DTC ecommerce · the spike the forecast never saw

Marketing scheduled the email blast two weeks ago. The forecasting system found out when the SKU sold out.

Demand forecasting predicts reorder timing from historical sell-through. It almost never reads the campaign calendar sitting in a different platform — so a scheduled blast that plausibly doubles traffic to a product page is a demand signal the forecast never gets to use.

The desire, and the forecast blind to its own campaigns

Nobody searching “demand forecasting software” wants a model of the past. They want a reorder that happens before a known campaign creates demand, not a stockout that shows up after it already did.

1. 1. The model — historical sell-through, done well. Seasonality, trend, velocity — every demand-forecasting platform models the past reliably. This is the well-known half.
2. 2. The gap — the campaign calendar lives elsewhere. A scheduled email blast or ad push that will plausibly spike traffic to a specific SKU sits in the marketing platform, planned by a different team, unread by the forecast.
3. 3. The close — feed the calendar into the forecast ahead of the send. A SKU featured in an upcoming campaign gets its forecast adjusted before the send goes out, triggering a reorder ahead of the spike instead of after the stockout.

This is a blind spot in what the forecast can see, not a data-quality problem inside it — the difference from /dynamic-pricing-software’s finding.

Which platform this runs on

• Ecommerce platform and email/SMS layer, by scale — the platform holding both the inventory forecast and the campaign calendar it should read.
• Dynamic pricing software: demand it cannot tell from returns — a different blind spot in the same demand signal.
• SMS marketing software: the opt-out it never reads — the same marketing-platform calendar, a different unread field.

Some links on the recommendation page are partner links: the vendor pays us if you sign up, your price does not change, and each button says which.

How to build it yourself, end to end

Five steps. Each names the vendor touchpoint from the picks above.

1. 1. Export the campaign calendar in SKU terms, not campaign names. The forecast needs to know which specific SKUs a scheduled send will feature, not just that a campaign is happening.
2. 2. Set a lead time the forecast update has to beat. The adjustment only helps if it lands with enough runway before the send to actually reorder and restock.
3. 3. Apply a conservative uplift, not a guess. Start from a past campaign’s actual traffic and conversion lift for a similar SKU, not an arbitrary multiplier.
4. 4. Flag the adjustment as campaign-driven in the forecast record. So the next planning cycle can tell a campaign-driven spike apart from a genuine trend change.
5. 5. Compare actual sell-through against the adjusted forecast after the send. The uplift estimate should get more accurate each cycle, not stay a one-time guess.

Step 2 is the one that determines whether this works at all — an adjustment that lands after the reorder window has already closed changes nothing.

Or have it scoped and built

Wiring the campaign calendar into inventory forecasting is scoped work against your specific platforms — the readiness assessment is where that gets mapped out before anything is built.