How to · Google Ads diagnostics

How to analyze Google Ads performance

Google Ads analysis has a correct order, and most accounts are analysed in the wrong one. Classify the search terms first, because query quality determines whether any other metric means anything. Then read impression share as two separate problems — losing to rank and losing to budget need opposite fixes. Only then look at CPC movement, and only one of its three causes justifies raising a bid.

Mayur Magar in ↗

Updated 2026-10-05 · 12 min read

To analyze a Google Ads account, work through:

What the account is actually configured to do — purchases, leads, calls or footfall
Search term classification, from profitable through to junk
Spend share by query tier — where the money actually goes
Impression share split into rank-lost and budget-lost
CPC trend week over week, and which of three causes explains it
Conversion volume against the thresholds smart bidding needs
Quality Score components where CPC is high

Classify every search term before anything else

This classification drives every later decision. Until you know what share of spend sits in each tier, impression share and CPC figures cannot be interpreted.

Tier	Definition	What to do
Q1 — Profitable	2+ conversions, or CPA at/below target, or ROAS at/above target	Protect and isolate into exact or tight phrase match
Q2 — Promising	No conversions, cost under half the target CPA, no junk signals	Leave running — it has not had its chance yet
Q3 — Unproven	No conversions, cost between half and one times target CPA	Watch closely; approaching the decision point
Q4 — Losing	Cost at/above target CPA with no conversions, or CPA 1.5× target, or ROAS below 0.7× target	Cap or cut
Q5 — Invalid	Contains free, cheap, diy, how to, job, salary, course, pdf, meaning, used, repair	Negate immediately, regardless of cost

The analysis, in order

Step 1

Establish what the account is for

Before computing a single CPA, check what the conversion actions actually are. An account built around store visits, phone calls or direction requests cannot be judged on purchase-derived CPA or ROAS, and doing so produces an analysis that is confidently wrong from the first table. If the account is footfall or awareness, swap the headline metrics for cost per call, cost per direction, CPM and reach, and say so explicitly.

Step 2

Classify the search terms and compute spend share per tier

Pull search terms for the last 30 days sorted by cost, and assign every one to a tier using the table above. Then compute what share of spend sits in each. This single number — the proportion of budget in Q1 versus Q4 and Q5 — tells you more about the account's health than any headline metric.

Q5 terms are eliminated regardless of cost. A query containing "job", "salary" or "free" is not going to convert into a sale no matter how cheap the click is.

Step 3

Read impression share as two separate problems

Low impression share on its own means nothing. What matters is why you are losing it, because the two causes have opposite fixes. Losing to rank is a bidding or relevance problem; losing to budget is a money problem. Act on it only when Q1 share is at least half of spend and the impression-share loss actually explains the gap.

Rank pressure is high if rank-lost impression share is 30% or more, or impression share is at or under 50% with rank-lost exceeding budget-lost.
Budget pressure is high if budget-lost is 20% or more, exceeds rank-lost, CPA is at or below target, and at least 60% of spend sits in Q1 to Q3.
If CPA is above target, do not raise bids to fix rank loss. Tighten the queries first — promote Q1 to exact, cut the Q3 and Q4 leakage.

Step 4

Diagnose CPC movement before touching a bid

Rising CPC has three distinct causes and only one of them justifies bidding more. Compare this week's average CPC against last week's and match the pattern.

CPC up with rank-lost impression share up — genuine competition. A bid increase is defensible, but only if CPA is at or below target.
CPC up with new junk terms appearing — match type leakage. The fix is negatives and tighter match types, not bids.
CPC up with CTR down — ad relevance has degraded. Fix the copy. Raising bids here pays more for an ad that is working less well.

Step 5

Check you have the volume smart bidding needs

Smart bidding strategies need signal to work. On Search, that means roughly 30 conversions in 30 days; on Performance Max and Shopping, around 50. Below those levels the strategy lacks the data to optimise and the account will underperform for reasons that have nothing to do with your targeting or creative.

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

Step 6

Decompose Quality Score where CPC is high

Quality Score is not a number to optimise directly, but its three components tell you where the cost is coming from: expected CTR, ad relevance and landing page experience. Identify which one is dragging and fix that specifically. Raising bids to compensate for poor quality is paying a permanent tax rather than fixing the cause.

Step 7

Assess PMax separately, and only once it is mature

Performance Max cannot be judged on the same timeline as Search. Give it 30 days at minimum — ideally 60 — or 50 conversions before drawing conclusions. Once mature, compare its ROAS against the Search campaigns as a ratio rather than in isolation.

Below 0.5× Search ROAS — poor. Monitor; consider pausing if it does not improve.
0.5–0.9× — below par. Fix assets and negatives before any scaling.
0.9× and above — healthy, and eligible to scale.
Asset-level data comes as performance labels rather than cost figures. Replace LOW assets with variations of BEST ones; PENDING means wait.

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.

A Search account with a $45,000 monthly budget and a $60 target CPA. The headline looks like an impression share problem.

Tier	Share of spend	Conversions	Read
Q1 — Profitable	38%	412	Carrying the account
Q2 — Promising	11%	0	Fine — still early
Q3 — Unproven	14%	0	Approaching the decision point
Q4 — Losing	27%	9	The actual problem
Q5 — Invalid	10%	0	Should never have run

The account is losing 34% of impression share to budget, which looks like an obvious case for more money. It is not. Budget pressure only counts as high when at least 60% of spend sits in Q1 to Q3 — here it is 63%, marginal, and CPA is above target.

37% of spend is in Q4 and Q5. That is roughly $16,600 a month going to queries that lose money or should never have matched. Adding budget before fixing that funds the leak.

The order is forced by the data. Negate the Q5 terms, cap or cut Q4, then recompute. Impression share will improve on its own as the same budget stops being spent on junk.

Only after that does the budget question become answerable, because the Q1 share and the CPA will both have moved.

What not to conclude too quickly

Google Ads analysis fails in a handful of recognisable ways:

Treating low impression share as a budget problem. Classify the queries first — it is often a leakage problem wearing a budget costume.
Raising bids to fix rank loss while CPA is above target. That buys more of something already unprofitable.
Chasing absolute top of page without CVR proof. If conversion rate at the top is not meaningfully better, the premium is vanity.
Judging PMax before it has matured. Under 30 days or 50 conversions, there is nothing to judge.
Applying CPA and ROAS to a footfall or awareness account.
Raising bids and budget at the same time. You will not know which worked.
Reading cost fields without converting from micros. A factor of a million is an easy way to produce a very wrong report.

How to automate this analysis

Classifying several hundred search terms, computing spend share per tier, splitting impression share into its two causes, comparing CPC week over week and checking conversion volume against bidding thresholds is a half-day exercise done properly — and it needs redoing every week to stay useful.

GoMarble runs the classification and the pressure diagnosis against the live account, so the output is the recommended move in execution order rather than a set of exports to interpret.

The shape of it

Connect the Google Ads account
Ask for the analysis, or put it on a weekly schedule
Search terms are classified and spend share computed per tier
Impression share is split into rank and budget pressure
Recommended actions come in execution order — AI Agents apply what you approve

The execution order matters as much as the findings: prune the junk, reclassify, recompute pressure, then move bids or budget.

Google Ads analysis checklist

☐ Confirm what the account's conversion actions actually are
☐ Swap in the right headline metrics for footfall or awareness accounts
☐ Pull 30 days of search terms sorted by cost
☐ Classify every term into Q1–Q5
☐ Compute spend share per tier
☐ Negate Q5 terms regardless of cost
☐ Split impression share into rank-lost and budget-lost
☐ Only act on impression share when Q1 share is 50%+ of spend
☐ Compare CPC week over week and identify which of three causes applies
☐ Confirm conversion volume meets smart bidding thresholds
☐ Decompose Quality Score where CPC is high
☐ Check PMax maturity before judging it, then compare as a ratio to Search
☐ Convert cost fields from micros before reporting anything

Questions people actually ask

What should I look at first when analysing Google Ads?

The search terms, not the headline metrics. Query quality determines whether anything else means anything — an account with a third of its spend on junk queries will show poor CPA, poor impression share and rising CPC, and all three are symptoms of the same cause. Classify the terms and compute spend share per tier before you form any other opinion.

Is low impression share bad?

On its own it tells you nothing. What matters is why you are losing it and what your queries look like. Losing impression share to budget while most of your spend is on profitable queries is a genuine case for more money. Losing it while a third of spend goes to junk queries is a reason to clean up, and adding budget would just fund the waste faster.

When should I increase bids in Google Ads?

When rank pressure is high, most of your spend is on profitable queries, and CPA is at or below target — and specifically when rising CPC is accompanied by rising rank-lost impression share, which indicates genuine competition. If CPC is rising because junk queries have started matching, or because CTR has fallen, bidding more makes it worse. And never raise bids and budget in the same move; you will not know which one worked.

How long should I wait before judging Performance Max?

At least 30 days, ideally 60, or 50 conversions. PMax needs more signal than Search because it spans more inventory, and judging it early produces decisions based on noise. Once it has matured, compare its ROAS to your Search campaigns as a ratio: below half of Search ROAS is poor, and at or above 0.9× is healthy and eligible to scale.

Why is my CPC rising?

Three causes, and they need different responses. If rank-lost impression share is also rising, competition has increased and a bid rise may be justified if CPA allows. If new irrelevant terms have started appearing, match types are leaking and you need negatives. If CTR has fallen, ad relevance has degraded and the fix is the copy. Only the first justifies paying more.

Where this method comes from

The query classification, the impression-share pressure logic, the CPC diagnosis and the maturity gates on this page come from the Google Ads search analysis methodology GoMarble's agents run against live accounts — including the Q1–Q5 tiers and their thresholds, the rank versus budget pressure definitions, the three CPC inflation cases, the smart bidding conversion minimums and the PMax maturity and ROAS-ratio tables. Thresholds are ours rather than published Google guidance.

Run the analysis on your own account.

Connect Google Ads and GoMarble classifies the queries, splits the impression share and names the move.

Start free trial Book a demo