When Google Ads reports no conversions, the first suspect is the counter, not the customers

A working campaign that looked like a failing one

I ran Google Ads for my own clinic on a small daily budget, a Performance Max campaign and a Search campaign for the high-intent local terms, and reviewed the click and conversion data myself. I have watched a conversion tag break and turn a campaign that was still producing enquiries into one that, on the dashboard, appeared to have died overnight. The clicks were there, the spend was there, and the conversions column had simply gone quiet.

I am a prescribing pharmacist, not an advertising specialist, and I built the clinic's website, Google Business Profile, Tag Manager container, analytics and cookie consent banner myself. Each sits in the chain between a customer filling in a form and a number appearing in an advertising report. When any link moves, the number changes, and the platform cannot know whether the customer or the counter went missing. The two readings lead to opposite decisions: if demand has fallen, cutting the budget is sensible; if only the measurement has failed, cutting the budget removes the enquiries that were still arriving.

What a conversion actually is, and why it is fragile

Google's description is mechanical: the advertiser adds a Google tag to the site; when someone clicks an ad a temporary cookie is placed on their device, and when they complete the defined action the system recognises the cookie and records a conversion ^[1]. The action can be a page load, which Google's setup guidance recommends when no transaction values are needed, or a click on a button or link ^[2]: in practice, a thank-you page or a submit button.

Many small businesses instead mark an event in Google Analytics 4 as a key event, an event that measures an action particularly important to the business, and create a Google Ads conversion from it, which Google says gives the same counts in both products and lets the advertiser bid against them ^[3]. That adds a third system to the chain: the site fires an event, Analytics records it, Google Ads imports it. Each design depends on one fragile thing, a URL, a button's markup or an event name, and when it changes the count falls to zero without an error, because the tag is simply never asked to fire.

How tracking breaks without anyone noticing

In my experience of maintaining the clinic's own stack, the breakages are side effects of ordinary maintenance. The list is not exhaustive.

• A form plugin is updated and the confirmation now appears in place, without loading the thank-you page whose URL the conversion was watching.
• A redesign or theme change drops or moves the Tag Manager snippet, so the container never loads on the pages that matter.
• A booking or form platform is swapped and the new one fires a different event name, or none, while the Analytics key event and the Google Ads import still wait for the old one.
• The thank-you page URL is tidied during a content edit and the conversion action still points at the old path.
• A tag is paused or deleted in Tag Manager, or a container version is published with a trigger that no longer matches.
• A consent banner is installed or reconfigured and begins blocking the advertising tag before consent, or never passes the consent signal to the tag at all.

The consent case is partly deliberate. Google's Tag Manager documentation says a tag set to require additional consent fires only if every specified consent type is granted when it is triggered ^[4]. Under basic consent mode, when the user does not consent, no data is transferred to Google at all and the tags are completely blocked; under advanced consent mode the tags send what Google calls cookieless pings instead ^[5]. Google's page on common measurement problems now lists an unexpected drop in tracking data for users in the European Economic Area after a consent mode update as a recognised issue ^[6], so the platform's own maintainers know a consent change can look exactly like a conversion collapse.

I set up the consent banner, Tag Manager and Analytics on the clinic site and on highregard.co.uk, and had to decide what loads before consent and what loads only after. That decision is the hinge. Require a consent type the banner never actually sets, and the advertising tag waits politely for a signal that will never come.

What UK consent law requires, and what it costs in data

The blocked tag is not a bug in the law. The ICO's guidance on cookies under the Privacy and Electronic Communications Regulations says a service must tell people the cookies are there, explain what they do and why, and get consent to store a cookie on the device, with narrow exemptions for transmission and for what is strictly necessary to provide the service the user asked for ^[7]. On advertising the ICO is blunt: a service using storage or access technologies for online advertising must get consent, none of the exceptions applies, and online advertising purposes have never been exempt ^[8]. A separate statistical purposes exception covers measuring how a service is used with a view to improvement, but it forbids tracking individual visitors and any use for online advertising ^[8]. A Google Ads conversion tag does not fit inside it, so a lawful UK site gates it behind consent.

Two field studies give a sense of the resulting gap. Utz and colleagues, working with more than 80,000 users of a German website, found that under 0.1 per cent allowed all third parties when offered a genuine opt-in choice, while highlighting the accept button lifted acceptance to 50.8 per cent on mobile and 26.9 per cent on desktop ^[9]. Nouwens and colleagues, who scraped the consent pop-ups on the top 10,000 UK websites, found that only 11.8 per cent met the minimal legal requirements they set, and that removing the opt-out button from the first page increased consent by 22 to 23 percentage points ^[10]. Those figures describe how design changes consent rates, not any particular clinic's rate, and both studies predate most current banners. The narrower inference is that a banner change can move the share of counted conversions by tens of percentage points without moving demand at all.

Google's answer is modelling. Analytics says it estimates key events it cannot observe directly, including those of unconsented users where consent mode is implemented, but reports none if there is not enough traffic to inform the model ^[11]. A clinic with a handful of bookings a week is unlikely to clear that threshold, so its consent gap stays a gap.

What Smart Bidding does when the signal disappears

Smart Bidding, in Google's definition, means bidding strategies that use Google AI to set bids for conversions or conversion value in every auction, and conversion tracking must be enabled to use it ^[12]. Google recommends judging performance over periods with at least 30 conversions ^[12], a volume a small account rarely reaches, which already makes its bidding sensitive to any change in the signal.

The algorithms are designed to follow the data. Google says Smart Bidding keeps updating to align with shifts in performance, uses adaptive historical weighting to rely more heavily on recent data, and with little or no conversion data falls back on query-level data from beyond the bid strategy ^[13]. Read against a broken tag, the system does not stop; it re-weights towards a recent past in which conversions are absent.

The more common outcome in a small account is subtler, because most have more than one conversion action. Google distinguishes primary actions, which can be used for bidding and reporting, from secondary actions, which are excluded from the conversions column used for bidding ^[14]. If the booking form was one primary action and a click on the phone number another, and only the form event breaks, the bidding now steers entirely towards phone clicks. The reported cost per conversion becomes a figure about the wrong thing, and it can look healthy.

Repair has side effects too: Google says Smart Bidding takes one to two conversion cycles in most cases to learn after any change to conversion goals or actions ^[15], so a silent break followed by a repair means two unchosen adjustments. That is a reason to detect the break in days rather than weeks.

Why the platform cannot tell you the signal is broken

Google Ads does have statuses for conversion actions. Its troubleshooter lists five: active; needs attention, where recording has not stopped; misconfigured, where conversions have stopped entirely due to a setup error or a broken tag; awaiting conversions, defined as no conversions recorded in the last seven days; and removed ^[16]. Awaiting conversions, it says, typically occurs when the action was created under 48 hours ago, campaigns are paused, or website traffic is insufficient ^[16].

That list of causes is the point. If the tag has been ripped out of the template, Google can sometimes detect its absence. But the failures above mostly leave the tag in place and loading. The thank-you page exists, it is just not visited; the container loads, it just never fires; the consent gate works, it just never opens. To the platform each is a healthy tag with no conversions, which it attributes to thin traffic or paused campaigns.

Google's companion page blurs the picture further. Reporting can take three to 24 hours to update, and conversions are attributed to the date of the click rather than the date of the action ^[17]. So the first days of a break do not look like a clean zero; they look like a soft decline, which is what a real fall in demand looks like. The recommended checks, a test conversion with Tag Assistant, confirming actions are set to primary and verifying tag placement ^[17], are sound but assume someone is looking. The dashboard gives that person every reason to suspect the campaign instead.

Demand collapse or measurement collapse: putting the site beside the platform

The reliable test is to compare three systems rather than one. The advertising platform reports clicks and conversions. Analytics reports sessions and the site's own lead events, whether or not anyone imported them into Google Ads. Search Console reports the clicks and impressions the site earned in organic results, which need no tag at all ^[19]. Genuine collapse shows up in all three. A broken tag shows up in one.

This is the method the HighRegard application uses, and I describe it plainly because it is the subject of this article. It connects Google Ads, Search Console and Google Analytics 4 read-only and runs a short set of checks each time the data refreshes; the verdict comes from reading them together.

Whether the signal stopped while activity continued

It compares the last seven days of platform conversions against the trailing four weeks, and makes the same comparison for sessions and for the site's own lead events: form submissions, WhatsApp clicks, call clicks. Conversions at or near zero while sessions and lead events held within their normal range is a stopped signal. If all three fell, the pattern is consistent with demand, and the verdict says so.

Which conversion actions steer the bidding

It counts how many actions are marked primary and what kind they are. Several primary actions with a click-type action outranking the booking action produces this symptom over time without any tag breaking ^[14].

Whether the booking event reaches both places

An event that appears in Analytics but not in Google Ads points at the import. One that appears in neither points at the site. One that appears in both, but at a fraction of the site's form submissions, points at the consent layer.

Whether the data is fresh

A source that has not delivered data for two days is treated as stale and reported before anything is scored, because a stale feed would itself look like a collapse.

The checks combine into a tracking-health score and one of four verdicts: healthy, degraded, likely tracking failure, or not yet assessable. Only checks that could run are counted. A verdict of likely tracking failure puts a plain warning at the top of the overview and blocks every recommendation that depends on platform conversions: budget increases, exact-match expansions, schedule and location changes, incrementality tests. Those items stay visible but cannot be accepted until the tracking item is done, because a low-ranked instruction to increase budget is still an invitation to act on a broken number. The demo workspace runs this scenario on synthetic data, showing a score of 69 out of 100, a verdict of likely tracking failure and four blocked actions ^[20].

When it really is demand, and when the comparison misleads

The other reading deserves its due. Demand does fall: a competitor opens nearby, a season ends, an ad is disapproved. In each case the site's own lead events fall with the platform's conversions, and the honest verdict is that the campaign, or the market, has changed. The method's value is that it says so rather than reflexively blaming the tag.

Platform and site can also diverge for reasons that are neither. A banner made more compliant reduces the share of visitors the advertising tag can see, and the studies cited earlier suggest that shift can be large ^{[9] [10]}. That is a lawful reduction in measurement, and the right response is to stop expecting the Google Ads conversions column to be a census. Attribution cuts the other way: because Google Ads assigns a conversion to the click date ^[17], late-arriving conversions make this week look worse than it will. A seven-day window against a four-week baseline reduces that distortion without removing it, which is one reason the application reports degraded rather than failure when the numbers are ambiguous.

Analytics itself can break. If one change removed both tags, the comparison sees two falling lines and reads them as demand, which is why Search Console clicks, which need no tag, are in the set.

A checklist for spotting it within days

The same comparison can be made by hand in under an hour and repeated weekly. The order matters.

1. Open Analytics and compare the enquiry event for the last four weeks against the previous four. If it is still arriving at roughly the usual rate while Google Ads shows conversions collapsing, do not touch the budgets.
2. Open Search Console's performance report and check that organic clicks have not fallen in step ^[19]. If organic and paid enquiries fell together, treat it as demand until proven otherwise.
3. Make a test enquiry from a device that has accepted the cookie banner and watch for the event in DebugView, which shows the events Analytics collects from a user in real time ^[18]. Then repeat from a device that has rejected the banner; if the event arrives only with consent, the gate is doing its job, and if it never arrives, the site is the problem.
4. Look for the test in the Google Ads conversions report the following day, allowing for the reporting delay of up to 24 hours ^[17], and remember that a status of awaiting conversions will not name a broken implementation ^[16].
5. Ask what changed on the site in the fortnight before the first zero: a plugin update, a form change, a theme update, a new banner, a URL edit. In my experience the answer is usually in that list.

The fault is nearly always in the site, where the event no longer fires; the consent layer, where the tag is blocked; or the import, where Analytics sees the event and Google Ads does not. Fix that first. Only then does the question of whether the campaigns are working have an answer.

What this means for a local business

For a clinic, a dental practice or a trades business spending a few hundred pounds a month, the stakes are asymmetric. A tag that breaks silently costs nothing on the day it breaks. It costs something every day afterwards: bidding that steers towards the wrong action, then a decision to cut a campaign that was still producing enquiries.

Three habits reduce the risk. Treat the site's own enquiry count as the primary measure and the platform's conversions as a derived one, checked against it weekly. Log site changes, however small. Make a test enquiry after every plugin, theme or consent change, and watch it arrive in Analytics before assuming anything about the campaign.

It also means accepting the consent gap. A UK site that follows the ICO's guidance will not count every enquiry in Google Ads, because advertising tags require consent and some visitors decline ^[8]. That undercount is lawful and permanent, so the right measure of a campaign is its trend against the site's own data.

What the evidence does not settle

Nothing cited here measures how often small-business conversion tracking actually breaks. Google publishes statuses and causes but no frequencies; the consent studies measure banner design on a handful of sites, years ago; my own observation is one account. The claim that measurement collapse is more common than demand collapse in a small account is my judgement from that experience and from the mechanics above, not a documented statistic.

Nor does it settle what Smart Bidding does in the days after a signal vanishes. Google says the algorithms re-weight towards recent data ^[13] and that changing conversion actions triggers a learning period of one to two conversion cycles ^[15], but it does not publish how quickly bids move after an unannounced break or how much spend is misdirected meanwhile. Any figure offered for that would be invented. Consent modelling is the third open question: Google withholds modelled key events below a confidence threshold ^[11] but does not say where it sits, so a small business cannot know whether its gap will be partly filled.

What remains is a method rather than a proof. Put the platform's number beside the site's own, and beside a source that needs no tag at all. When they move together, believe the fall. When one drops alone, look at the counter before the customers. How often that second case arises among small advertisers is a question nobody has yet published an answer to.