SKU-level forecasts that release working capital, not just improve MAPE.

AI for Demand Forecasting

T7 builds hierarchical, segmented demand forecasts tied to a dollarised scoreboard your CFO actually reads — not just a MAPE number your data team defends.

The problem

A single global model rarely wins across velocity classes, promotions and new-product introductions.

MAPE improvements don't translate to inventory decisions without a working-capital view — so forecasts sit in dashboards nobody acts on.

How T7 solves it

We segment SKUs by velocity, volatility and lifecycle, then choose a model class per segment — statistical for stable, gradient-boosted for promoted, analogue-priors for new products.

Forecasts feed a weekly working-capital dashboard: excess inventory carrying cost, markdown risk, stockout lost sales and expedited freight — split by category and region.

Continuous evaluation with backtesting, drift monitoring and automated retraining keeps the system honest.

The workflow

Step 1

Segment & baseline

SKU segmentation, exogenous feature engineering (weather, price, promo, calendar) and a baseline model per segment.

Step 2

Forecast & reconcile

Bottom-up forecasts reconciled hierarchically across SKU → category → region → total.

Step 3

Translate to dollars

Weekly dashboard translating forecast error into working-capital impact for the S&OP meeting.

Outcomes we ship for

-4.2pt
MAPE improvement vs. incumbent
$11.8M
Annualised working capital release
-31%
Stockouts on A-class SKUs

The stack

Python / Prophet / statsforecastLightGBM / XGBoostSnowflake / BigQueryMLflow / AirflowPower BI / Looker dashboards

Frequently asked questions

How much history do you need?

Ideally 24 months at daily grain per SKU. We handle sparse and intermittent demand with dedicated model classes and analogue priors for new products.

Do you integrate with our planning system?

Yes — forecasts publish to SAP IBP, o9, Kinaxis, Anaplan or a custom S&OP process via API or scheduled export.

How is the CFO involved?

The dollarised scoreboard is designed for finance leadership: forecast error → working capital → P&L impact, weekly.

Ready to Build Your AI Product?

Talk to a senior AI consultant from T7 about your industry, workflow, or product idea. Free, no commitment — reply within one business day.

  • · AI feasibility & architecture review
  • · Product / MVP roadmap
  • · Integration & automation strategy