The short answer
The instinct is to cut whatever had the worst cost per lead last month. Where the sales cycle is longer than a month, that number describes a period that has not finished producing results.
Cut by intent instead: broad prospecting first, high-intent search last.
A sensible order
- Broad prospecting with no intent signal.
- Upper funnel awareness activity.
- Retargeting beyond a short window.
- Non-branded search on generic terms.
- High-intent search and branded terms, last.
Branded search looks like an easy cut because those people would find you anyway. Some would; the ones who see a competitor's ad instead would not.
Do not judge on an incomplete period
| Sales cycle | Judge on |
|---|---|
| Days | Last month is fine |
| Weeks | Look back a quarter |
| Months | Look back further, expect noise |
| Highly seasonal | Compare with the same period last year |
| Very few sales | Use qualified enquiries, not closed deals |
The last row matters for high value services. With a handful of sales a year, closed deals are too sparse to judge a campaign by, and qualified enquiries are the honest proxy.
Protect the things that take time to rebuild
- Anything feeding a long pipeline
- Campaigns that have accumulated useful optimisation history
- Activity supporting an imminent seasonal peak
- Anything you would struggle to restart quickly
- Tracking, which costs little and is painful to rebuild
Turning off a campaign that has learned a lot and restarting it later is not free. The restart begins without that history.
Reduce before removing
Lowering a budget keeps a campaign alive and keeps the data coming. Switching it off entirely ends both, and the decision is harder to reverse.
Where the cut is temporary, reducing is almost always the better option even though it looks less decisive.