Kevin Didelot
Co-founder & CTO, Solya
Co-founder & CTO of Solya. CentraleSupélec engineer with three master's degrees, coding since age 13. Built three companies — one sold — including an IT consulting firm and a product management tool. Specialist in IT architecture for AI, data, and cybersecurity: years designing infrastructure for large enterprise clients across cloud, data platforms, and AI systems. At Solya, responsible for the platform architecture across the Data, Intelligence, Orchestration, and Application Layers.
65 articles
LinkedInDigital merchandising is a decision function
Digital merchandising gets confused with the storefront. The real surface to digitize is the decisions: assortment, allocation, replenishment, markdown.
Inventory optimization is a decision, not a forecast
Most inventory optimization projects solve a math problem that retail no longer has. The leverage moved from the formula to the decision around it.
Kill the analytics queue: let merchants query live data
Every ad-hoc data request is a decision delayed by days. The fix isn't another dashboard — it's letting merchants query live data in plain language.
What every retail CEO should ask about a decision platform
A CEO rarely buys a decision platform — they authorize a change in how the company decides. Here are the questions that tell you whether it will stick.
Generative AI in retail: where it helps, where it's hype
Generative AI is transforming retail's language tasks — search, content, copilots. It is not what decides a markdown or a replenishment. The two get conflated.
Merchandise financial planning software: plan vs decision
A merchandise financial plan sets the sales, margin and open-to-buy targets. The decisions that hit or miss them live in other tools, mostly by hand.
Prescriptive analytics in retail: what it is, where it stops
Prescriptive analytics is the top rung of the analytics ladder: it recommends an action. In retail it stops one step short of executing the decision.
Demand planning vs forecasting: where retail goes wrong
A forecast predicts demand; demand planning decides what to do about it. Confusing the two is why sharper numbers rarely change the result.
Retail data stack vs decision stack
Your modern data stack is composable and complete — and it terminates in a chart. The decision stack is the tier above it, and most retailers are missing it.
WMS vs decision platform in retail: where each one wins
Most retailers confuse WMS and decision platform. They solve different problems, they need each other, and conflating them wastes six-month RFPs.
What is yield management in retail? Price, inventory, time
Yield management started in airlines. In retail it becomes a coordinated arbitration of price, inventory, time and channel — here is the definition.
What is dynamic pricing in retail? A 2026 decisioning view
Dynamic pricing in retail isn't one discipline but two — rules and decisions. Here's the 2026 definition that tells them apart.
Retail KPI dashboards aren't decisions — and why that matters
Most retail teams run 12–40 dashboards. None of them has ever closed a P&L gap on its own. Here's why the KPI dashboard trap costs more than it looks.
Decision intelligence platform: the architecture, in 2026
What a decision intelligence platform actually contains: the four layers, the decision lifecycle, and the things that quietly break in production.
Inventory optimization: consulting engagement or decision platform?
Every CFO eventually asks whether inventory optimization belongs in a consulting engagement or in a platform. The honest answer depends on one variable.
Best demand planning software 2026: a decisioning-first guide
Most demand planning evaluations rank forecast accuracy. The split that actually predicts production value is forecasting-first vs decisioning-first.
Autonomous supply chain: the honest state in 2026
Most 'autonomous supply chain' pitches are rules plus an LLM wrapper. Here's where supply chain autonomy actually works in 2026 — and where it breaks.
AI agents in the retail decision stack: types, fit, anti-hype
Language agents are chatbots. Decision agents change the margin. Here's the taxonomy that cuts through the noise in 2026.
AI inventory management software in 2026: the vendor landscape
AI inventory software comparisons collapse three very different categories into one. The real 2026 split is what predicts 18-month ROI.
What is operational AI in retail?
Operational AI doesn't advise, it acts: it makes the decision, respects the business rules, and drives execution. The distinction that matters in retail.
What is a decision layer in retail?
A decision layer is the part of the retail stack that turns data and models into executed decisions under the chain's own business rules. Here's the definition.
Size-curve decisions: why footwear retail needs its own logic
Footwear is a size-matrix problem. A style with units left but a broken size run is dead on the floor — yet most systems still count aggregate units.
The supply chain VP's playbook for AI agents in retail
If an agent re-orders across your network and gets it wrong, it gets it wrong at scale — and you're accountable. How to automate decisions with guardrails.
Retail decisioning glossary: 15 terms that matter
A clean, opinionated glossary of the 15 terms that define retail decisioning — decision layer, operational AI, sell-through, demand sensing and more.
How to choose a retail decision platform: a guide
You've decided to buy a retail decision platform. The feature checklist won't tell you which one survives production. Here are the criteria that will.
The retail allocation engine: where the season is actually decided
Initial allocation is the most leveraged decision in retail. Every later replenishment lives inside the box it set — yet most chains still allocate by formula.
A retail CDO's guide to decision platforms
You spent years building a modern data stack. A decision layer doesn't replace it — it does the one thing the stack was never built to do.
The real cost of stock-outs: three layers deep, not one
Overstock has been measured to death. Stock-outs are still treated as a temporary nuisance — and the bill is far heavier than the missed unit suggests.
Promo lift is a decision, not a marketing number
Most retailers measure promo success on absolute uplift. That number is wrong in three structural ways — and the year-end bill is rarely seen for what it is.
Why the merch director owns retail AI — not IT
What to buy, how to allocate, when to mark down: the decisions a retail AI automates are merch decisions. So is ownership. Why delegating it to IT backfires.
Integrating a retail decision engine without the rot
A decision engine reads from and writes back into your core systems. Here's the integration architecture that survives production — and how to prove it.
Markdown automation under regulated sale periods
Most markdown engines assume you can discount anything, any day. In regulated markets you can't — and a rule-blind engine can't be deployed.
How to RFP a retail decision platform
Most retail decision-platform RFPs pick the wrong vendor because they test the wrong thing. Here is a reusable structure that selects on adoption and execution.
Dynamic pricing in retail: rules dressed up, or actual decisions?
Most platforms sold as dynamic pricing are rule engines with a slicker UI. Here's a buyer's test to tell rule-based pricing apart from decisions.
Decision platform vs forecasting platform vs BI
Retail buyers conflate three distinct categories: BI describes, forecasting predicts, a decision platform decides. Here is what each is actually built for.
Sport retail runs two inventory logics at once
The core wants continuous replenishment; the seasonal drop wants sell-through and markdown. Run both on one logic and you lose on both walls.
Decision intelligence vs business intelligence
Decision intelligence and business intelligence are different paradigms, not tool tiers. Here is the structural line between them.
Data sovereignty and GDPR for a retail decision AI
A retail decision layer mostly runs on aggregates, not customer identities — which makes sovereignty and GDPR an architectural choice, not a tax.
A CFO's guide to retail AI ROI: which P&L line, and when
Most retail AI ROI decks don't survive a finance review. Here's how to tell cost-saving AI from margin-generating AI — and where it actually moves the P&L.
Continuous replenishment vs. the weekly meeting
The weekly replenishment cadence is an artifact of a pre-data era. Every week your team meets to decide is a week the network drifts further from optimal.
Build vs buy a retail decision platform: an honest framework
You have a warehouse and a data-science team, so building the decision layer in-house looks obvious. Here's the framework that tells you when it isn't.
Big-ticket retail: slow turn inverts the playbook
In furniture and big-ticket home retail, slow turn and heavy logistics amplify the cost of every wrong allocation and late markdown.
AI grocery decisions: why perishables break the order model
In grocery, the question was never how much to order. It is how much to order given spoilage, shelf-life, and the markdown clock running all day.
Managing the DIY long tail: slow movers and project baskets
In DIY retail, cutting a slow-moving SKU can lose the whole project basket. Velocity per SKU is the wrong lens for a long-tail catalogue.
AI assortment planning: the highest-leverage decision in retail
Assortment is set once a season and lives 6–9 months. No in-season tuning recovers a bad pre-season call — which is exactly where most assortment AI fails.
AI assortment and allocation in beauty retail
Beauty's SKU explosion makes assortment and store-level allocation the make-or-break calls. Here's why spreadsheets break and what a decision layer changes.
Retail returns are a feedback signal, not a logistics problem
Most retailers handle returns as a reverse-logistics flow. Treated as a signal feeding allocation and buying, the same returns recover 200–400 bps of margin.
Demand sensing in retail: forecasting for decision-readiness
Forecast accuracy is the wrong KPI. A 92%-accurate forecast that lands after the replenishment cut-off is worth less than a 78% forecast that lands in time.
Why retailers lose money between stores without knowing it
Most retail leadership teams discuss network performance store by store. But the biggest leaks aren't inside the stores — they're between them.
Why your BI tools don't make decisions (and never will)
Retail dashboards have never been clearer, and yet the same problems keep eating margin. Here's why BI tools can't fix what they were never built for.
Why 80% of retail business rules are misused in data systems
Recommendations get rejected, teams revert to Excel, projects stall. The root cause is rarely the model — it's that the rules live in heads.
Why 20+ store chains need centralized decisions
Past roughly 20 stores, the methods that built your chain start working against it. The threshold is mathematical, not organizational.
Why 70% of retail markdown decisions are still manual (and expensive)
Most chains still set markdowns from a spreadsheet on Monday morning. Here's why the manual habit persists — and what it actually costs in net margin.
What the top retailers share: a closed decision → execution loop
Copying Zara or Costco doesn't work — what sets them apart is invisible: a closed decision-to-execution loop most retailers lack.
The real cost of overstock: what your ERP isn't telling you
Ask a retail CFO what overstock costs them and you'll get a number. It's almost always a fraction of the real bill — here's where the rest hides.
Why the future of retail isn't data, but automated decisions
The data race is over — everyone has the same infrastructure now. The next decade of retail competition will be won on something else: decision automation.
The invisible problem of modern retail: decision fragmentation
Ten retail teams will name ten different problems. Above all of them sits one structural issue nobody talks about: decisions never meet each other.
How retail data becomes useless without a decision layer
Ten years of data investment, and operational KPIs barely moved. The missing piece isn't more data — it's the layer that turns it into executed decisions.
From stock to cash: how decisions actually drive retail performance
Stripped of everything else, retail is a stock-to-cash converter. The real lever isn't margins or contracts — it's the speed and quality of daily decisions.
Why automating workflows isn't enough if decisions stay manual
RPA and workflow tools made retail processes faster — but the same decisions, just executed faster, won't move margin. Here's the trap, and the way out.
From forecasting to decision: why ML models aren't enough in retail
Retail forecasts have never been more accurate, yet operational results haven't budged. Here's the gap between predicting and deciding — and how to close it.
5 weak signals a product should be liquidated
Every retail product has a pivot moment — weeks before numbers show it, months before the markdown. Most systems miss it.
3 classic mistakes that kill retail data projects
For every retail data project that ships into production with measurable impact, three or four POC successes quietly die. Here's why — and how to avoid it.
AI Buying Agent Fashion Retail — Solya, Decision Runtime
An AI buying agent for multi-store fashion retail that decides + explains + executes. 4-layer architecture, native PoS integration, 4-week deploy.
How to avoid end-of-season overstock in fashion
End-of-season overstock? -22% on average with Solya. 7 concrete levers, quantified, validated by 80+ fashion retailers. Complete 2026 guide.