Q4 budget planning when last year's data lies to you

This post is an example of scheduled publishing. Its date is in the future, so it is hidden on the live site and appears automatically once that date passes. In development it is visible so you can preview it.

Q4 planning almost always starts the same way: pull last year's Q4, apply a growth multiplier, submit the budget.

That works when last year was normal. It quietly fails when it was not, and "was not" happens more often than people remember, because anomalies get forgotten within about six months.

Check whether your base year is usable

Before you multiply anything, ask what was unusual about the comparison period:

• Was there a stockout during peak?
• Did a competitor exit or enter the market?
• Was there a platform change, such as a tracking break, an iOS update, or a policy shift?
• Did you run an unusually aggressive promotion that pulled demand forward?

Any of these makes the base unreliable in a direction you can estimate but not measure precisely.

Build from weekly demand, not quarterly totals

A quarterly total hides the shape. Q4 revenue is not spread evenly. It concentrates into a handful of weeks, and which weeks those are depends on your category.

Pull weekly data for the last three years, index each week against that year's Q3 average, and look at the shape rather than the absolute. The shape is far more stable across years than the totals are, which makes it the better thing to forecast from.

Reserve budget for the week you can't predict

Whatever your plan says, hold back 15–20% unallocated.

Peak season produces surprises in both directions: a channel that unexpectedly clears its target and could absorb more, or a competitor bidding irrationally that makes your planned channel uneconomic. A fully committed budget cannot respond to either.

Set the stop rule before you start

The hardest decision in Q4 is when to stop spending on something that isn't working, because everyone is busy and the season is short.

Decide the threshold in advance, a specific CAC or margin figure, checked on specific dates. Write it down. Agreeing a stop rule in October is a five-minute conversation. Having the same conversation in the middle of peak week is an argument.