Five tracking mistakes that lead Google Ads to bad decisions

Automated bidding in Google Ads doesn't know what good business looks like. It only knows what the conversion data shows it. If the data is wrong, the system isn't failing. It's doing exactly what you asked, just not what you wanted.

These are the five mistakes I keep finding in accounts. You don't need a developer to spot any of them.

1. Too many conversions marked as primary

If form submissions, phone click-throughs and newsletter signups all sit in the primary column, bidding treats them as equal. Newsletter signups are the easiest to get, so that's where it'll push.

Check: Goals, Conversions, Summary. See what has primary status. Anything that isn't a business result belongs in secondary.

2. No value attached to the conversion

Without a value, every conversion is worth the same. To the system, a 200 EUR enquiry and a 20,000 EUR enquiry look identical.

If you can't pass a real value, an estimate that varies by type still beats nothing. A fixed value per service type, for example.

3. Double counting

If a conversion comes in once through the Google Ads tag and again through a GA4 import, it's counted twice. The giveaway is a suspiciously good ROAS that never shows up in your bank account.

Check: Compare the Google Ads conversion count with the real number of orders or leads in your CRM or shop, for the same day. If Ads shows more, look for the duplicate.

4. The conversion window doesn't match the real decision time

The default is 30 days. If your sales cycle is six weeks, some conversions never make it back to the campaign that drove them. If it's set to 90 days for an impulse buy, you're crediting purchases to campaigns that had nothing to do with them.

5. Consent mode is missing, or wired the wrong way round

If you advertise in the EU and consent mode isn't set up properly, you either lose some of your conversions or collect data you're not allowed to collect. Both are problems, just different ones.

What to do with this

You can work through all five in an afternoon. The order matters: the primary list first, then duplicates, and only then value and window. If the first two are wrong, there's no point tuning the rest.

If you're not sure the data in your account is sound, don't optimise the campaign. Fix the measurement first.