When to pause a campaign (and when to wait another week)

Pausing too early kills campaigns that would have worked after learning. Pausing too late burns budget on something that was never going to work.

The difference between the two is having a criterion set before you start — because mid-month, with a bad number on screen, the decision turns emotional.

First of all: is the campaign the problem?

Half the campaigns paused for "not performing" had broken measurement. Thirty seconds of checking avoids switching off something that was working:

• Has the conversion event fired in the last 3 days?
• Is the form submitting? Run a test.
• Are leads arriving in the CRM, in the right queue?
• Did someone change the landing page?

If conversions dropped to zero overnight, it is almost never the campaign. A bad campaign produces bad results, not no results.

The cut-off criteria

A new campaign with no conversions

Use spend, not the calendar:

pause when accumulated spend passes
2 to 3× the target cost per conversion, with zero conversions

If the target is $200 per lead and the campaign spent $600 without one, the probability of it delivering at target cost is low. That is not impatience, it is evidence.

A new campaign with expensive conversions

Here patience is justified. Exit learning before judging: 7 to 14 days or around 50 conversions in the ad set, whichever comes later.

Judging on day three is judging the algorithm while it learns, not the campaign.

An old campaign that got worse

Two weeks against the two before, not one day against another.

The last row is the most important and the least obeyed. A campaign swinging within range needs no intervention — and intervening is the most common cause of self-inflicted instability.

A campaign that delivers leads and no sales

This one requires a different decision, because the problem may not be the campaign.

Before pausing, check: are the leads being contacted? How fast? Is this campaign's close rate worse than average, or is the whole funnel slow?

If its close rate is worse than average with equal handling, it is the campaign: it brings the wrong audience. If the close rate matches and sales do not come, the problem sits after media — and pausing does not fix it.

What to do instead of pausing

Pausing is the most drastic action and not always the right one.

Cut the budget. Half the spend keeps the campaign and its learning alive while you test something else with the rest.

Swap the creative. If the structure is right and the piece tired, the piece is the problem.

Narrow the audience. A campaign with high CPL is often delivering to too many people.

Change the optimization event. If the campaign delivers volume and not quality, the event may be wrong. See quality events vs volume events.

The cycle that destroys accounts

Switching on and off repeatedly is worse than either decision.

Every reactivation puts the ad set back into learning. Three cycles in a month and the campaign spent the whole month learning, never delivering with stability — and each cycle produces a bad number that motivates the next cycle.

If you decided to pause, pause properly. If you will reactivate, treat it as a new campaign: starting budget, learning window, fresh cut-off criterion.

The note that stops the repeat

A campaign paused without a note gets recreated in six months, identically, by whoever arrived after you.

PAUSED 7 Nov · acme_tofu_leads_consult_2026-09
Spent        $4,200 over 23 days
Result       31 leads, 1 sale
Why          CPL fine ($135) but 3% close rate against a 9% average
Hypothesis   audience too broad, brought curiosity
Next         test the same creative with a narrow interest audience

The "next" line is what turns a paused campaign into learning. Without it, you just spent $4,200.

The summary

Check measurement before blaming the campaign. Use accumulated spend against target cost, not days on the calendar. Give a new campaign time to exit learning. Prefer cutting budget, swapping creative or narrowing the audience before pausing. Never enter the on-off cycle. And write down the reason, with a hypothesis for the next test.

For the weekly routine where this decision appears, see a weekly optimization routine.