When and why to use a forecast override
A forecast override is a manual adjustment applied on top of a statistical demand forecast when you have information the model cannot see. Statistical methods like exponential smoothing and moving averages learn from historical sales data, but they cannot anticipate events that have not happened before: an upcoming Lightning Deal, a new PPC campaign, a competitor going out of stock, or a supplier delay that will shift your next shipment window.
The tension with forecast overrides is balancing data-driven forecasting against human judgment. Sellers who override every SKU every month are essentially doing manual forecasting with extra steps. Sellers who never override miss obvious signals that the data has not yet captured. The goal is to reserve forecast overrides for situations where you have concrete, specific intelligence that meaningfully changes the demand picture.
Good candidates for a forecast override include: planned promotions with known traffic multipliers, confirmed competitor listing removals, Subscribe & Save enrollment spikes, seasonal events your model has not seen before (first year selling that product), and supply disruptions that will cause an extended stockout followed by a demand surge when you restock.
Forecast override framework
When to apply a forecast override:
| Trigger | Override Direction | Typical Multiplier |
|---|---|---|
| Lightning Deal planned | ↑ Increase | 1.5× to 3.0× for that week |
| Major PPC campaign launch | ↑ Increase | 1.2× to 1.8× over 2-4 weeks |
| Competitor goes out of stock | ↑ Increase | 1.3× to 2.0× while they are down |
| Supplier delay confirmed | No change to demand, shift timing | N/A |
| Product approaching EOL | ↓ Decrease | 0.5× to 0.8× tapering down |
Example: Lightning Deal override
You sell a stainless steel water bottle at $42 ASP with a 65-day lead time. Your statistical forecast for next month is 380 units based on exponential smoothing. You have a Lightning Deal scheduled that will feature the product on the Deals page for 6 hours.
Based on past Lightning Deals for this product, you have seen a 2.5x lift in daily sales on the deal day, tapering to 1.3x for the following 5 days from the residual traffic and Best Seller rank boost. Here is the override calculation:
| Period | Base Daily | Override Factor | Adjusted Daily |
|---|---|---|---|
| Deal day (1 day) | 12.7 | 2.5× | 31.7 |
| Residual (5 days) | 12.7 | 1.3× | 16.5 |
| Normal (24 days) | 12.7 | 1.0× | 12.7 |
Adjusted monthly forecast: 31.7 + (5 × 16.5) + (24 × 12.7) = 31.7 + 82.5 + 304.8 = 419 units. That is 39 units above the base forecast. At $42 ASP, you need an extra $1,638 in inventory to cover the promotion. Without the forecast override, you risk stocking out during the post-deal traffic bump, which wastes the Best Seller rank you just paid for.
FBA-specific considerations
FBA sellers should reserve forecast overrides for Amazon-specific events that statistical models cannot anticipate. Prime Day is the most obvious case: every July, demand spikes 2 to 5x for participating SKUs, and your statistical forecast trained on June data has no way to know it is coming. The same logic applies to Black Friday, Cyber Monday, and the December holiday surge.
Subscribe & Save enrollment is another override candidate. If you just secured a strong S&S customer base for a new product, your statistical forecast has not seen that committed monthly demand yet. Override upward by the S&S enrollment count to avoid stockouts that would break the auto-ship cycle and damage your enrollment rate.
Competitor intelligence drives more overrides than most sellers realize. If a major competitor in your category gets their listing suspended, their share of category demand redistributes within days. The smart move is a temporary upward override of 1.3x to 2.0x while they are down, scaling back as they return. Without the override, you ride the wave for 60 days, then crash into a stockout right as they return.
Common mistakes
- Overriding too frequently. If you are overriding more than 20% of your SKUs every month, you are essentially doing manual forecasting and the statistical model is noise. Reserve a forecast override for situations with concrete intelligence the model genuinely cannot capture.
- Not tracking override accuracy. Most sellers never check whether their gut-call overrides actually beat the statistical forecast. Track override results: compare the overridden forecast to actuals, and compare it to what the unmodified statistical forecast would have produced. If your overrides are not beating the model on average, the discipline of overriding less is worth more than your judgment.
- Applying overrides without adjusting safety stock. A forecast override changes your demand prediction, but if you do not also recalculate safety stock based on the new demand level, you will either over-buffer (excess inventory) or under-buffer (stockout risk during the very promotion you are overriding for).