Wrongly set-up conversion tracking reports a success that doesn't exist and shifts the budget to the wrong place. Seven points to check before trusting the numbers.
The most dangerous number in ad reports is not zero but the wrong number. When you see zero, you know there's a problem; an inflated conversion count, on the other hand, pleases you and leads you to shift the budget to the wrong campaign. You can do the following seven checks in your own account, without needing anyone.
1. When was the last conversion recorded?
In Google Ads, on the Tools → Conversions screen, the last recorded date appears next to each conversion action. If this date is days ago, there are two possibilities: either conversions genuinely aren't coming, or tracking is broken. These are very different problems and both are urgent.
2. Is the same conversion counted twice?
This is the most common cause of inflation. When the code is added both site-wide and to the thank-you page separately, or set up both via a tag manager and via the theme, a single purchase is reported twice.
A simple test: make a test conversion (fill in the form or place a low-value order) and look at how many appear in the report. If you did one action and the report says two, the matter is clear.
3. Is what's being counted really a sale?
If "page view" or "all form submissions" is chosen as the conversion, the numbers come out high but are meaningless. Spam form submissions in particular inflate this number quickly. Make sure the counted event is the action that's genuinely valuable to the business: purchase, quote request, phone call.
4. Are phone customers measured?
In the service sector, a significant portion of customers don't fill in a form; they call directly. If clicks on the site's phone number aren't defined as a conversion, most of the business the ad brings never appears in the report. The campaign seems "not working" and gets shut off.
5. Is the conversion window suitable for your business?
The default tracking window is 30 days in most accounts. In businesses where the decision is made quickly (food ordering, emergency service), this period is longer than needed and can attribute sales to irrelevant clicks. In businesses where the decision takes months (architecture, enterprise software), 30 days is short and hides the real contribution.
6. Is the measurement defined in more than one place?
If the same event is counted both by Google Ads' own tag and imported via Analytics, the total figure can become double the reality. Choose a single source for an event; don't leave both open.
7. How much do ad blockers and cookie consent cut off?
A portion of visitors reject cookie consent or use an ad blocker; these visits don't show in browser-side measurement. This doesn't mean the setup is broken — but it explains part of the difference between platform figures and your own accounting records. Take this into account when comparing figures.
Why don't platform figures match each other?
If Google Ads says 40 sales, Meta says 25 sales, and your total sales are 50, no one is lying. Each platform credits to itself the sale of a user who touched its own ad. The same customer may have both seen the Instagram ad and searched from Google. So platform reports are not added up; your own sales records are the single true reference. Platform figures, meanwhile, are used to compare between campaigns.
Frequently asked questions
The conversion count dropped but sales are the same. What happened?
Usually there's been a change on the tracking side: the site was updated, the thank-you page address changed, or the cookie consent bar was renewed. First verify the measurement, then question the campaign.
Is server-side measurement necessary?
It's not mandatory. It's meaningful in high-volume accounts where browser-side losses are large enough to affect business decisions. In small accounts, the setup and maintenance cost can outweigh the extra accuracy it provides.
Does a test conversion spoil the report?
A few test records don't distort the statistics meaningfully; but note the day you tested and take that day's data into account when interpreting it.
