Meta Ads ROAS Dropped Suddenly: What to Check First

Troubleshooting · 6 min read · 3 Nov 2026 · By the Cofoundr team

When Meta ads ROAS dropped suddenly on a client account, check in this order: whether real revenue actually fell, whether attribution or tracking changed, what was edited in the last seven days, and only then creative fatigue and auction pressure. Most panic comes from a reporting change rather than a real performance change, so confirming the drop against backend revenue saves you from pausing campaigns that were fine. This guide gives you that order, the evidence to look at in each step, and what to tell the client while you investigate.

Did Meta ads ROAS really drop, or did the reporting change?

Start outside Ads Manager. Pull the client's revenue from Shopify, Stripe or their CRM for the same days and divide it by total Meta spend for a blended number. A practitioner checklist from Niblin makes the same point: if revenue is stable but Meta ROAS dropped, you have an attribution problem, not a performance problem. Niblin is a third party vendor, so treat its checklist as a practitioner view rather than Meta guidance.

If blended revenue fell by roughly the same amount as Meta ROAS, the drop is probably real and you keep going down this list. If revenue held steady, skip ahead to attribution and tracking before you touch a single ad set.

Write the two numbers side by side in your notes: Meta-reported ROAS and blended ROAS, for this week and the previous one. That table tells you in a minute which branch of the checklist you are on, and it gives you something concrete to send the client. Remember that blended revenue includes sales from email, organic and search, so a stable blended number does not prove Meta is fine. It only proves the business did not suddenly stop selling.

Has the attribution window or comparison setting changed?

ROAS is a ratio of credited conversions to spend, so anything that changes how many conversions get credited moves it. Check the attribution setting on the columns you are looking at, and check that this week and last week are compared with the same setting. Meta has changed attribution options over time, so read Meta's current Business Help Center documentation rather than relying on memory.

Also check the date range. Recent days can still be filling in as conversions are reported late, so a ROAS number for the last 48 hours often looks worse than it ends up. Compare completed periods, not partial ones. If the client keeps asking why Meta and their store disagree, our guide on duplicate conversions covers one common cause of the mismatch.

Is tracking still sending events?

A broken pixel or Conversions API connection makes ROAS fall without a single ad changing. Open Events Manager and confirm the purchase event is still arriving, that the count is in line with the store's order count, and that nothing was changed on the site, checkout or tag manager recently. Site theme updates, checkout app changes and consent banner changes are frequent culprits on client sites you do not control.

A useful signal is spend staying normal while reported purchases fall off a cliff on one day. That pattern usually points to tracking rather than demand. Tell the client early so their developer can check, and see facebook pixel not firing for the checks.

Compare the event count in Events Manager with the store's order count for the same days. Some gap is normal because of consent choices, browser limits and attribution rules, but a gap that suddenly widens is the signal. Check that event deduplication is still working if both the pixel and Conversions API are active, since double counting can also make ROAS jump before it falls.

What was edited in the last seven days?

Ask who changed what. Account activity history shows budget, audience, creative and bid changes by person. Niblin lists budget changes above 20 percent, and changes to audience, creative, bid, optimization event or placements as things that can restart learning. Other sources disagree on the exact budget threshold and on whether pausing counts, so confirm the current list of significant edits in Meta's Help Center instead of treating any single number as fixed.

If an ad set re-entered learning, results are expected to be unstable for a while. Judge it after the learning status settles, not on day two. If it is stuck, the fix is usually to consolidate ad sets or optimize for an event that happens more often; our learning limited fix walks through the options.

Is creative fatigue or frequency the cause?

Look at frequency and click-through rate by ad over the last 14 days. Niblin suggests frequency above 3 in a week and a CTR drop of 20 percent or more from baseline as signs it is time to refresh creative. These are rules of thumb from one vendor, not Meta thresholds, and they vary by audience size and budget.

The more reliable test is whether the ad that carried the account has declined while a newer ad is still healthy. If the top spender is falling and nothing new is being tested, fatigue is likely. If everything fell on the same day, look at tracking or the auction instead. Our guide on facebook ads frequency too high covers what to do next.

When you refresh creative, add new ads to the existing ad set or a test campaign instead of rebuilding the account. That keeps what has already been learned and lets you compare old and new ads on the same audience. Avoid pausing the whole set of ads that used to work until the new ones have enough data to replace them.

Has auction pressure or a seasonal shift changed your costs?

If tracking is fine and nothing was edited, look at CPM and conversion rate separately. A CPM jump with a steady click-through and conversion rate means you are paying more for the same attention, which is common around busy sale periods when many advertisers compete. A steady CPM with a falling conversion rate points to the landing page, the offer, stock or price rather than Meta.

Break the account down by placement and device as well. A single placement or platform quietly taking more budget can drag the blended number down. If CPM is the problem, meta ads cpm too high goes deeper.

What should you tell the client while you investigate?

Send a short, factual note the same day: what you saw, which checks you have done, and when you will report back. Do not promise a fix before you know the cause. Wait several days before judging changes, because reacting to every daily wobble makes results less stable.

Keep a simple log of what you checked and when, so the next time ROAS dips you can compare against it. Agencies running many accounts also set up alerts so a drop is caught the same day rather than at the weekly report. Cofoundr ranks client accounts daily by what needs attention and can alert on spend changes, but a spreadsheet and a morning routine work too.

If the drop is real and you cannot find a single cause, say so. Clients usually accept 'we have ruled out tracking and attribution, and we are testing new creative this week' far better than a confident guess that turns out wrong. Set a date for the next check and stick to it.

Frequently asked questions