Solutions / Stock forecasting

Inventory & stock-out forecasting

Out-of-stock risk per SKU and location, visible one to three weeks ahead, with a reorder point and quantity that account for supplier lead time and its variability.

Operations Inventory ML
7–21 days ahead
stock-out risk per SKU
Per location
not one number for the network
Lead-time aware
including how unreliable it is
Who it is for

Retail and logistics operations planning replenishment across multiple locations or warehouses.

Short answer

Demand forecasts, current stock, open purchase orders and supplier lead times are combined into a stock-out probability for each SKU and location, updated daily. Where the risk crosses your threshold, the system proposes a reorder quantity that reflects the service level you chose, and the lead-time variability of that particular supplier rather than a number typed into the ERP years ago.

The problem

Stock-outs are discovered when a customer cannot buy, and the reorder points were set by someone who has left.

01

Reorder points are frozen

A minimum set three years ago has survived a change in demand, in suppliers and in packaging. It still triggers, and it triggers at the wrong moment in both directions.

02

Lead-time variability is ignored

A supplier who averages fourteen days but sometimes takes thirty is planned as fourteen. Safety stock covers the average and the business runs out on exactly the exceptions that matter.

03

Overstock and stock-out live side by side

Cash sits in slow items in one warehouse while a fast one runs dry in another. Without a per-location view, both look like a single blended service level that hides them.

How it works

From trigger to result, step by step.

01

Reconcile what stock you actually have

System stock versus counted stock, in transit, reserved, damaged, and per location. A forecast on top of an inventory nobody trusts produces confident nonsense.

02

Measure supplier lead times as distributions

Not the contract number — the observed spread from your own purchase orders, per supplier and per item class. Variability is what safety stock is actually paying for.

03

Turn demand forecasts into stock-out probability

The demand forecast, its error, current stock and open orders combine into a probability of running out inside the lead time. That probability, not the raw forecast, is what triggers action.

04

Set service levels by item class

A core product and a long-tail item do not deserve the same protection. Service level per class is a business decision, and it directly determines how much cash sits in stock.

05

Deliver proposals into the buying process

Reorder proposals land in your ERP with quantity, timing, the reason and the risk it avoids. Buyers approve or adjust, and adjustments are tracked so the rules improve.

Before / after

What changes on the ground.

Today, by hand
×Reorder points set years ago and never revisited
×Lead times treated as a fixed contract number
×Stock-outs discovered at the moment of sale
×Overstock and shortage hidden in a blended figure
With the automation running
Stock-out probability per SKU and location, daily
Safety stock sized from observed lead-time variability
Risk visible 7–21 days before it bites
Service level chosen deliberately per item class
What you get

Delivered, not demoed.

A reconciled stock picture across locations and in-transit
Lead-time distributions measured from your own purchase history
Stock-out probability and reorder proposals refreshed daily
Service-level policy per item class, agreed with you
Delivery into your ERP with approval, adjustment and tracking
Built with

We build in your stack rather than moving you onto ours. The list below is what this solution most often connects to — other systems are a scoping question, not a blocker.

Python statsmodels LightGBM SciPy BigQuery dbt Airflow 1C SAP Power BI
Time to production6–10 weeks
Build priceFixed quote
First stepFree mini-audit
Honest limits

When this is not the right solution.

·If system stock does not match physical stock, fix that first. Every number this produces inherits that error, and the project will be blamed for it.
·If you have a handful of SKUs from one reliable supplier, a spreadsheet with a sensible buffer is genuinely enough.
·If purchasing is driven by supplier deals and volume discounts rather than demand, the constraint is commercial, not analytical.

Questions we get about this one

Demand forecasting predicts how much will sell; this converts that, plus stock, open orders and lead time, into the probability of running out and what to reorder. They are usually built together.

Bring us the process that hurts.

The mini-audit is free: we take your version of this process apart and tell you plainly whether automating it pays. If it does, you get a scope and a fixed price.