elasticity provide practical responses to processing and responsiveness constraints. Global resource competition outside the local perimeter, and the cost of changing the economic model, remain unresolved in the public evidence. (8, 31)
Product and architecture integrity
| Criterion | Score |
|---|---|
| Architectural coherence | 4/10 |
| System-boundary clarity | 5/10 |
| Security seriousness | 2/10 |
| Software parsimony and ownership cost | 3/10 |
| Compatibility with programmatic and agent-assisted operations | 5/10 |
Average: 3.8/10.
Architectural coherence — 4/10. Shared-model and integration mechanisms are described, but the PlaniSense acquisition and current decomposition work leave material questions about module boundaries, duplication, and migration. A common commercial identity is ahead of the publicly inspectable architectural account. (16, 30)
System-boundary clarity — 5/10. The planning perimeter and connection to ERP, manufacturing, and commercial systems are recognizable. API-triggered pipelines and validation are explained at a functional level; precise execution ownership, synchronization guarantees, and failure semantics are not. (1, 18)
Security seriousness — 2/10. Generic controls are listed, but certification receives most of the detailed attention. Tenant containment, privileged access boundaries, and agent permission design need technical substantiation. (13, 18, 19)
Software parsimony and ownership cost — 3/10. Multiple documented technology generations, continuing refactoring, and a growing delivery ecosystem create maintenance obligations. The public record does not quantify those costs or show how the customer avoids carrying them through upgrades and repeated configuration. (15, 16, 17, 21)
Compatibility with programmatic and agent-assisted operations — 5/10. APIs, events, reusable pipelines, and configurable agents provide useful entry points. Public documentation does not establish complete configuration export, reproducible automated deployment, or a safe unattended decision interface spanning the suite. (13, 18)
Technical transparency
| Criterion | Score |
|---|---|
| Public technical documentation | 4/10 |
| Inspectability without vendor mediation | 3/10 |
| Portability and lock-in visibility | 3/10 |
| Implementation-method transparency | 5/10 |
| Evidence density behind technical claims | 5/10 |
Average: 4.0/10.
Public technical documentation — 4/10. Product pages explain functions and inputs, while the historical paper exposes an actual formulation. The current solver design, uncertainty representation, API contracts, and security boundaries remain mostly beyond public inspection. (7, 9, 18)
Inspectability without vendor mediation — 3/10. The paper permits independent examination of one historical method. The contemporary platform is presented through descriptive pages and demonstrations rather than runnable examples or independently reproducible current evaluations. (1, 7)
Portability and lock-in visibility — 3/10. Integration support demonstrates data movement, not the exportability of configured models, rules, agent behavior, and their test histories. The architecture transition makes that distinction particularly important for existing users. (13, 16, 18)
Implementation-method transparency — 5/10. Recruitment and the AWS case disclose technical and delivery work usually absent from the product narrative. They identify testing, ongoing optimization, migrations, and partner governance, though their expected customer effort is not quantified. (15, 20, 21)
Evidence density behind technical claims — 5/10. Published research, attributable engineering requirements, and named operational cases support several substantive conclusions. They do not substantiate the broadest claims about unlimited scale, continuous adaptation, and improved decisions under permanent uncertainty. (7, 12, 16, 25)
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 4/10 |
| Technology roadmap integrity | 5/10 |
| Conceptual sharpness | 4/10 |
| Economic accountability | 4/10 |
| Defensibility in an agentic-software world | 4/10 |
Average: 4.2/10.
Technical seriousness of public communication — 4/10. Concrete solver and infrastructure evidence sits beside inflated historical AI language and current sweeping claims about the fidelity of scenarios. The score rewards the former while treating the latter as a recurring communication weakness. (7, 11, 15, 28)
Technology roadmap integrity — 5/10. The cloud migration and continuing development provide evidence of delivery. Successive end-to-end scheduling propositions, unfinished architectural migrations, and the recent agent release require further follow-through before the latest transformation claims earn comparable credit. (13, 15, 16, 26, 30)
Conceptual sharpness — 4/10. Sunstice recognizes that disruption and decision latency matter. Its public answer still conflates faster analysis, consistent scenarios, and an improved economic choice without sufficiently explaining the transitions between them. (2, 8, 14)
Economic accountability — 4/10. Customer cases report useful outcomes, and the consulting role explicitly emphasizes business results. Full lifecycle costs, attribution, and responsibility for persistent model errors remain unexposed. (20, 23, 24, 25)
Defensibility in an agentic-software world — 4/10. The manufacturing models, implementation knowledge, and scheduling functionality offer more substance than a conversational interface. Helios nevertheless preserves the planning workflow as its main unit of work, without evidence that Sunstice has substantially reduced the human effort required for ordinary supply chain decisions. (7, 13, 14)
Sources
Sources were reviewed on November 4, 2026. Vendor case studies report the vendor’s and customer’s stated outcomes. Dates below follow the individual source rather than the current website’s resource-listing dates.
Sunstice platform. Sunstice, undated. Establishes the current planning and revenue-management perimeter.
[2] FuturMaster becomes Sunstice
Rebrand announcement. Sunstice, January 21, 2026 on the vendor page. Describes the name change, Structured Agility framework, and opposition to sequential planning cycles.
Demand planning. Sunstice, undated. Forecasting, collaboration, promotions, and new-product functions.
Supply planning. Sunstice, undated. Distribution, inventory, production, scheduling, and sourcing inputs and constraints.
[5] Integrated business planning
Integrated business planning. Sunstice, undated. S&OP, execution, financial planning, and demand-to-executive review processes.
Revenue growth management. Sunstice, undated. Pricing, promotions, cost-to-serve, assortment, and financial planning functions.
[7] Published perishable-distribution optimization research
Multi-start Iterated Local Search for Two-echelon Distribution Network for Perishable Products. S. Kande, C. Prins, L. Belgacem, and B. Redon, ICORES 2015, pages 294–303; DOI 10.5220/0005224902940303. FuturMaster-affiliated research describes the mathematical formulation, APS heuristic, implementation, customer-derived instances, and computational results.
Local Optimization announcement. FuturMaster, late 2025. The page displays December 22, while its body dateline says November 22. It explicitly separates global calculation outside working hours from daytime local recalculation.
Intelligent optimization. Sunstice, undated. Describes the shared-model global/local approach and scenario comparison.
[10] AI for demand forecasting
Forecasting technology. Sunstice, undated. Data treatment, segmentation, model selection, and external demand drivers.
Dynamic digital twin. Sunstice, undated. Network Insight Graph, live representation, constraints, and scenario claims.
Data model. Sunstice, undated. Graph structure, dimensions, time series, and scalability claims.
[13] Helios agent announcement
Helios launch. Sunstice, June 11, 2026. Expert agents, configurable Agent Builder, analysis automation, and retained human control.
[14] Agentic AI and planning work
How agentic AI tackles the hidden work slowing down planning decisions. Sunstice, 2026. Explains the analytical tasks agents handle and the business decisions left to people.
FuturMaster AWS case study. Amazon Web Services, undated, discussing 2021–2022 decisions and results. Documents earlier hosting, managed Oracle requirements, Lambda/Fargate/EC2, and provisioning improvements.
[16] Software architect recruitment
Senior Fullstack Software Architect. Sunstice, current recruitment. Specifies technical debt reduction, monolith decomposition, frontend migration, and the Java/Spring/React/Docker stack.
[17] Historical full-stack recruitment
Senior Fullstack Developer. FuturMaster employer listing hosted by WeLoveDevs, approximately 2025 according to the listing’s relative date. Names Java 17, Spring Boot, React, PostgreSQL, MongoDB, AWS, and container deployment.
Data integration. Sunstice, undated. File and API interfaces, no-code preparation, validation, triggering, monitoring, retries, and access controls.
Certification announcement. FuturMaster, October 22, 2025. Defines the management-system scope and process commitments.
[20] Professional-services recruitment
PS Senior Consultant. Sunstice, current recruitment. Functional design, configuration, testing, hypercare, continuous optimization, and business-outcome responsibilities.
[21] Partner delivery strategy
Partner Enablement Lead. Sunstice, current recruitment. Describes partner quality governance and the target of approximately half of delivery activity through partners within two to three years.
[22] Scheduling product management
Product Manager — Scheduling. Sunstice, current recruitment. Explains the tradeoff between customer projects, roadmap commitments, technical debt, and maintenance.
Haribo planning case. Sunstice, undated current account. Forecast, service, at-risk inventory, and IBP process claims.
[24] Eat Happy customer account
Eat Happy planning case. Sunstice, undated current account. Ultra-fresh operating constraints and reported German sales increases.
[25] Guerlain customer account
Guerlain scheduling case. Sunstice, undated current account. Fragrance production, tank scheduling, capacity, and customer explanation of operating benefits.
[26] FuturMaster executive interview, 2016
Michel Ramis interview. Voxlog, March 18, 2016. Firsthand account of the cloud/web roadmap, machine-learning plans, and Ortems scheduling partnership.
[27] FuturMaster executive interview, 2017
Michel Ramis interview. Voxlog, June 8, 2017. Promotion-model research, laboratory improvements, customer pilots, and planned HTML5 interface redevelopment.
[28] FuturMaster consultant’s AI commentary
How artificial intelligence can help the food industry. Stephanie Duvault-Alexandre, FuturMaster consultant, Robotics & Automation News, May 23, 2018. Records the vendor-associated claims about AI learning and forecasting before the LLM era.
FuturMaster and PredictHQ announce a strategic partnership. PredictHQ, December 12, 2023. Specifies Forecast at Scale, event-data integration, and the external information supplied.
FuturMaster and PlaniSense join forces. FuturMaster, February 17, 2025. Describes the acquisition, prior partnership, integration ambitions, and PlaniSense’s module perimeter.
[31] Adaptive and elastic engine
Adaptive and elastic engine. Sunstice, undated. Cloud scaling, operational monitoring, scenario publication, and continuity claims.