Why your Google Ads are not converting
A diagnosis by the report you open, for a Birmingham account
Eight reports inside your own Google Ads account, what each shows when ads are not converting, and the fix for each, with the Birmingham cases we see most.
Written by Daniel Stoychev, who runs the accounts · 5 September 2026 · about 10 minutes
An account that spends without converting is nearly always failing in one of eight places, and each place has a report inside your own Google Ads account that shows it. Open them in the order below and the cause is usually visible within an hour: the searches were wrong, the places were wrong, the hours were wrong, the page was too slow, the tracking was broken, the ads were thin, you were outbid, or the offer itself does not work in the auction. The first five are the common ones in Birmingham accounts we read, and they are all fixable this week.
Before anything else: are the conversions being counted at all?
Go to the conversion actions summary. If there is no action for calls and no action for the form, the account is not failing to convert; it is failing to see. If there are actions, submit the form yourself and make a call over a minute, and check that each appears once. An account that counts a thank-you page refresh as a lead, or counts every three-second misdial, is teaching the bidding to buy the wrong thing, and looks like it is converting when it is not. Half of the Birmingham accounts we read for the free check have something wrong here, and it changes the meaning of every other report.
The eight reports, in the order to open them
| Report | What it shows when the account is not converting | The Birmingham case we see most | The fix |
|---|---|---|---|
| Search terms | Spend on searches that could never have bought: jobs, courses, manuals, other towns, the trades you do not do | A Digbeth shopfitter paying for "shopfitting jobs"; a Solihull solicitor paying for Coventry searches | Exclude them, weekly, at the right match type; tighten the keywords that let them in |
| Locations, by user location | Clicks from outside the area you serve | A twenty-mile radius from the centre buying Dudley and Redditch clicks for a Kings Heath business | Presence-only targeting, postcode districts, place names on the negative list |
| Ad schedule, by hour and day | Clicks at hours nobody answers, or nobody buys | A trade paying for midnight clicks to voicemail; a restaurant bidding at 10am for dinner | A schedule from the enquiries-by-hour view, bid adjustments up and down |
| Landing page, in PageSpeed Insights | A page that takes five or six seconds to be usable on a phone, or a form that does not submit | A homepage doing every campaign's job; a hero image the size of a film | One page per campaign, fast, with the number and the form in the first screen |
| Devices | Spend on a device that never converts | B2B firms paying for Sunday phone clicks; trades paying for desktop clicks that never call | Device bid adjustments from the conversion rate, not from a guess |
| Ads and assets | Ad strength poor, three headlines where fifteen are allowed, no call or location asset | A phone-led business with no call button on the ad | Full responsive search ads per ad group, assets on every campaign, pins only where a claim must stay |
| Auction insights | Outranked by the same competitors on the searches that matter | A franchise outbidding an independent on the generic term | Move to the specific searches they ignore; improve the page so the same bid ranks higher |
| Conversions by campaign, over three months | Every other report is clean and the cost per enquiry is still above what an enquiry is worth | A low-margin service where the click price exceeds the margin at any conversion rate | Stop, honestly, and put the money into rankings or a different offer |
The searches were wrong
Open the search terms report for the last ninety days and sort by cost. Read down the list and mark every search that could never have become a customer. In most unmanaged accounts the marked rows are a fifth to a third of the spend, and in a city with five universities and a big jobs market the culprits are predictable: course, apprenticeship, salary, jobs, free, DIY questions, product manuals, the towns you do not serve. Each one goes on a shared negative list at the match type that blocks it without blocking anything you want. Then look at how they got in: usually a broad match keyword added by a recommendation, or a phrase match keyword that is one word long.
The places were wrong
Two settings decide this. The location option, which by default shows ads to people who have merely shown interest in your area, including people abroad, and should be set to people physically in it. And the radius, which in a conurbation like the West Midlands quietly includes towns you do not serve. The locations report, viewed by user location rather than by target, shows where the clicks actually came from. If it says Wolverhampton and you are in Solihull, the fix is postcode districts and a place-name negative list.
The hours were wrong
Segment conversions by hour of day and day of week. A trade that answers 7am to 7pm and runs ads around the clock is paying for evening clicks that reach voicemail; a restaurant that wants dinner bookings and bids evenly all day is paying for lunchtime clicks from people at their desks. The fix is an ad schedule with bid adjustments, set from the enquiries-by-hour view and corrected monthly from the same view.
The page was too slow, or the wrong page
Put the landing page into PageSpeed Insights and read the mobile score. Then load it on your own phone with the wifi off. If it takes more than about three seconds to be usable, or the number is not a tap-to-call link, or the form is nine fields long, the account is buying visits from people who leave before they see the offer. A homepage sent every campaign's traffic is the most common version of this: it answers every question and therefore none. One page per campaign, repeating the search in the first heading, is the fix, and it lowers the cost per click as well through Quality Score.
When everything is clean and it still does not pay
Sometimes the honest diagnosis is that the searches cost more than the margin can bear at any realistic conversion rate. A low-margin service in a sector where franchises bid hard on every term can be in that position. The report should say so, with the sum, and the answer is to stop and put the money into search rankings, a different offer, or a different channel. An agency paid a percentage of spend has no reason to write that paragraph; it is one of the reasons the fee model on this site is flat.
The order matters. Fix the tracking first, because until the account can see enquiries, every other report is describing the wrong thing. Then searches, places and hours, because they are cheap to fix and losing money today. Then the page. Then, and only then, the bids.
Follow-up questions
My ads get clicks but no calls. Where do I look first?
The conversion actions, to check calls are being counted at all, then the landing page on your own phone with the wifi off. Clicks without calls usually means the page is slow, the number is not a tap-to-call link, or the calls are happening and not being counted.
How much of a typical Birmingham account's spend is wasted on the wrong searches?
In the unmanaged accounts we read, a fifth to a third, and it is visible in the search terms report sorted by cost. The free check on this site reports the figure for last month.
Should I pause the campaign while I fix it?
Pause the parts that are clearly wrong, such as a keyword pulling in job searches, and leave the rest running so the fixes can be judged against data. Pausing everything resets the learning and makes the comparison harder.
How long after a fix should I wait before judging it?
Two weeks for a change to searches, places or hours; a month for a change to bidding, because Smart Bidding relearns. Judging after three days is judging noise.
What if the tracking has been wrong for a year?
Then the account has been optimising for the wrong thing for a year and the historical data is unreliable. Fix the tracking, give it a month of clean data, and treat that month as the new baseline.