What should I pause first when ROAS drops?

Rule out the impostors first

Every item below produces a chart that looks exactly like a performance collapse, and every one has a different fix. Pausing your way through them costs a week and teaches you nothing.

• Attribution lag. Yesterday's ROAS is not final and the last few days are still moving. If the drop is confined to the most recent window, wait.
• A tracking break. A deploy that dropped the pixel, a consent banner change, a server-side event failing quietly. Establish whether conversions fell or whether recorded conversions fell. Those are very different sentences.
• The site. Checkout errors, a payment outage, a page that got slow, a product out of stock. Ads deliver perfectly into a broken funnel and the ad account takes the blame.
• Seasonality and auction pressure. Compare against the same period last year rather than last week. See why is my Meta CPM going up.
• A change somebody made, inside the account or outside it. Read the change history first, then check whether a promotion ended, a competitor started one, or a price went up. Most collapses have a date attached.

If the answer is anywhere in that list, pausing campaigns makes the problem harder to diagnose and does not fix it.

The order to cut in, most confident first

• Nothing at all, if tracking is broken. Fix measurement first. Decisions made on wrong numbers are worse than no decisions.
• Ads with meaningful spend and zero conversions across a window longer than your usual time from click to purchase. This is the least ambiguous waste in any account.
• Search terms and placements, before campaigns. Adding a negative or excluding a placement is surgical. Pausing the campaign is amputation.
• Ad sets that have missed target across enough conversions to be a decision rather than a run of bad luck.
• Whole campaigns last, and only when the layers underneath failed to explain it.

The principle is to cut at the smallest unit that explains the problem. Most accounts do the reverse, pause at the top, and lose the winners that were hiding inside.

What not to do

• Do not pause everything. You destroy the comparison you need to work out what happened, then pay learning costs to restart.
• Do not cut budgets across the board. Broad cuts reset delivery learning on both Meta and Google, which makes next week worse and appears to confirm your fear.
• Do not change five things at once. You will fix it and never know which change did it, which means you cannot do it again.
• Do not react to one day. One bad day is weather.
• Do not do any of this at 11pm. Every catastrophic account decision has a timestamp after 10pm and a reason that sounded excellent at the time.

The check almost nobody runs

Spend concentration. In most accounts a small number of ads carry most of the waste, and a ranked list of spend against conversions answers the question faster than any dashboard view. It is also the report nobody builds, because it means pulling ad-level data and sorting it.

The second underused check is comparing the same window across your other platforms. If Meta and Google both fell in the same week, the cause is probably not inside either account. Answering that needs a tool that can see both at once, which is the whole argument for connecting one.

A twenty minute routine

• Confirm the conversion event is still firing and that the numbers still reconcile to the store.
• Pull spend and conversions by ad for the last 14 and 30 days, sorted by spend.
• Read the change history for the week the drop started.
• Compare the same window on your other channels, and against the same weeks last year.
• Then act at the smallest unit that explains it, one change at a time.

Most of that is retrieval. Muze reads it across Meta, Google, Amazon and read-only Shopify in one session, and wasted spend detection produces the ranked list directly. Pauses preview before they run. The longer walkthrough is how to find wasted ad spend.

Related questions

How long should I wait before pausing an ad?

Long enough to accumulate spend worth several times your target cost per acquisition with nothing to show for it, and longer than your usual time from click to purchase. Pausing on day two of a considered purchase cycle is pausing on no information.

Should I pause or lower the budget?

Lowering is usually the reversible option, and reversible is what you want while you are still diagnosing. Pausing an ad set that turns out to be fine also costs you the delivery learning it had accumulated.

My ROAS dropped but revenue is flat. What happened?

Usually spend rose without conversions following, or the attribution window shifted what got credited where. Check spend and conversions separately before touching anything. A ratio moving tells you nothing about which side of it moved.

Should I pause everything and start again?

Almost never. You lose delivery learning, the historical comparison and the ability to diagnose, and you keep whatever underlying problem caused the drop. Rebuilds feel decisive and are usually avoidance.

Can AI decide what to pause?

It ranks candidates well, because that is retrieval and sorting. Deciding is different work: it needs your margin, your stock position and your read of the week, none of which are in the ad platform. With Muze every pause previews before it runs.

Keep reading

Other answers

See every answer