Forecasts your buyers will actually trust.
Stokk runs an ensemble of models per SKU per location and picks the best fit. It handles seasonality, promotions, new-product launches and long tails — and writes down what it picked, so your team can challenge it.
Sits on top of your ERP. Reads stock, writes transactions, never duplicates the source of truth.
Most forecasts are one model, applied to everything.
Legacy planners pick a single global model and force every SKU through it. New launches look like outliers. Seasonal SKUs over-shoot in summer and miss in winter. Long-tail items get rounded to noise.
- Per-SKU, per-location — never a global average
- Daily snapshots so you can see what changed and why
- ABC classification baked into safety-stock decisions
- Promotions and events handled as first-class inputs
- Override anything — Stokk learns from your edits
Capabilities, in plain language.
Multi-model ensemble
EWMA, linear, seasonal, naive — Stokk runs them all and picks the best fit per SKU per location, every cycle.
Seasonality without tuning
Detects yearly, weekly and daily patterns automatically. No manual seasonality curves, no maintenance.
Events calendar
Promotions, sales weekends, holidays and store openings feed in as demand multipliers.
ABC + lifecycle classes
A-class SKUs get tighter forecasts and shorter cycles. C-class SKUs aren't allowed to absorb working capital.
Forecast snapshots
Every forecast is versioned daily. Compare, audit, explain — to finance, to suppliers, to yourself.
MAPE & bias tracking
Stokk tracks its own forecast accuracy and flags drift before it bites. No black box.
What it looks like when forecasting actually works.
Two views of the system in motion: how forecast accuracy improves cycle over cycle, and how the events calendar feeds promotions and shutdowns into the model.
- seasonMay 4Mother's Day prep
- promoMay 12Spring promo · 20% off home decor
- shutdownMay 18–25Acme Foods · supplier shutdown
- seasonJun 1Summer reset · A-class only
The numbers Stokk customers report after the first cycle.
After the first full ordering cycle, lost-sale events on A-classified items roughly halve.
Buyers stop building proposals from scratch. The Brief lands; they review and approve.
Stokk's per-store sizing and lateral transfers free cash that was sitting on a pallet.
Continuous, offline-capable counts replace the once-a-year shutdown count.
AI as a forecasting team-mate, not a forecasting black box.
Stokk picks the model per SKU per location automatically — and shows the reasoning. Buyers see why the system chose seasonal vs linear, and what it would predict if you switched.
The supplier intelligence files (SIFs) you write feed into the forecast: shutdowns, recall risks, planned promotions. The model respects your knowledge instead of competing with it.
How AI gets used here
Decisions stay explainable. Every recommendation has a written reason and a human approval step. Your data isn't used to train shared models. Claude's prompts are scoped per request.
Plays nicely with the systems you already pay for.
Common questions about this module.
Do we have to pick the model?
No. Stokk runs the ensemble and picks per SKU per location. You see the choice and can override it, but you never have to make it.
How long do you need before forecasts are useful?
Two cycles is typical. Stokk uses your historical sales (24+ months ideal, 12 months acceptable) to calibrate, then re-evaluates accuracy nightly.
What about new SKUs with no history?
Stokk uses cohort signals from similar SKUs to seed the forecast, then re-fits as real history accumulates. Buyer overrides are respected during the cold-start window.
Other parts of the operating system.
See Forecasting on your data.
A 20-minute demo on your ERP, your SKUs, your stores. We do the work; you decide if it earns its place in your operating system.