Review of Colibri, S&OP Software Vendor

By Léon Levinas-Ménard

Last updated: November, 2025

Colibri is a French cloud-native supply chain planning (S&OP) software vendor operating under the domain colibri-snop.com, positioned as a mid-market solution for demand planning, supply planning and S&OP processes. The company was created as an internal initiative at the IT group VISEO around 2014 and later incorporated as a separate SAS in late 2017, while remaining a VISEO subsidiary, with about 35 employees across Boulogne-Billancourt and Lyon and annual revenues in the low single-digit millions of euros.¹²³⁴⁵⁶ Functionally, Colibri delivers a three-module suite—VISION (demand planning), FLOW (supply planning) and PILOTE (S&OP)—complemented by a pre-packaged “E-Colibri Vision” offer and, more recently, a set of AI/automation extensions branded as “Super Best Fit”, “Data Sensing” and AI assistants.¹⁷⁸⁹¹⁰¹¹¹² Technically, the platform is a multi-tenant SaaS application built on Microsoft Azure, using per-customer SQL databases and a .NET/JavaScript stack, exposing a web UI, an Excel interface and REST APIs for integration with ERPs and CRM systems such as Salesforce.¹²³¹³¹⁰¹¹ Colibri’s forecasting logic relies on a catalog of statistical and machine-learning models with automatic “best-fit” selection and optional exogenous variables, while its supply planning and S&OP modules implement rule- and heuristic-based planning across multi-echelon networks and capacity-constrained scenarios.⁷⁸¹⁴⁹¹⁰¹¹ Over roughly a decade the company has accumulated a client base of about 100 customers and 1,500 users, mostly mid-sized manufacturers, distributors and brands such as IZIPIZI, Puressentiel, Safran Nacelles, Asmodee and Isla Délice, with multiple independent case studies and trade-press articles confirming production use.¹³⁶¹⁵¹⁶¹⁷¹⁸¹⁹²⁰²¹ At the same time, Colibri’s public materials remain high-level: while the AI features are described in detail at the marketing level, there is no technical documentation of model architectures or optimization formulations, so its technology is best assessed as solidly aligned with mid-market state-of-practice—rather than demonstrably at the frontier of supply chain analytics.

Colibri overview

Identity, history and ownership

Colibri is positioned as a “Cloud Supply Chain Planning solution” used to manage demand, supply, distribution, forecasts, replenishments and S&OP.¹¹⁵ Multiple sources converge on a two-stage history:

2014–2017 – product phase inside VISEO. Colibri’s own recruitment page describes it as an “éditeur de solutions innovantes de Supply Chain Management” and explicitly as a startup subsidiary of the VISEO group growing strongly “since 2014”.² An earlier interview with CEO Nicolas Commare in French logistics media describes Colibri as a new-generation, cloud-based supply chain management tool “edited by the VISEO group”.²²
From December 2017 – separate legal entity. French company registries show that COLIBRI (SIREN 834 242 703) was incorporated as a SAS on 19 December 2017 under NAF code 62.02A (IT systems and software consulting), with its registered address at 94–96 rue de Paris, 92100 Boulogne-Billancourt.⁴⁵

Welcome to the Jungle (WTTJ) lists Colibri as an IT/SaaS company with creation year 2017, 35 employees, average age 31 and revenue 4M (currency not specified).³ That same profile states that Colibri is a subsidiary of the VISEO Group, a 3,000-person multi-technology consulting firm with an international footprint.³ The French industrial initiative “La French Fab” similarly describes Colibri as a supply chain planning solution for demand, distribution, forecasting, replenishments and S&OP, explicitly cloud-based on Microsoft Azure and giving VISEO’s address as the company coordinates.²³

Financial data from Pappers for COLIBRI SAS shows:

Revenue €901k in 2019, up 36.5% vs 2018.
Positive net result in 2021 after losses in earlier years, with 7 recorded employees in 2019.⁴

More recent trade-press coverage from Supply Chain Magazine indicates that Colibri’s SaaS S&OP revenue reached €3.3M in 2024, growing nearly 20% year-on-year, with around 110 clients and 35 collaborators.⁶ This aligns with WTTJ’s 35-headcount figure and suggests a small but established vendor.

Taken together, the most coherent interpretation is:

Concept and product started around 2014 as a VISEO internal initiative.
SAS entity created in 2017, still majority-owned and backed by VISEO.
Today Colibri is a small, specialised SaaS vendor with roughly 35 staff and low single-digit million euros of annual revenue.²³²³⁴⁵⁶

Product suite and scope

Colibri’s solution portfolio is organised around three core modules plus a pre-packaged variant:

VISION – Demand planning. The Vision module is described as “demand planning made easy”, providing forecasts, collaborative workflows and dashboards to manage sales forecasts.¹⁰¹¹ It allows multi-level hierarchies (e.g. customer, product family, brand), multiple units (quantities, value), different time buckets and a library of “known and proven statistical models” for forecasting, alongside simulation screens where users can test scenarios before committing a forecast.¹¹ Third-party listings (GetApp, SoftwareAdvice, Logistica Efficiente) all characterise Colibri as a cloud demand and supply planning tool based on statistical and historical models with collaborative features and visual dashboards.²⁴¹³
FLOW – Supply planning. Flow targets distribution and replenishment planning, working by exceptions and managing multi-supplier, multi-echelon contexts.¹⁰⁷ Supply Chain Magazine’s “Colibri S&OP en mode Machine Learning” article describes Flow as covering distribution, replenishments, production and purchasing for multi-site networks, with features for grouped orders, management of minimum and multiple order quantities and visibility on stock and capacity impacts across the network.⁷
PILOTE – S&OP / strategic planning. Pilote is the S&OP layer (“manage your sales and operation planning from A to Z”), providing scenario simulations, comparison of aggregated plans and strategic decision support.¹⁰ Colibri’s own content emphasises that Pilote is used to build and compare long-term scenarios, align demand with capacity and financial objectives and support cross-functional collaboration during S&OP cycles.⁷¹⁵
E-Colibri Vision – pre-packaged demand planning. E-Colibri Vision is a preconfigured, volume-limited version of the demand-planning module, marketed at €500/month with:
- Immediate start,
- Two dimensions with up to four levels each,
- Volume capped at 30,000 SKUs,
- Pre-packaged environment, monthly data updates and limited support, with no contractual commitment and a 30-day free trial.¹²²¹ It is also offered free of charge to students who provide a proof of enrolment.¹²²¹

In addition, Colibri increasingly promotes AI/automation add-ons:

Super Best Fit – an AI-based module comparing and selecting the most suitable forecasting algorithms (statistical, machine-learning, deep-learning) for each use case;
Data Sensing – an exogenous-variable module integrating external drivers (weather, price changes, promotions, competitor stock-outs) and measuring their impact via correlation;
AI assistants – conversational agents expected to execute tasks in the platform, interact with users in multiple languages and perform complex analyses using web data.²⁶⁸¹⁴²⁴⁹¹⁶

These features are promoted as optional extensions that “enrich” the core S&OP offer rather than replacing the existing workflow.⁶⁹¹⁶

Technology and architecture

Colibri is built as a multi-tenant SaaS application on Microsoft Azure. The public architecture and security descriptions indicate:

Per-customer databases: each client has its own application database in Azure SQL, alongside a shared authentication database, with the Colibri Solution Web Application connecting to the appropriate DB at runtime.¹²³¹⁰
Authentication and security: a separate authentication web application using OpenID and OAuth2, accessed over HTTPS with SSL certificates (GeoTrust), optional two-factor authentication, and annual penetration tests by a qualified third party.²³¹⁰
Integration layer: web browser and Excel as user interfaces, plus a REST API for exchanging data with customer information systems and connectors for CRM/ERP systems such as Salesforce.¹¹⁰¹¹¹⁵

Job postings and tech profiles suggest the underlying stack:

Backend: C#/.NET with MS SQL Server, deployed on Azure. A developer job advert mentions C#.NET, jQuery, AngularJS, MS SQL Server and Visual Studio Team Services.²
Frontend: JavaScript single-page app (originally AngularJS, likely evolved), plus Excel integration.
Data/AI: internships in “IA & Python – Creation of intelligent agents for Supply Chain” and “software development assisted by AI” point to Python as the primary language for data-science work.³

In architecture terms Colibri is therefore a conventional ASP.NET + SQL multi-tenant SaaS, not a custom execution environment or data platform.

AI and optimization capabilities

Colibri’s AI/ML story has evolved from basic statistical models to more ambitious automation:

Baseline statistical forecasting. Vision has always exposed multiple statistical models and “best-fit” logic for series-level forecasting.⁷¹¹ Colibri’s own and partner materials highlight its ability to handle aggregated/disaggregated forecasts, promotional adjustments and collaborative overrides.⁷²⁴
Super Best Fit. The Super Best Fit module, launched in 2025, is described as an AI-based system that compares and selects the “most relevant algorithms” (statistical, machine-learning, deep-learning) per use case, relieving planners from choosing models manually and promising “the best forecasts possible”.⁶⁸⁹ No technical details (algorithms, training procedure, error metrics) are disclosed.
Data Sensing. Data Sensing is framed as a causal/exogenous module: it ingests external variables (weather, promotions, price variations, competitor stock-outs, etc.), quantifies their impact through correlation and allows users to incorporate these effects into their forecasts.⁸¹⁴⁹ Again, the underlying modelling technique (e.g. regression vs tree-based vs neural) is not public.
Automation and AI assistants. The same wave of releases introduces features for automatic safety stock calculation by priority/product type, optimization of constrained plans and learning recurring planner actions to automate them, plus AI assistants that will be able to perform tasks, talk to users in any language and analyse data fetched from the internet.⁶⁸¹⁴¹⁶

From a technical scrutiny standpoint:

The conceptual architecture—an ensemble of forecasting models with automated selection, exogenous variables, and heuristic automation—is consistent with mainstream demand-planning practice.
There is no public algorithmic documentation or open benchmarks that would allow independent validation of claims such as “always the best models” or “best forecasts possible”.⁶⁸⁹
There is no evidence of full probabilistic forecasting (modelling demand as distributions rather than point predictions) or of mathematically formulated stochastic optimization; “optimization of constrained plans” appears in marketing language but with no mathematical detail.⁷⁶⁸

Clients and commercial maturity

Colibri reports more than 100 customers and 1,500 users.¹⁶ While that figure is self-reported, independent trade-press and partner sources confirm a non-trivial installed base.

Notable named clients with external confirmation:

IZIPIZI (eyewear brand) – Supply Chain Magazine reports that in late 2019 IZIPIZI selected Colibri’s Vision and Flow modules after an RFP, replacing heavily Excel-based forecasting; the implementation supports S&OP and inventory management, and is explicitly cited as a step-change in their planning.¹⁶
Puressentiel (natural health brand) – Both Supply Chain Magazine and Voxlog report that Puressentiel moved from Excel to Colibri’s Vision module for monthly sales forecasts in France, Belgium and Switzerland, with a three-month implementation and improved planning horizon and collaboration.¹⁷¹⁸ A digital press release adds that Puressentiel chose Colibri S&OP “and its support” to gain maturity in forecasting, again highlighting a three-month deployment.¹⁹
Asmodee (board game publisher) – A press article describes Puressentiel and Asmodee jointly choosing Colibri S&OP as a strategic decision, with both projects deployed in three months to consolidate their sales-forecasting processes.²⁰
Isla Délice (food manufacturer) – Another press roundtable with IZIPIZI and Isla Délice covers how both companies use Colibri S&OP to improve supply chain performance and how Colibri is integrating AI into its offer; while details are light, the event confirms real-world deployment.²¹
Safran Nacelles and GGB Bearing Technology (industrial) – Trustfolio and trade mentions identify these as Colibri Vision users for aeronautical spare parts and industrial bearings forecasting, integrated with SAP.¹¹³¹⁵

These references are consistent with a mid-market focus, primarily in Europe but with some international reach. Supply Chain Magazine’s 2025 piece notes around 110 clients, 20 new projects in 2024 and “first large-scale international projects”, which is compatible with a vendor that is established yet still small in absolute size.⁶

Colibri vs Lokad

Colibri and Lokad both operate in the broad space of supply chain analytics but adopt quite different technical and conceptual approaches.

On the functional surface, both provide cloud-based tools for demand forecasting, inventory and supply planning, and S&OP, and both market AI/ML capabilities. Colibri offers a relatively traditional module structure—Vision (demand), Flow (supply), Pilote (S&OP)—backed by a SQL-based SaaS stack and focused on mid-sized companies seeking to move off Excel into a configurable off-the-shelf tool.¹⁷¹⁰¹¹ Lokad, by contrast, presents itself as a “quantitative supply chain” platform whose core deliverable is a bespoke, code-defined optimization application that directly outputs prioritized decisions (orders, allocations, production, pricing) based on probabilistic forecasts and economic drivers.²⁵²⁶

At the data and modelling level, Colibri’s public material describes a best-fit model selection system over a catalog of statistical and machine-learning models and, more recently, deep-learning options, with optional exogenous variables via the Data Sensing module.⁷⁶⁸¹⁴⁹ Forecasts are described and used as point estimates in examples (monthly sales per country, etc.), and no explicit mention is made of predictive distributions, Monte Carlo simulations or the like. Lokad, on the other hand, explicitly claims to have pivoted to probabilistic forecasting around 2016, modelling full demand and lead-time distributions and using these distributions as the foundation for inventory and pricing optimization.²⁷²⁶ Lokad’s technical documentation emphasises distribution-based forecasts (“probabilities for every possible future demand value”) and probabilistic lead times, and integrates this into optimization routines.²⁷²⁶ In other words, Colibri appears to remain in a mostly deterministic, point-forecast paradigm with ML-based enhancements, whereas Lokad’s value proposition centres on probabilistic forecasting and uncertainty-aware decision-making.

From an optimization standpoint, Colibri talks about “optimization of constrained plans”, automatic safety stock and automated adjustment of plans using AI, but does not publish mathematical formulations or algorithms.⁷⁶⁸¹⁴⁹ The most reasonable interpretation is that Colibri implements heuristic or rule-based DRP and safety-stock calculations, enhanced with ML for parameter suggestions. Lokad, by contrast, publishes technical content describing custom stochastic optimization algorithms, such as Monte-Carlo-based methods and, in public lectures and blog posts, the use of probabilistic forecasts and economic loss functions to rank discrete decisions.²⁸²⁹³⁰²⁶ Lokad’s M5 competition results—6th overall and 1st at the SKU level among 909 teams—are externally verifiable and used to support its claim of world-class probabilistic forecasting performance.²⁸²⁹³¹ There is no comparable external benchmark for Colibri’s forecasting engine; claims such as “always the best models” or “best forecasts possible” remain unsubstantiated beyond marketing narratives.⁶⁸⁹¹⁶

Architecturally, Colibri runs as a standard ASP.NET/SQL multi-tenant web app with per-customer databases, REST APIs and Excel integration.¹²³¹⁰¹¹ Lokad’s own sources describe an in-house domain-specific language (“Envision”) and a distributed execution engine sitting on top of Azure storage, with an event-sourced data store and a focus on code-driven, fully programmable pipelines.²⁶ That leads to different implementation and operating models:

Colibri emphasizes configuration within a fixed application—hierarchies, parameters, planning groups—with limited exposure of underlying code; customisation is largely done by Colibri consultants via configuration and, at most, light scripting.
Lokad exposes coding as the primary interface; its “supply chain scientists” and client teams write Envision scripts defining data transformations, probabilistic models and optimization logic, effectively building a tailored application per client.²⁶

Commercially, Colibri is a smaller, mid-market vendor with around 110 customers and €3.3M revenue in 2024, backed by VISEO and addressing companies that want a packaged solution with relatively fast deployment (often cited at around three months).²³⁷⁶¹⁶¹⁷²⁰ Lokad, founded in 2008, is also relatively small in headcount but targets larger and more complex supply chains with a consultancy-heavy model, as evidenced by case examples with major retailers and industrials and by partnerships documented by third parties such as Brightpearl.³²

In practical terms:

A business seeking a modular, UI-driven APS replacement with standard statistical/ML forecasting and classic DRP/S&OP features may find Colibri easier to adopt: the product resembles a modernised APS with AI bolted-on and is implementation-focused rather than modelling-first.⁷²⁴¹³¹¹¹⁵
A business prioritising probabilistic, decision-centric optimisation and open access to the modelling layer—at the cost of greater complexity and dependence on specialised data-science skills—is closer to Lokad’s positioning.²⁵²⁷²⁶

Colibri and Lokad thus do not approach supply chain problems the same way: Colibri aims to simplify and accelerate traditional planning workflows with AI-enhanced APS functionality, whereas Lokad attempts to reframe planning as a quantitative, programmatic optimisation problem under uncertainty.

Technology and architecture in depth

Multi-tenant SaaS design

Colibri’s technical documentation and public diagrams indicate a straightforward SaaS architecture:

Each customer environment corresponds to an application database in Azure SQL, hosting their transactional and master data.
A separate authentication database and web application handle identity, using OpenID/OAuth2, and all user access is via HTTPS.¹²³¹⁰
The main Colibri Solution Web Application is multi-tenant at the application layer but connects to one database per company, a common pattern that simplifies data isolation and customer-specific schema evolution.¹²³¹⁰

Security-related content emphasises:

TLS/SSL certificates (GeoTrust) for all traffic.
Optional two-factor authentication.
Regular external penetration testing by a PASSI-qualified provider (a French accreditation for security auditors).²³¹⁰

While detailed implementation documents are not public, there is no indication of unusual or advanced architectural features such as event sourcing, message queues, or large-scale distributed computation. This is consistent with Colibri’s focus on daily or monthly batch planning, which rarely requires real-time streaming architectures.

Stack and integration

The WTTJ tech section and job adverts make the stack relatively clear:

Backend: C# / .NET on Azure, with SQL Server/Azure SQL.
Frontend: JavaScript SPA, originally AngularJS plus jQuery, likely modernised to later Angular versions over time.²³
Tooling: Visual Studio / VSTS, suggesting a standard Microsoft-centric CI/CD pipeline.²
Data science: Python-based prototypes and AI work mentioned in internship titles relating to intelligent agents and AI-assisted development.³

Integration points include:

A REST API used by customer systems; details are not fully documented publicly but the architecture diagram shows external “SI” connecting via REST.¹¹⁰
Salesforce AppExchange connector and Microsoft marketplace listing, pointing to integration into CRM and Microsoft ecosystems.¹⁵³³
Excel connectors or export mechanisms for planners who still heavily rely on spreadsheets.¹¹⁰¹¹

This combination is conventional and more than sufficient for the planning horizons Colibri targets (typically monthly planning cycles).

Product capabilities and deployment

Demand planning (VISION)

Vision is the core demand-planning module and the centre of most case studies. Colibri’s site and partner content describe Vision as:

Managing collaborative forecasts across organisation levels and markets.
Providing statistical forecasts using “known and proven models” selected per series.
Offering simulation workbenches where planners can test forecast scenarios before validation.
Supporting multi-level hierarchies (products, customers, regions) and units (volume, value).¹⁰¹¹¹⁵

Supply Chain Magazine’s machine-learning article notes that Colibri focuses on avoiding aggregation artefacts: aggregated demand curves can appear smooth and trendless while hiding divergent behaviour at a detailed level, leading to incoherent disaggregated forecasts.⁷ Vision is positioned as a tool that works at the right granularity to avoid such pitfalls, with aggregated views primarily used to monitor coherence rather than drive the core calculation.⁷

Case studies illustrate typical use:

Puressentiel uses Vision to calculate monthly sales forecasts for France, Belgium and Switzerland, replacing manual Excel-based processes and enabling more collaborative decisions.¹⁷¹⁸¹⁹
Safran Nacelles uses Vision for spare parts demand forecasting, feeding SAP-based planning.¹³¹⁵

Supply planning (FLOW)

Flow extends the data and forecasts from Vision into supply planning:

It manages distribution and replenishment across networks, including multi-supplier configurations and multi-site stock.¹⁰
It operates via exception-driven workflows, focusing planner attention on shortage risks, late deliveries, and items below safety stock.⁷¹⁰
Planning groups can be configured by criteria such as supplier, ABC class or warehouse, with group-specific parameters and user responsibilities.⁷

Trade-press coverage of Flow’s evolution highlights:

Capabilities to handle multi-echelon flows, grouping purchase orders under minimum quantity or batch constraints, and giving visibility on the effect of decisions throughout the network.⁷
Integration with Vision so that supply proposals are directly tied to the current forecast, with the ability to re-run plans after forecast revisions.⁷¹⁶

Flow thus behaves like a modern DRP engine designed for planners comfortable with exception lists and grouped ordering, rather than a full-blown optimization solver.

S&OP and strategic planning (PILOTE)

Pilote is described as the module for:

Building, simulating and comparing S&OP scenarios.
Visualizing impacts on capacity, inventory and service.
Aligning stakeholders around a consensus plan.¹⁰¹⁵

The roadmap built with Supply Chain Movement frames Colibri as a “collaborative S&OP in the cloud” solution, targeting companies suffering from “forecasting chaos” and Excel-based planning, and presenting a series of maturity steps culminating in integrated S&OP processes supported by Pilote.¹⁵ Pilote functions primarily as a scenario manager and visualization layer atop Vision and Flow rather than as an independent optimizer.

Deployment and roll-out

Colibri’s own messaging and independent case studies consistently emphasise rapid deployment:

Supply Chain Magazine notes that Colibri S&OP can be operationally deployed in around three months for typical projects, covering demand, distribution, forecasting, replenishments and S&OP through its three modules.⁷
Puressentiel’s project is reported as completed in three months, extending planning horizons and improving collaboration; the same time frame is echoed in a digital press release describing gains in forecast reliability.¹⁷¹⁸¹⁹
A press article on Puressentiel and Asmodee highlights that both companies implemented Colibri S&OP in three months, with a strong emphasis on the vendor’s accompanying expertise.²⁰
IZIPIZI’s adoption—Vision and Flow after an RFP—took place over a few months, with the company moving from Excel to a cloud solution more capable of handling its rapid growth and complex distribution.¹⁶

The pattern is:

Discovery / RFP, often involving comparisons with other APS vendors.¹⁶²⁰
Configuration of hierarchies, planning groups, alerts and parameters, plus integration with ERP (e.g., SAP).
Historical data upload and cleansing, including segmentation and anomaly handling.
Training, go-live and later expansion (e.g., adding Flow or Pilote after Vision).

These timeframes are realistic for mid-size deployments and do not suggest unusually heavy implementation overhead.

Machine learning, AI and optimization: assessment

From a sceptical standpoint, Colibri’s AI and optimization claims can be summarised as:

Super Best Fit automates model selection across a mix of statistical and ML/deep-learning algorithms; its existence is well documented in multiple independent sources.⁶⁸¹⁴⁹¹⁶
Data Sensing ingests exogenous variables and quantifies their influence, enabling planners to adjust forecasts based on correlated external signals.⁸¹⁴⁹
Automation features span automatic safety stock calculation by priority/product category, automated plan adjustments and the ability to learn recurring user actions.⁶⁸¹⁴
AI assistants are announced as forthcoming capabilities that will execute tasks in the platform and provide conversational support for analyses.¹⁴¹⁶

However:

There are no published technical details (architectures, hyperparameters, evaluation protocols) for the ML models used in Super Best Fit or Data Sensing. Public texts mention classes of methods (statistical, ML, deep learning) but no specifics.⁶⁸⁹
There are no independent benchmarks or competitions demonstrating Colibri’s forecasting performance on standard datasets.
Optimization is described with terms like “optimization of constrained plans” and “better anticipation and performance”, but no mathematical formulations (e.g., mixed-integer programming, stochastic programming) or solver details appear in public sources.⁷⁶⁸¹⁴

The safest technical interpretation is that Colibri implements:

A model-tournament / best-fit forecasting engine, which is a reasonable and increasingly standard approach.
A causal feature module for exogenous regressors.
A set of heuristics and business rules for safety stock and constrained planning, enhanced by pattern recognition over historical planner actions.

That places Colibri broadly at the current mid-market state-of-practice, rather than at the cutting edge of probabilistic modelling or stochastic optimization.

Commercial footprint and client evidence

Colibri’s commercial maturity is supported by several independent indicators:

Revenue and customer base: Supply Chain Magazine reports €3.3M revenue in 2024, nearly 20% growth, about 110 clients and 35 collaborators.⁶
Legal and financial filings: Pappers shows a ramp from sub-€1M revenue and negative results in 2018–2019 to positive net income by 2021, consistent with a scaling SaaS business.⁴
Headcount and organisation: WTTJ confirms 35 employees split between Boulogne-Billancourt and Lyon, with VISEO as parent.³

On the client side, there is overlapping evidence from vendor, partners and independent media for a roster of mid-sized companies across different sectors:

Fashion/accessories (IZIPIZI), consumer health (Puressentiel), food (Isla Délice), entertainment (Asmodee), industrial manufacturing (GGB, Safran Nacelles), among others.¹³¹⁵¹⁶¹⁷¹⁸¹⁹²⁰²¹

There is no sign of tier-1 global retailers or FMCG giants as flagship references, but given Colibri’s scale, this is consistent.

Assessment of state-of-the-art claims

When evaluating Colibri’s technology against the frontier in supply chain analytics, several points emerge:

Cloud-native architecture: Colibri’s Azure-based, multi-tenant ASP.NET/SQL design is contemporary but not unique; many modern SCP tools share a similar architecture. There is no evidence of architectural innovations comparable to custom domain-specific languages, event-sourced data platforms or large-scale probabilistic computation engines.¹²³²³¹⁰¹¹
Forecasting approach: The move from simple statistical models to an ML/deep-learning-augmented best-fit engine mirrors broader trends in enterprise forecasting. Without external benchmarks or algorithmic detail, Colibri’s Super Best Fit must be considered a credible but unverified implementation of those ideas.⁷⁶⁸¹⁴⁹
Probabilistic vs deterministic: Public sources do not mention full probabilistic forecasting (distributions) or Monte Carlo–based decision-making; all examples rely on point forecasts (e.g., monthly volumes) and classical safety stock logic. This contrasts with vendors (including Lokad) that publish documentation around probability distributions and stochastic optimization.²⁷²⁶
Optimization depth: “Optimization of constrained plans” and “automatic safety stock” appear more as descriptive labels than as references to specific mathematical programs or solvers. Given the lack of detail, it is safer to assume heuristic and rule-based logic rather than advanced OR algorithms.⁷⁶⁸¹⁴
Transparency and reproducibility: Unlike some research-driven vendors, Colibri does not publish technical white papers, code or academic collaborations; its AI features are primarily documented through marketing materials and trade-press articles.

Overall, Colibri looks like a technically competent, modern APS for mid-market supply chains, with sensible use of ML/AI to automate model selection and integrate exogenous data. It does not, based on public evidence, demonstrate the kind of deeply documented, probabilistic and optimization-centric stack one would label “state-of-the-art” in a strict research sense.

Conclusion

Colibri is a small but established cloud supply chain planning vendor, originating as a VISEO initiative in 2014 and formalised as a separate French SAS in 2017, now generating around €3.3M in annual revenue with roughly 35 employees and 110 customers.²³⁴⁵⁶ Its software suite—Vision for demand planning, Flow for supply planning, Pilote for S&OP and the E-Colibri Vision pre-packaged offer—targets mid-market companies looking to replace Excel- or legacy-based planning with a configurable, cloud-based APS.¹⁷¹⁰¹¹¹²¹⁵ Architecturally, Colibri is a straightforward multi-tenant ASP.NET/SQL SaaS on Azure, with per-customer databases, web/Excel front-ends and REST-based integration, and security practices aligned with typical enterprise SaaS.¹²³²¹⁰¹¹

On the analytics side, Colibri has evolved from pure statistical forecasting to a model-tournament approach (Super Best Fit) integrating ML and deep learning, plus exogenous variables via Data Sensing and automation features for safety stocks and constrained plans.⁷⁶⁸¹⁴⁹ These capabilities align with current practice in many modern SCP solutions, but are documented at a high level only; there is no public technical documentation or benchmarking that would allow an independent assessment of algorithmic novelty or performance. Optimization appears to be primarily heuristic and rule-based rather than grounded in explicit stochastic or mathematical programming models.

Commercially, Colibri’s reference base—IzIziPI, Puressentiel, Asmodee, Isla Délice, Safran Nacelles, GGB, among others—demonstrates real-world deployments across diverse sectors, typically implemented in about three months and integrated with existing ERPs.¹⁵¹⁶¹⁷¹⁸¹⁹²⁰²¹ This positions Colibri as a credible choice for mid-sized companies that need a packaged, forward-compatible APS with some AI enhancements and are willing to work within a standard module architecture rather than building a bespoke optimisation environment.

When contrasted with more research-driven platforms such as Lokad, Colibri’s value proposition is more about simplifying traditional planning processes with modern SaaS and ML, rather than fundamentally redefining planning around probabilistic optimisation. For organisations prioritising ease of deployment, a familiar modular structure and incremental AI support within an APS paradigm, Colibri offers a pragmatic, well-supported option. For those seeking deeply programmable, probabilistic and decision-centric optimisation, a different type of platform is likely more appropriate.