Self-built product
Keel
Demand forecasting and inventory planning for retailers importing with long lead times, running on your sales CSV.
A complete product I designed, built and tested myself. Client work is in Case studies.

Long lead times punish a bad forecast
When stock takes weeks to arrive, a retailer has to order well before demand shows up. A simple average of recent weeks reacts too late, and the result is stockouts at the worst time or cash tied up in the wrong products.
What runs out, what to order, and whether to fly it in
Keel forecasts weekly demand per SKU and simulates 300 futures of demand and supplier delays per SKU to show what runs out before stock arrives. Order quantities come from safety stock at a target service level, rounded to case packs and minimums, and air freight is suggested only when the protected margin covers the cost.
Suggestions become editable purchase orders by supplier, fitted to a budget, and a what-if panel tests a stronger Black Friday or a port delay. It runs on your sales CSV.
The right method for each product
For each SKU, Keel chooses between damped trend, seasonal average, Croston’s method for slow sellers and a launch curve for new products. Stockout days are filled with estimated demand first, so a sell-out doesn’t look like low demand.
Tech stack
- TypeScript
More accurate than the spreadsheet method
On 8 held-out weeks, Keel’s forecast error was 20.7%, against 26.7% for a 4-week average.
Screens from Keel

One product: projected stock and a suggested order.

Forecast accuracy on held-out weeks.
Want something like this for your business?
Tell me what you need it to do. I’ll reply within 24 hours on weekdays with questions or a suggested first step.
Start with the $150 scoping session, credited toward the build.