paid · 9 min read
Google Ads are not converting: work out which of the three problems you have
"Not converting" describes three separate failures — clicks that never become enquiries, enquiries nobody can use, and conversions the platform records that the business never sees. They have different causes and different fixes, and the account will not tell you which one you are looking at unless you ask it properly.
Written by Zubair Afzal, FounderUpdated
First, be precise
Three different things people mean by "not converting"
These get described in the same sentence and diagnosed as one problem. They are not one problem, and the first job is deciding which one you are actually looking at. It takes about an hour and it prevents weeks of work in the wrong place.
- Clicks arrive and nothing is submitted.
- The platform shows traffic and few or no conversions. Either the people arriving do not want what is being offered, the page does not answer the search they made, or the form is failing silently. This is the version where the fault is most likely to sit between the query and the page rather than in bidding or budget.
- Enquiries arrive and none of them are usable.
- Forms are being submitted, so the mechanics work. The people submitting them are wrong: out of area, out of budget, looking for a service you do not offer, or looking for a job. This is a query-selection problem almost every time, and the evidence sits in the search terms report rather than in the conversion column.
- The platform reports conversions the business has never seen.
- The account looks healthy and nobody in sales recognises the numbers. Something is being counted that is not a lead — a page view, a scroll, a duplicate submission, a click on a telephone number that never connected. This is the most common and the most damaging, because automated bidding has been optimising towards it for months.
- Enquiries arrive, are usable, and never become customers.
- Worth separating because the advertising is doing its job. The loss is in the offer, the pricing against alternatives, or in how long somebody waits for a reply. This is not a Google Ads problem at all, and treating it as one produces expensive account changes that cannot possibly help.
The sequence
Six checks, in the order that finds the fault fastest
Work through these in order and stop when something explains what you are seeing. Do not change bids or pause anything until you have finished.
Open the conversion actions table and read what is being counted
Look at every conversion action, what it counts, and whether it is included in the primary goal set that bidding optimises towards. Check the counting setting — whether every submission counts or only one per interaction — and look for anything firing on a page load. This is where the majority of unexplained accounts are explained.
You get: A list of every conversion action and whether it represents a real business outcome
Reconcile one week against reality
Take a single completed week. Count actual enquiries from your inbox, call log and CRM, and compare with what the platform reported for the same days. A large gap in either direction is a measurement fault and it invalidates everything else you were about to analyse.
You get: A written comparison of platform conversions against real enquiries for one week
Read the search terms report, sorted by cost
Take the last thirty to ninety days, sort by cost descending, and read the top hundred queries. Mark each one relevant or not. This is the single most informative hour available in a search account, and it tells you whether you are buying the intent you think you are buying.
You get: A marked list of high-cost queries with negatives identified
Check where the money actually went
Segment by network to see how much was spent on search partners and display expansion rather than on Google search itself. Check match types, and check whether a Performance Max campaign is absorbing queries a search campaign was meant to serve. Budget frequently leaves through a door nobody opened deliberately.
You get: A spend breakdown by network, match type and campaign type
Follow one query through to the page as a customer would
Pick a high-cost relevant query. Search it on a phone, click your own advertisement, and go through the whole thing without any inside knowledge. Does the page say back what you searched for, within the first screen? Is there one obvious action? Does the form actually send, and where does it go?
You get: A recorded walkthrough of the query, ad and landing page on a mobile device
Measure how long an enquiry waits
Take the last twenty enquiries and record the time between submission and the first genuine human response. Paid search enquiries are frequently comparing several suppliers in one sitting, and a reply the following morning arrives after the decision has been made. This costs nothing to fix and is almost never measured.
You get: A response-time figure for the last twenty enquiries
The same symptom, and the evidence each cause leaves behind
Once you have run the checks, this maps what you found to where the fault sits. The point is to avoid working on the account when the problem is elsewhere.
| Dimension | What you will see | Where the fault is |
|---|---|---|
| Conversions high, sales unaware of them | Platform numbers far exceed enquiries you can find, or the same enquiry appears several times. | Conversion setup. Wrong trigger, wrong counting setting, or a soft action feeding the bidding goal set. |
| Enquiries from the wrong people | Search terms report full of queries adjacent to what you sell, or from outside your service area. | Query selection: match types too loose, negatives missing, or location settings set to interest rather than presence. |
| Relevant queries, no submissions | Good queries, decent click volume, almost nothing submitted, and a high exit rate on the landing page. | The page. Message match, a form that fails on mobile, or too many competing actions on one screen. |
| Spend rising, search impressions flat | A large share of cost sitting in search partners, display expansion, or a Performance Max campaign. | Where the budget is being spent, rather than how it is being bid. A settings problem, not a strategy one. |
| Performance fell after switching bidding | A clear turn in results dated to a bidding change, with conversion volume that never stabilised. | The conversion data underneath the automation, not the automation. It is buying more of what it was told to value. |
| Good leads, no customers | Enquiries are genuinely qualified, and the close rate is poor or the leads go cold. | Outside the account entirely: the offer against the alternatives, or how long somebody waits for a reply. |
Practical
What to have open while you do this
None of this requires a tool you do not already have. It requires about half a day and a willingness to read the reports rather than the summary screen.
- The conversion actions table, showing what each action counts and whether it feeds the bidding goal set
- The search terms report for the last thirty to ninety days, sorted by cost
- A network segment on the campaigns, separating search from partners and display expansion
- Your own inbox, call log or CRM for one complete week, for the reconciliation
- The landing pages, opened on an actual phone rather than a narrowed browser window
- A record of the last twenty enquiries with the time of the first human response
The uncomfortable part
Two causes no amount of account work can reach
The first is the offer. Search advertising puts you in front of somebody who is comparing you against everything else on the same page, usually within a few minutes. If your price, guarantee, availability or proposition is weaker than the alternatives they are looking at, better advertising delivers more people to the same conclusion faster. Nobody selling account management wants this to be the answer, and it is sometimes the answer.
The second is what happens after the form is submitted. Enquiries generated by paid search are frequently made in a batch — three or four suppliers contacted in one sitting — and the first credible reply has an enormous advantage. A business replying within minutes and a business replying the next working day are buying the same clicks at the same price and getting different results from them.
Both of these are cheaper to fix than anything inside the account, and both are usually skipped because they belong to somebody other than whoever runs the advertising. If the diagnostic sequence above clears the account, this is where to look next, and it is worth looking before increasing the budget.
Questions
What people ask when the spend is not producing anything
How do I know whether my conversion tracking is even correct?
Count real enquiries for one week from your inbox, phone log and CRM, then compare that number with what the platform reported for the same week. If they are close, the tracking is broadly sound. If the platform reports substantially more, something is being counted that is not a lead.
The usual culprits are a conversion action firing on a page load rather than a submission, the same submission counted more than once, and a soft action such as a click-to-scroll left in the primary goal set where it feeds bidding.
Should I pause the keywords that cost the most and produce nothing?
Not as a first move. Expensive terms are often expensive because they carry the most intent, and cutting them is the fastest way to reduce spend while also reducing revenue. It also starves an automated bidding strategy of the data it needs.
Look at the search terms report first. Frequently the keyword is fine and the queries it matched are not, which is a negatives and match-type problem rather than a reason to remove the keyword.
Why did performance get worse after we switched to automated bidding?
Usually because the conversion data it was given does not describe a real business outcome. Automation is very effective at buying more of whatever it is told counts, so a goal set that includes page views, newsletter sign-ups or duplicate form submissions produces exactly that at scale.
Fix the conversion actions first, let the system relearn, and only then judge the bidding strategy. Judging it on top of bad data tells you nothing about the strategy.
The clicks look relevant but nobody enquires. What now?
Compare the query, the ad and the page as a sequence. Somebody searched a specific thing, was promised something in the ad, and landed on a page that either repeats that promise immediately or does not. Most of the loss is in that third step.
Then check the mechanics: does the form work on a phone, does it send anywhere, is there a phone number visible, and is there one obvious action rather than five competing ones.
How long should I leave changes before judging them?
Long enough to accumulate a meaningful number of conversions, which depends on your volume rather than on a fixed number of days. A low-volume account judged weekly is being judged on noise, and reacting to noise is itself a cause of poor performance.
Also allow for conversion lag. If your enquiries typically arrive several days after the click, the most recent period will always look worse than it is.
Could the problem be nothing to do with Google Ads?
Frequently, and two causes in particular sit outside the account. The first is the offer: if what you sell is priced or positioned poorly against the alternatives on the same results page, better advertising delivers more people to the same decision.
The second is response time. Enquiries from paid search are often comparing several suppliers at once, and a reply the next morning arrives after the decision. Neither of these is fixable inside the interface.
Send us the account and we will tell you which of the three it is
Read-only access is enough. You get the finding and the evidence behind it, including the cases where the account is fine and the problem is the offer or the response time. That answer is free and occasionally annoying.
Last updated