When to pause an ad (and when to bring it back)

Search "when to pause an ad" and you will find thresholds. A click-through floor here, a spend multiple there. The specific numbers vary by author. The confidence never does.

Here is what almost nobody writes about when to pause an ad: the pause is the easy part. Our system has paused 2,340 weak ads across the accounts it operates. It has also brought 875 of them back, not as an apology but as arithmetic: auctions move, active ads tire, and yesterday's pause can become today's best cost per sale. If your pausing process has no path back, you are not optimizing. You are deleting, slowly.

When should you pause an underperforming ad?

When the evidence says its budget would work harder somewhere else. Not when it annoys you, and not because it had one bad week. An ad earns its pause the same way it earned its spend: with data.

The framing matters more than the trigger, though. A pause is a hypothesis, not a sentence. The hypothesis reads: this account performs better without this ad. Like any hypothesis it can be wrong, and the only way to find out is to keep watching after you act.

Most advertisers stop watching. The ad drops into the paused pile, the pile becomes a graveyard, and nobody audits a graveyard. So every wrong pause becomes permanent by default.

When to bring a paused ad back

When the numbers flip. A paused ad is a standing option, not a closed case. Auction costs drift, the surviving ads accumulate fatigue, and the ad that lost in March can beat everything active in July.

My favorite example from our own operations. The system paused an ad and let its replacement take over the traffic. Then it did the thing humans rarely do: it kept measuring its own decision. The replacement was converting 10% worse. So the original ad returned to the account entirely on its own, with no one asking for it. The system watched the consequence of its choice and reversed it. We never lead demos with this, but it may be the most honest thing the product does.

Of the 2,340 ads paused, 875 came back this way. More than a third. That number is not a report card on the pauses; it is a portrait of how much markets move. Each pause was right against the prices of its day, and the days changed. A system that keeps comparing paused ads with live ones captures that movement automatically. A manager who never reopens the paused tab never will.

The same reading that retires can also promote

Pausing is one output of reading performance. It is not the only one. The same system that paused 254 tired keywords also promoted 14 to ad groups of their own, with dedicated ads and budgets, because they were performing too well to stay diluted where they sat. The signal that spots a keyword outgrowing its ad group is the same signal that retires the tired ones. That loop is what Attribution runs across every account it manages, overnight, with no checklist and no mood.

Why we barely pause anything on Meta

A cross-channel note, because the answer to when to pause an ad depends on where the ad lives. On Meta, the system evaluated 4,415 ads and paused 5. That restraint is deliberate. Meta's delivery already prunes on its own, shifting impressions toward what works, and a hasty pause mostly interferes with a machine that was handling it. So we made patience the standing rule for that channel and let the auction do most of the retiring.

Google Ads is different terrain: keyword-level economics, ads competing inside a group. Pausing there is a sharp and useful tool. The rule travels with it, though. The tool is only safe when it cuts both ways.

Both directions, the pause and the return, execute without anyone confirming them. That is deliberate. A pause that waits for your yes runs on your calendar instead of the account's, and the whole lesson of the 875 is that the account's calendar wins. Reversibility is what earns that freedom: nothing the system does here is a one-way door.

So, when to pause an ad? When the data says pause, and only if you are equally prepared to hear the data say come back. Write down what the pause is supposed to achieve, then check whether it did. If you cannot watch the consequence, maybe do not make the cut, or hand the cut to a system that never stops watching.

A pause is a decision about a moment, not about an ad. The 875 that came back are what it looks like when a system refuses to confuse the two.