Direct answer
Cut by evidence, not by how easy a line is to cancel. First go the lines nobody can say what they bought, then duplicates reaching the same buyer with the same message, then experiments past their review date, then spend supported only by the platform's own report. Protect the decisions work, the channel your own records show closing deals, sales follow-up, basic measurement, and long-cycle brand spend, which should be reduced slowly if at all.
When a budget has to come down, the order of cuts usually follows convenience. The first things to go are whatever has a monthly cancellation clause, whatever is visible on the budget sheet, and whatever has no one in the room to defend it. That order has nothing to do with what the spend produced.
There is a better order, and it starts from a different question. Not "what can we cancel?" but "what can we show that this bought?" Spend that can answer goes last. Spend that cannot answer goes first, however long it has been in the plan.
Why the instinct is to cut the wrong thing
Two reflexes cause most of the damage. The first is the pro-rata cut: take the same share from every line and call it fair. It feels fair because nobody is singled out, and that is exactly the problem. Evidence is ignored, and the line that was quietly closing deals loses as much as the one that produced a quarterly deck.
The second is cutting what is furthest from this quarter's sales. Brand and positioning work looks like a luxury when the target is short-term. Binet and Field's analysis of the IPA Databank, published as The Long and the Short of It, found that activation effects fade quickly while brand effects build over months and keep working for longer. That does not mean brand spend is always right. It means cutting it saves money now and shows its cost later, at a time when nobody will connect the two.
Cut what you can't explain first. Protect what you can.
The order I would cut in
| Order | What goes | The test |
|---|---|---|
| First | Lines nobody can say what they bought: tools, subscriptions, retainers that produce activity and reports | Can the owner of the line name one decision or deal it influenced? |
| Second | Duplicates: two channels, suppliers or agencies reaching the same buyer with the same message | If one stopped tomorrow, would any buyer notice? |
| Third | Experiments past their review date, with no result agreed in advance | Was a result and a date written down before it started? |
| Fourth | Spend supported only by the platform's own report, with no match in your CRM or invoices | Does your own count agree with the dashboard? |
| Protect | The decisions work, the channel your records show closing deals, sales follow-up, basic measurement, long-cycle brand spend | Would losing it change what happens on Monday morning? |
The order matters because the first three categories can usually be cut with little visible effect. If nothing changes after they go, that is information too: you were paying for something that was not working. If something does change, you have found a line that was doing more than you knew, and you can restore it with evidence behind the decision.
What to protect, and why
Protect the decisions first. Positioning, the offer and the written audience cost little and every other line depends on them. A company that cuts its way down to execution without them has made the next budget cycle harder, not easier. Why growth stalls even as marketing spend rises describes the same pattern from the other side: more spend, same missing decisions.
Protect the channel your own records show is closing deals, even if it looks unglamorous. Protect sales follow-up, because marketing that generates enquiries nobody answers quickly is a cost, not an investment. And protect basic measurement, because a cut made without being able to see its effect cannot be reversed with any confidence. If your tracking cannot tell you what a cut changed, the first priority is to be able to.
If your buyers take months to decide, reduce brand spend slowly and in steps, with a review date, not in one quarter. The evidence above says its effects build gradually. It also says the cost of removing them arrives gradually, which is why it so rarely gets blamed.
Cut in steps, with a date
Whatever the order, do not cut once and walk away. Take the first tranche, write down what you expect to happen, set a date to look, and then decide on the next. A cut that is reviewed is a decision. A cut that is not is a one-way door. The case for treating every euro spent as a decision to be defended is the argument of Reports, Not Revenue.
Working pattern, from experience — not a published study: in budget reviews I have been involved in, the lines that survive cuts are rarely the ones that earned their place. They are the ones with a champion in the room. The exercise that helps most is to ask every line owner to write one sentence beginning "This bought us…" and to strike out whatever cannot finish it.
- Ask every line owner to complete the sentence "This bought us…" in writing.
- Rank lines by evidence, not by size or ease of cancellation.
- Cut the lines that cannot answer first, then duplicates, then experiments past their date.
- Check platform-reported results against your own CRM before treating them as evidence.
- Protect the decisions, the proven channel, sales follow-up and measurement.
- Cut in steps, write down the expected effect and set a date to check.
