The signal fusion methodology behind 94% forecast accuracy.

Weather is the oldest demand driver in CPG, and it's still the most systematically ignored. Temperature deviations from seasonal baseline — not the temperature itself — drive meaningful demand shifts. A warm October in the Upper Midwest doesn't just slow soup velocity; it accelerates cold-brew coffee, carbonated beverages, and outdoor snack categories simultaneously. These are category-specific correlations that play out on a 3–10 day lag from the weather event.

Heatvelo pulls from regional weather API feeds at zip-code resolution, normalizes against 5-year temperature baselines, and maps deviation magnitude against your specific SKU category's historical weather-demand correlation. The model flags events where the deviation exceeds a 1.5-standard-deviation threshold — the level at which demand response becomes operationally material. Precipitation events (particularly unexpected precipitation disrupting outdoor occasions) are tracked separately from temperature, with their own lag and category correlation structures.

Regional specificity matters here. A Pacific storm system has a different demand cascade than a Great Plains cold front — not just in magnitude but in which SKUs respond first. Category-specific weather sensitivity is calibrated per SKU type in your catalog.

The sequence matters more than the magnitude. A TikTok recipe video featuring a specific ingredient doesn't create instant shelf-clearing — it creates a 3–5 week demand build that accelerates before it peaks. By the time POS data shows the demand spike, you're already in week three of a five-week window. You've missed the production commitment window. Social trend velocity is the signal that tells you where demand is building before your scan data knows it.

Heatvelo monitors mention volume velocity — the rate of acceleration of mentions, not the mention count — across platforms where CPG category trends originate: primarily short-form video and recipe communities. The key metric is mention velocity (rate of change), not raw mentions. A static high mention count means the trend is already established and priced into demand. A rapidly accelerating mention count for a previously-low-mention ingredient combination is the signal with forecasting value.

Ingredient-level tracking is also critical. The trend often starts with an ingredient (cucumber-mint, yuzu, black sesame) before brands can react. Heatvelo maps ingredient trend velocity to your specific SKU formulations, flagging which products in your catalog are ingredient-exposed to a building trend.

CPI changes don't reduce grocery spending — they redirect it. A sustained inflation period (CPI above 4% for 3+ consecutive months) predictably shifts basket composition within CPG categories: consumers trade down within protein (premium protein bars to mid-tier), trade across packaging size (bulk vs. single-serve), and substitute between categories (premium snacks to private label staples). These are not random behavioral shifts — they follow measurable macro pressure with a 6–12 week lag from the macro event to full POS response.

Heatvelo tracks CPI at the category level (not just headline CPI), consumer sentiment indices, fuel price trends (which affect both consumer driving behavior and CPG distribution costs), and category-specific spending indices. The model maps which of your SKUs are in macro-sensitive price tiers and recalibrates their demand forecasts based on the current macro signal environment.

The value here is in the lead time: 6–8 weeks of warning that a trade-down shift is building gives your supply planning team time to pre-position mid-tier SKU inventory before the POS data shows any movement at all. That's the difference between meeting the demand wave and scrambling to explain a stock-out to your retail partners.

The POS baseline is the foundation that all external signals are applied against. Your historical scan data — by SKU, by channel, by region — establishes the demand baseline pattern: seasonal curves, promotional lift history, distribution change events, and inherent SKU growth or decline trends. This baseline is not static: Heatvelo continuously updates it as new scan data arrives, so the baseline reflects your most recent 12–24 months of true demand history.

The baseline also establishes per-SKU signal sensitivity. Not all SKUs respond equally to the same weather signal or macro pressure. A trail mix SKU sold primarily in outdoor-activity markets has high weather sensitivity. An everyday staple protein bar has low weather sensitivity but high macro sensitivity. The POS baseline analysis calibrates these sensitivity weights per SKU — so the signal fusion model isn't applying the same weights to every product in your catalog.