Models don’t know your margins.
A forecast that ignores cost structure, MOQs and supplier lead times will route you toward the wrong buy every season.
02Intelligence Layer
Your retail logic,
finally executable.
An AI that respects your metrics, your constraints, your priorities — and proposes one decision per SKU·store·action that you can trust, override, or auto-execute.
Book a demo
Explore another layer
The intelligence problem
Generic AI doesn’t understand your retail business.
Off-the-shelf models optimize for averages. Retail decisions are about exceptions — the SKU·store pairs where the default answer is wrong.
Rules live in spreadsheets.
Minimum coverage by category, never-out-of-stock SKUs, exchange windows — encoded in one head and one Excel, applied inconsistently across teams.
Every team re-invents the wheel.
Buying computes margin one way, finance another, store managers a third. Nobody’s using the same definition of the same number.
How it works
Inside the Intelligence Layer.
Three stages. You configure the logic. The AI respects it.
01DefineActive
Build your metrics and signals.
Start from a pre-built retail library (sell-through, coverage, residual margin) and adapt the formulas to your reality. Every metric is shared across teams.
- 60+ pre-built retail metrics
- Full editor: formulas, aggregations, time windows
- One definition, used by every team and every agent
01 · Definelive
Metric · sell-through
sum(units_sold) / sum(units_received) window: rolling_4w
w-8w-7w-6w-5w-4w-3w-2w-1
02ConstrainActive
Encode your business rules.
Visual rule builder with no code. Hard constraints (never break MOQ), soft preferences (favor the donor store closest to demand), exception flows.
- Pre-built templates for common retail rules
- Versioned, with diff view and rollback
- Bypass with reason, every override audited
02 · Constrainlive
min cover ≥ 5 days
1247 hitsno transfer < 12 units
312 hitsmax markdown ≤ 30%
89 hitsnever out — top 20%
draftversioned · diffable · rollback
03RecommendActive
One decision per SKU·store·action.
The AI produces a single, explainable recommendation per item, ranked, with the reasoning and the alternatives. You approve, override, or auto-execute.
- One recommendation per SKU·store, ranked
- Full reasoning trace — every input visible
- Override with reason; the model learns from it
03 · Recommendlive
Decision · Allocation
Send 38 units of SKU-2417 → Lyon-Bellecour
Demand
+38%
Cover
2d
Margin
62%
Capabilities
The logic center for every retail decision.
Six capabilities to encode, evaluate and trust the logic that runs your network.
Metric factory
Pre-built library + custom editor. One definition shared across teams and agents.
Signals & alerts
Trigger on threshold, trend, or anomaly. Routed to the right team, with context.
Business logic center
Visual rule builder, templates, versioning, rollback. No code required.
Constrained AI
Forecasts and optimization that respect hard rules and soft preferences.
Explainable decisions
Every recommendation comes with its reasoning trace. No black box.
Scenario simulation
Test a rule change on last season before shipping it.
What you can do with it
All case studiesThe logic center for every retail team.
Real teams running rules, forecasts and recommendations on Solya — without rebuilding the same logic in each tool.
Retail was never built for AI.
We’re rebuilding it.
“Last season we marked down €80k of stock in the wrong stores — and missed as much from stockouts. With Solya, both numbers are on the dashboard before the decision, not after.
Stop watching the dashboard.
Start running the decisions.
A 30-minute walkthrough on your own data shape. We’ll show what Solya would decide for your network this week.