It happens on a Tuesday. The account that has been steady for months reports a return on ad spend a third lower than last week, nothing obvious has changed, and the first suggestion in the room is to reduce the budget until it recovers.

Sometimes that is correct. More often it makes the problem permanent, because cutting spend on a learning-based bidding system removes the very signal it needs to recover. Before you touch a budget, work through these five in order. They take about an hour and they tell you which kind of drop you are looking at.

1. Is it a reporting drop or a revenue drop?

Start outside the ad account. Look at total orders or total qualified leads in your own system — the shop admin, the CRM, the bank — for the same period. If the business number is flat and only the platform number fell, you have a measurement problem, not a performance problem.

This is by far the most common cause after a platform or browser update, and the fix is tracking work, not budget work. If you cut spend in response to a measurement drop you will turn a reporting artefact into a real decline.

2. Did the mix change underneath the average?

Blended ROAS is an average, and averages hide the thing that actually moved. Break the period down by campaign, by audience, and by placement, then look for a segment whose share of spend grew. A single campaign scaling into a worse audience will drag the whole account down while every other line stays healthy.

This is the second most common cause and it is entirely self-inflicted: an automated bidding strategy found somewhere cheap to spend, and cheap is not the same as valuable.

An average that falls is not evidence that everything got worse. It is usually evidence that something got bigger.

3. How old is the creative that is carrying the spend?

Sort your ad sets by spend and look at the launch date of the top three. If the answer is “months ago”, you are not looking at a platform problem. Frequency has climbed, the audience has seen the ad enough times to stop reacting, and efficiency decays on a curve that looks exactly like an algorithm change.

The tell is that impressions and reach stay flat while click-through rate slides gently over several weeks rather than dropping in a single day. Platform changes tend to be cliffs. Fatigue is a slope.

4. What did the competition do?

Auction pressure is invisible from inside your own account, but it is not invisible. Cost per thousand impressions rising across every campaign at once, on the same days, with no change on your side, is a market signal rather than an account signal. Seasonal peaks, a competitor’s launch, or a category-wide event will all do it.

If this is the cause, cutting spend hands the season to whoever is still bidding. The right response is usually to defend the highest-intent inventory and let the expensive upper funnel breathe.

5. Did anything change on the other side of the click?

Ads do not convert. Landing pages, checkouts and forms convert. A theme update, a new consent banner, a slower page, a payment method that started failing on mobile — any of these will present in the ad account as a performance drop, because the platform can only see the conversion event that never arrived.

Check conversion rate from click, not just cost per acquisition. If clicks held and conversion rate fell, the problem is downstream and no amount of bidding work will fix it. That is conversion and analytics territory.

What to actually do once you know

Only one of those five causes is fixed by spending less, and it is the auction one — and even then, selectively rather than across the board. A measurement drop needs tracking work. A mix problem needs structure and exclusions. Fatigue needs new creative on a cadence rather than one heroic batch. A downstream problem needs the page fixed.

Write down which one you think it is before you change anything, then change one thing. The discipline is not glamorous, but it is the difference between an account that recovers in three weeks and one that never quite does.