How to · Google Ads diagnostics

How to find underperforming Google Ads campaigns

A campaign can look underperforming for four different reasons, and only one of them is a bad campaign. It might be spending on the wrong queries, starved of impression share, below the conversion volume its bidding strategy needs, or simply too new to judge. Work out which before you cut anything, because pausing a starved campaign and pausing a failing one look identical on the way in and very different a month later.

Mayur Magar in ↗

Updated 2026-10-05 · 10 min read

Before calling a campaign underperforming, check:

Conversion volume against what the bidding strategy needs
Campaign age and whether it has matured
Query mix — how much spend sits in losing and invalid tiers
Impression share, split into rank-lost and budget-lost
Whether the target it is judged against is realistic
Whether the conversion metric matches the campaign type
Whether tracking is intact for that campaign's conversion action

Four reasons a campaign looks bad

Each needs a different response. Only the last one is a pause.

Pattern	What it really is	The response
Under 30 conversions in 30 days on Search	Not enough signal for smart bidding	Consolidate or change strategy, do not judge yet
PMax under 30 days old or 50 conversions	Immature	Wait. There is nothing to read
High budget-lost impression share, CPA on target	Starved, not failing	This is a scale candidate, not a cut
A third of spend in losing and invalid queries	Leaking, not failing	Prune the queries, then re-measure
Mature, sufficient volume, clean queries, CPA above target	Genuinely underperforming	Fix the targets or cut

The process

Step 1

Check volume before anything else

Smart bidding needs roughly 30 conversions in 30 days on Search and around 50 on Performance Max and Shopping. Below those levels the strategy cannot optimise, and the campaign will underperform for structural reasons regardless of how well it is built. This is not a campaign quality problem and cutting it will not help.

Do not judge Search on fewer than 30 conversions or PMax and Shopping on fewer than 50. Below that, you are reading variance rather than performance.

Step 2

Check maturity

A Performance Max campaign under 30 days old — ideally under 60 — has not had time to learn. Judging it at two weeks and pausing it is a reliable way to never have a working PMax campaign. The same logic applies more loosely to any campaign that has recently had a significant structural or bidding change, since those reset learning.

Step 3

Classify the queries before blaming the campaign

Pull the search terms for the campaign and assign them to tiers. If a large share of spend is sitting in losing or invalid queries, the campaign is not underperforming so much as leaking — and the same campaign will look competent once the leak is closed. Negate the invalid terms, cap or cut the losing ones, and re-measure before passing judgement.

Step 4

Separate starved from failing

A campaign hitting its CPA target but losing a large share of impression share to budget is one of the best things in the account, not one of the worst. It is being prevented from spending on demand it can convert. Look at budget-lost impression share alongside CPA before reading a low conversion count as weakness.

Budget-lost high, CPA at or below target, spend concentrated in profitable queries — scale candidate.
Rank-lost high, CPA above target — do not raise bids. Tighten queries first.
Both low and CPA above target — this is the genuine underperformer.

Step 5

Check it is being judged on the right metric

An account configured for calls, store visits or direction requests cannot be judged on purchase CPA or ROAS. Nor can a lead-gen account be judged on ROAS unless offline conversion import is configured, because the conversion value will be zero and every ROAS figure derived from it will be meaningless.

Step 6

Confirm tracking for that specific conversion action

A campaign whose conversions fell to zero while spend continued is the signature of broken tracking rather than collapsed performance. Check the conversion action is still firing at its expected daily volume before treating the campaign as failed.

Step 7

Then act, in the right order and one change at a time

Once a campaign is genuinely underperforming on clean data, fix it in sequence: prune the junk queries, reclassify what remains, recompute the pressure, then move bids or budget. Do not stack changes — changing budget, bids and structure together makes it impossible to learn anything from the result.

Scaling caps apply in both directions: no more than 20% per step on Search and Shopping, and no more than 50% on a mature PMax campaign. And budget is a campaign-level control — you cannot shift money between keywords, queries, devices or asset groups, so a recommendation phrased that way is not executable.

A worked example

Illustrative example. The numbers below are made up to show the method. They are not a client account and not a benchmark.

Four campaigns flagged as underperforming on a monthly review. Only one actually is.

Campaign	Spend	Conversions	CPA vs target	Verdict
Brand — Search	$3,100	19	−52%	Fine. Low volume is normal for brand
Generic — Search	$14,800	41	+88%	Leaking: 31% of spend in losing and invalid queries
PMax — new	$6,200	22	+34%	18 days old. Too early to judge
Shopping	$9,400	63	−11%	Starved: 41% impression share lost to budget

Only the generic Search campaign needs work, and the work is query hygiene rather than a pause. Nearly a third of its spend is on terms that lose money or should never have matched; cleaning that up will move CPA without touching bids.

The Shopping campaign is the opposite of a problem. It beats target CPA and is losing 41% of impression share to budget, which means demand it can convert profitably is going unserved. It is the best scale candidate in the account.

The new PMax campaign should be left alone. At 18 days and 22 conversions it is under both the maturity and volume gates. Any verdict here is noise.

The brand campaign's low conversion count is structural. Brand terms are cheap and limited in volume; judging it on absolute conversions rather than CPA misreads what it is for.

What not to cut

The campaigns most often cut wrongly:

Anything under the volume gate. Below 30 conversions on Search or 50 on PMax, you are reading variance.
A PMax campaign under 30 days old.
A campaign starved of impression share while hitting target. That is a scale candidate wearing a disguise.
A brand campaign judged on conversion volume rather than efficiency.
A campaign whose conversions went to zero while spend continued — check tracking before concluding anything.
A lead-gen campaign judged on ROAS with no offline conversion import configured.
A campaign that just had a structural change. Learning has reset; give it time.

How to automate this

Separating starved from failing means holding four things together per campaign — conversion volume against the bidding threshold, age against the maturity gate, query mix by spend share, and impression share split by cause. Doing it across an account by hand is why it tends to get done quarterly instead of weekly.

GoMarble applies those gates continuously, so a campaign is only flagged once it has actually cleared the bar for a verdict.

The shape of it

Connect the Google Ads account
Campaigns are checked against volume and maturity gates first
Query mix and impression-share cause are computed per campaign
Only campaigns with enough signal to judge are flagged
Recommended fixes arrive in execution order — AI Agents apply what you approve

A tool that flags every campaign above target CPA produces a list you learn to ignore. The gates are the point.

Underperforming campaign checklist

☐ Check conversion volume against the bidding strategy threshold
☐ Check campaign age against the maturity gate
☐ Pull search terms and classify them by tier
☐ Compute spend share in losing and invalid tiers
☐ Split impression share into rank-lost and budget-lost
☐ Compare CPA against target before reading conversion count
☐ Confirm the metric matches the campaign and account type
☐ Confirm the conversion action is still firing normally
☐ Check whether a recent change reset learning
☐ Fix in order: prune, reclassify, recompute, then move bids or budget
☐ Keep changes to one lever at a time, capped at 20% on Search and Shopping

Questions people actually ask

How do I know if a campaign is underperforming or just new?

Check it against two gates before anything else. On Search, fewer than 30 conversions in 30 days means smart bidding does not have the signal it needs. On Performance Max and Shopping, the bar is around 50 conversions, plus at least 30 days of age and ideally 60. A campaign below either gate is not underperforming — it is unjudgeable, and the only honest verdict is to wait or consolidate it into something with enough volume.

Should I pause a campaign that is above target CPA?

Not before checking why. If a third of its spend is going to junk queries, the campaign is leaking rather than failing, and query hygiene will move CPA without any structural change. If it is losing impression share to rank while CPA is already above target, raising bids makes it worse and tightening queries is the fix. Pause is the right answer only for a mature campaign with sufficient volume and clean queries that still misses target.

What is the most common reason a Google Ads campaign underperforms?

Query leakage — budget going to searches that were never going to convert, usually through broad or loose phrase matching. It shows up as high CPA, poor impression share and rising CPC all at once, which makes it easy to misdiagnose as three separate problems. Classifying search terms by spend share usually finds it in minutes.

Can I fix an underperforming campaign by increasing the budget?

Only when it is starved rather than failing — hitting its CPA target while losing significant impression share to budget, with spend concentrated in profitable queries. In that specific case more budget buys more of something that works. In every other case it funds the existing problem faster. And scale in steps of no more than 20% on Search and Shopping, so you can tell whether it worked.

Why did conversions drop to zero while spend kept running?

That pattern is almost always tracking rather than performance. Real performance declines are gradual and partial; conversions going to exactly zero while impressions and clicks continue is the signature of a broken conversion action or tag. Check that the conversion action is still firing at its expected daily volume before touching the campaign at all.

Where this method comes from

The gates, the query classification and the starved-versus-failing distinction on this page come from the Google Ads methodology GoMarble's agents run against live accounts — the smart bidding conversion minimums, the PMax maturity gate, the Q-tier classification, the rank versus budget pressure split, the scaling step caps and the guardrail that budget is a campaign-level control rather than something that can be moved between keywords or asset groups. Thresholds are ours rather than published Google guidance.

Find the ones that are actually failing.

Connect Google Ads and GoMarble applies the volume and maturity gates before flagging anything.

Start free trial Book a demo