How do you know when to shut a marketing channel off?
When its marginal cost per booked job has stayed above the ceiling for the work it produces across more than one full revenue cycle, and the jobs it does produce are not unusually valuable. Before cutting, check whether the channel is being under-credited — assisted, branded and repeat bookings often hide inside other sources. Then cut cleanly enough to measure the effect on total bookings, not just that channel's.
Two questions before you cut
First, is the channel genuinely unaffordable, or does it merely look worse than a channel that is harvesting demand this one created? Under-crediting is common enough that cutting on last-touch data alone is a coin flip in some accounts.
Second, is the problem the channel or your handling of it? A source whose leads book poorly because nobody answers after five o'clock is not a media problem. Check answer rate, speed to contact and booking rate on real opportunities before blaming the source — call data answers this in a day.
Cut in a way you can learn from
A clean cut is one you can measure: turn the channel off entirely, or off in a defined set of territories, for a defined window, while holding everything else steady. Slowly starving a channel over months produces data nobody can interpret.
Then watch total booked jobs, not just the channel's line. The whole question is whether the demand it was producing was incremental. If total bookings barely move, you found duplicate demand. If they fall roughly in line with the channel's reported volume, the reporting was honest.
Not all channels shut off the same way
- Paid search and social. Reversible within days. Turning them off is a cheap experiment.
- Purchased leads. Reversible immediately, but relationships and account standing can take time to restore.
- Search and content programs. Not reversible on the same timescale. Ranking decays slowly and recovers slowly, so pausing is not a test — it is a decision with a long tail.
- Brand and community presence. Effects are diffuse and lagged; a short holdout will almost always look like it proved the channel worthless.
What to do with the freed budget
Do not immediately redeploy it. Hold it for at least one cycle so you can observe what the cut actually did. If you cut one channel and raise another in the same week, you have run two experiments simultaneously and can conclude nothing from either.
Then reallocate against marginal cost, not average, and document the decision. That discipline is the difference between a budget that improves each year and one that just churns — the recurring theme across marketing intelligence work.
And record what you expected to happen before you cut. Without that, whatever occurs afterward will be explained as the thing you already believed.
Topics: channel decisions · budget · testing · attribution
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.