Connect the operating data
Sales, inventory, inbound orders, promotions, supplier rules, and purchasing limits enter one controlled planning cycle.
How it works
Sales, inventory, inbound orders, promotions, supplier rules, and purchasing limits enter one controlled planning cycle.
Missing supplier constraints, missing unit costs, low-confidence forecasts, and purchasing-limit overruns are surfaced as management exceptions.
Each SKU receives a demand forecast, inventory position, reorder decision, risk classification, and forecast-versus-actual record.
You receive supplier-grouped PO recommendations and a 13-week purchasing-cash schedule. Your team retains final approval.
How accuracy is established
For each managed SKU, up to 24 candidate forecasting methods drawn from the demand-planning literature — including Croston's method and its variants for intermittent demand, exponential smoothing, and the Theta method that won the M3 forecasting competition — are replayed against held-out recent history with a rolling-origin backtest across multiple horizons. A challenger method only replaces the simpler baseline when a Diebold-Mariano significance test, Bonferroni-corrected for the number of methods compared, confirms the improvement is not backtest noise.
How far forecast units were from actual units, weighted by sales volume.
Whether the method systematically planned above or below actual demand.
A method only wins if it beats the baseline by a statistically and economically material margin.
Forecasts are compared with actual outcomes after every completed cycle.
Backtests estimate historical performance; they do not guarantee future demand. Seasonal-naive models join the candidate pool once a SKU has enough history to test monthly and quarterly cycles (52+ weeks) and annual cycles (104+ weeks) — a newer SKU is not yet modeled for seasonality. Promotions, stockouts, launches and supplier changes are reviewed separately because history alone cannot describe them.
Versus a forecasting dashboard
Inventory Planner, Cogsy, and similar tools compute a forecast and hand it back to you as another screen to check. Someone on your team still has to read it, decide what it means, and turn it into a PO every cycle.
We run the forecast, build the PO recommendation, and only route it to you when something needs a decision. No dashboard to check on a schedule — a plan lands in your inbox, ready to approve.
Ready to see your own numbers?