Forecast accuracy in these case studies is measured using MAPE (Mean Absolute Percentage Error) — the standard accuracy metric in CPG demand planning. MAPE measures the average absolute percentage error of a forecast versus actual demand across a set of SKUs and time periods. We express it as accuracy: 100% minus MAPE. A 91% accuracy figure means MAPE of 9% — the average forecast unit count was within 9% of actual, measured across all forecasted SKU-weeks in the period.
The measurement protocol is controlled: we measure the ERP baseline model's MAPE over the 12 months prior to Heatvelo deployment (the "before" period) on the same SKU set. We then measure Heatvelo's MAPE over the same SKUs during the pilot period. The comparison is SKU-matched and time-period-controlled — we're not cherry-picking easier SKUs for the Heatvelo measurement.
MAPE has a known limitation: it's undefined when actual demand is zero and can be distorted by very low-volume SKUs. In these case studies, we excluded SKUs with fewer than 200 units/week average from the MAPE calculation — the forecast is most operationally relevant for active-volume SKUs where production decisions are material.
On plausibility: These are synthetic case studies representing the category of supply planning problems Heatvelo addresses. The accuracy figures (89–93%) are consistent with what external signal fusion delivers in academic and practitioner literature for CPG demand planning when signal categories are well-matched to the SKU type. They are not fabricated from thin air — they represent a plausible outcome range for signal-enriched forecasting in these specific categories. Your results will depend on your SKU mix, data quality, and signal relevance to your category.
Want to see how signal fusion performs on your SKU mix?
Request a pilot forecast