The short answer: open your conversion actions, make a real test conversion yourself, then compare Google Ads’ count with the orders, calls or form submissions in your own records for the same dates. If the two disagree by more than roughly a quarter, or the same action is counted twice, your tracking is not trustworthy and any automated bidding strategy is learning from bad data. The whole test takes about twenty minutes.
This matters more than keyword choices or ad copy. Smart bidding strategies optimise toward whatever you have defined as a conversion. If that definition is wrong, the system becomes very efficient at producing the wrong result. We have seen accounts where a “conversion” was a page view of the contact page, and the campaign was proudly delivering cheap conversions that were never customers.
Why does a green status not prove tracking works?
In Google Ads, a conversion action shows a status such as “Recording conversions” or “No recent conversions”. These statuses only tell you that the tag fired or did not fire at some point. They say nothing about whether it fired at the right moment, once per customer, with the right value. A tag placed on a thank-you page that people can revisit will keep “recording conversions” every time someone bookmarks and reopens it.
So the status is the starting point, not the answer. Five checks follow, in order.
Check 1: What does the conversion actions list say?
In Google Ads, go to Goals, then Conversions, then Summary. For each action, note three things:
- Source: website, phone call, imported from Google Analytics or an offline upload.
- Include in “Conversions”: whether it counts toward the main column and toward bidding.
- Count setting: “Every” or “One”.
For purchases, “Every” is right because one person can buy twice. For leads, “One” is usually right, because one person filling in a form three times is still one lead. A lead action set to “Every” is a classic cause of inflated lead numbers.
Also look at whether secondary actions such as newsletter sign-ups or button clicks are included in the main conversions column. If you are bidding toward “add to cart” or “page view” as if it were a sale, bidding will chase cheap but worthless actions.
Check 2: Does a live test conversion register?
Do it yourself. Open your website in a private browser window, click one of your own ads if you can do so without paying for repeated clicks, or at least go through the funnel directly, and complete the action: submit the form, place a small test order, or call the tracked number. Then use Google Tag Assistant or the tag diagnostics in Google Ads to confirm the tag fired once, on the right page, after the action was completed.
Then wait. Conversions can take hours to appear, and some take longer depending on your conversion window and how they are modelled. Check the next day. If nothing appears and you are certain the tag fired, look at whether consent settings are blocking it. For European visitors this is a common cause, and we wrote about the fix in our guide to Consent Mode v2 and enhanced conversions.
Check 3: Does Google’s count match reality?
This is the most revealing step. Take a recent period, say the last 30 days, and write down two numbers:
- Conversions that Google Ads reports for the campaigns.
- Actual orders, booked appointments, form entries or answered calls in your own system (shop back end, CRM, call log or inbox) over the same dates.
They will not match exactly, and they should not. Google counts only conversions it attributes to ads, while your system includes organic, direct and email customers. So the comparison is about plausibility. Google Ads reporting 140 conversions when your whole shop took 90 orders is impossible and proves overcounting. Google Ads reporting 12 when you know your phone has been ringing from the ads daily suggests undercounting.
As a rough working rule, if Google’s number is above your total real number, or differs from what you can reasonably attribute by more than 20–30%, investigate. That threshold is our own rule of thumb, not a Google standard.
Check 4: Are conversions being counted twice?
Duplicates usually come from one of four causes:
- Both a Google Ads tag and a Google Analytics import are marked as primary for the same action, so the sale is counted twice.
- The thank-you page fires the tag every time it loads, including reloads and returning visits.
- Two plugins or a plugin plus manual code are both installed.
- A call is counted both as a call from the ad and as a website call conversion.
To check, filter the conversions column by conversion action name and compare the counts. If two actions show nearly identical numbers and both are primary, one of them should be set to secondary. Pick the most reliable source, usually the tag fired from the confirmed-order event, and demote the rest.
For businesses that depend on phone calls, tracking has its own traps; our article on call tracking and offline conversion import covers how to connect calls to real customers.
Check 5: Are the values right?
If you sell products, confirm that the value passed matches the order total and the currency is correct. A common mistake is passing a fixed value of 1 or sending a value including shipping and tax when you want to optimise on margin. If you run lead generation, assign a value to leads based on what a lead is worth to you, even a rough one, because value-based bidding needs it. Our post on break-even ROAS shows how to calculate the number that tells you what a conversion must be worth.
Check that the conversion value for the test order appears correctly in the report. If the value arrives as zero or as a text string with a currency symbol, the tag is reading the wrong field from your page.
What should you do with what you find?
Fix in this order:
- Remove or demote duplicate and secondary actions from the main conversions column.
- Correct counting (one for leads, every for purchases).
- Repair the tag, the trigger or the consent settings.
- Only then look at bids, budgets and creative.
After a clean-up, expect a short period of unstable performance. Reported conversions will probably drop, because inflated ones are gone, and smart bidding may re-enter a learning phase. Do not read the drop as the campaign getting worse. It is the first accurate reading you have had.
If, after tracking is fixed, you still get clicks but no real enquiries, move on to our seven-day diagnostic for ads that get clicks but no leads.
How often should you repeat the test?
Run this test whenever something upstream changes: a new website theme, a new form plugin, a checkout redesign, a cookie banner update or a switch of payment provider. These changes silently break tags more often than any change inside Google Ads. A good habit is a five-minute version each month: one test conversion, one count comparison.
The reason we put tracking ahead of everything else is simple. An account with average ads and honest data improves steadily. An account with excellent ads and dishonest data wastes budget and nobody notices, because the dashboard says it is working.
A short example of what the gap looks like
Imagine a small clinic that sees Google Ads report 60 “leads” in a month, while the reception log shows 22 new-patient enquiries from all sources. The first reaction is usually to ask which source is lying. The better question is what each system is counting. In this illustration, which uses invented round numbers, the Google tag fires on a page that shows after the booking form loads, so everyone who opens the page counts, including people who never press submit. Moving the tag to fire only on the confirmation message brings the Google count to 18, a figure that is plausible against the log. Cost per lead jumps from a flattering number to a real one, and the campaign decisions that follow are better because they rest on reality.
That is why we prefer to start every account review with this test. It is not exciting, and it will not make the dashboard look better, but it makes everything else in the dashboard believable.