Forecast demand with precision using AI
SKU × Store × Day forecasts combining your sales history, future events, and your team's commercial decisions — in a single source of truth.
- Lost sales reduction
- 33%
- Inventory reduction
- 25%
- Forecast accuracy
- 92%
What demand planning is, and what makes it different in retail
Demand planning is the process of estimating how much of each product will sell, at each point of sale, in each period, and turning that estimate into inventory decisions. In retail, demand forecasting has a difficulty of its own: thousands of SKUs, hundreds of stores and a demand that moves with promotions, weather, holidays, paydays and launches. A monthly forecast by category is no use for deciding how much to ship tomorrow to one specific store.
Celes plans demand at the level where decisions are made: SKU × store × day. The model learns from every day of sales, takes in the commercial decisions the team knows before the system does, and measures its own accuracy against what actually sold. The forecast stops being a number argued over in a meeting and becomes the base on which stores are replenished, inventory is distributed and purchase orders are placed. To go deeper, read what demand planning software should do beyond the forecast and how to calculate safety stock.
A forecast built on your data
Celes builds a base forecast at the SKU × Store × Day level from your complete sales history and everything known about the future: seasonality, holidays, weather, local events, category trends, and more. The model learns continuously — every day of operation makes it more accurate.
Your commercial decisions reflected in the forecast
Promotions, price changes, launches, seasonal campaigns — activities your commercial team plans drive future demand. Our forecasting engine recognizes the impact of these activities and automatically reflects the expected demand uplift.
Summer Promo -20%
Promotion
National Holiday — Jul 20
Holiday
New SKU — Cluster A
Launch
Price change $2,400→$2,100
Price
Full flexibility for your planning teams
Planners can add building blocks or adjust the forecast when they have information the model doesn't. The system is the foundation; team knowledge is the complement.
María Ramírez
2h ago
Building block: Back to school +15%
Cat. Notebooks · 42 stores
Julián Cárdenas
4h ago
Manual adjustment: +800 units
SKU Corn Flour · Warehouse 1
AI that monitors and acts on exceptions
AI agents continuously monitor the forecast and detect anomalies — unusual patterns, abrupt changes, atypical behavior. When rules allow, they adjust automatically. With Celes Autopilot, this capability expands even further.
AI AgentAutopilotAnomaly detected
28% demand drop for SKU-8821. Possible supplier issue.
Automatic adjustment applied
Forecast recalculated. +$42K in recovered sales this cycle.
“Celes allows us to ensure patients always find what they need, when they need it, while keeping our finances healthy.”
Celes Autopilot takes this further.
The forecast doesn't just predict — it self-optimizes. An AI agent continuously fine-tunes parameters, validated by a deterministic optimization engine.
Explore Celes AutopilotFrequently asked questions
What is demand planning in retail?
Demand planning is the process of estimating how much of each product will sell, in each store and each period, and turning that estimate into inventory decisions: how much to replenish, transfer and buy. In retail it is done at SKU × store × day level, because that is the level at which decisions are made.
How does Celes forecast demand?
Celes trains a model for every product and store combination on sales history, promotions, weather, holidays and seasonality, and retrains it every day with actual sales. The team can add what the model cannot know (a campaign, a launch), and the forecast measures its own accuracy against what actually sold.
How is Celes different from my ERP's forecasting module?
An ERP usually forecasts by category or product at national level with fixed statistical methods. Celes forecasts every SKU in every store per day with machine-learning models that learn from each sale, and connects that forecast directly to replenishment, distribution and purchasing.
What data does Celes need to start?
Sales history by SKU and store (ideally two years), current inventory, product and store masters, and the logistics parameters: lead times, pack sizes and delivery frequencies. It connects to the existing ERP, POS or WMS; nothing has to be replaced.
How long does implementation take?
Between 3 and 12 weeks from contract signature to running in production, depending on stores, categories and integrations; the first stores are usually live within the first weeks. That is a fraction of what is typical in planning software, where an implementation can take more than a year. A dedicated Customer Success team accompanies the whole process.
Want to see it in action?
Talk to our team and we'll show you how it works with your operation.