Sales has one new inquiry. Google Ads shows two conversions. Before changing the campaign, I’d check what those two conversions actually represent. Two tags could be counting the same form, or one conversion could be a form submission and the other a qualified lead recorded later. Those are different problems.

The same goes for a missing lead. If the form didn’t save the inquiry, fixing the tag won’t get it to sales. And if the tag fired, that still doesn’t tell you whether Google Ads can attribute the event to an eligible ad interaction.

I’d pick one form and trace one submission. Write down what you expect at each step, then stop at the first place where the evidence doesn’t match.

1. Confirm the form actually saved a lead

Start with the event you want to measure: a successful demo request, for example. A click on the submit button isn’t the same thing. A visit to a thank-you page is only a useful proxy if that page reliably follows a successful submission.

Use an approved test form or coordinate a clearly marked test record so sales knows not to follow up. Try an invalid submission first, then send one valid submission. Check the form backend or CRM for the record. Note the timestamp and time zone, the form, and an internal test reference. You don’t need to put someone’s email or phone number in the debugging notes.

2. Watch what fires on success and on reload

Use Tag Assistant to inspect the path. Look for the expected conversion tag or source event when the form succeeds. Check that a validation error doesn’t send it. Then refresh the confirmation page, if there is one, without submitting the form again. If the event fires again, check whether the trigger is tied to a page view instead of a successful submission.

Write down the exact conversion action you expect. A direct Google Ads website tag and an imported Google Analytics event can both describe the same form, but they’re separate actions if you set them up that way. Check which actions are recording and which ones the campaign uses before changing either one.

3. Compare the action and the date, not just the total

A direct visit for a tag test is useful for verifying the setup. It doesn’t have to show up as a paid conversion. To check reporting, compare an eligible, attributed lead when you have one. Give the data time to process, then segment by conversion action. Check that you’re in the right account and campaign and looking at the right date range.

The standard Google Ads conversion columns generally report against the ad interaction date. The conversion-time columns are more useful when you’re comparing with the date a lead arrived in the CRM. Matching dates still doesn’t make the totals identical: attribution rules, reporting delays, and differences in the events you’re counting can all matter.

Check primary and secondary actions before treating a blank Conversions column as broken tracking. A secondary action generally shows in All conversions and isn’t used for bidding. Custom goals are an exception: an included secondary action can be used for bidding. Look at the actual campaign goal before making a change.

What I’d do with a duplicate

Hypothetical example: A demo form creates one CRM record. In Tag Assistant, the same website conversion tag fires once on success and again on a confirmation-page reload. I’d fix the repeat trigger, then rerun the same test. If the same tag needs to handle retries and your setup supports it, Google documents how a unique transaction ID can prevent repeat counts for the same conversion action. Use a stable, non-identifying reference per submission, not one fixed value for every lead.

If two different conversion actions are counting the same submission, an ID on one action won’t fix the overlap. Decide which action represents that event for bidding and reporting. Keep genuinely different milestones, such as submitted and qualified, separate and name them clearly.

The One versus Every setting is a counting rule for a conversion action after an ad interaction. It doesn’t fix a tag that fires twice or deduplicate a person across separate actions. I wouldn’t use it to hide a problem we can reproduce.

Copy this test record

  • Expected event: One saved demo request, not a button click. Form and destination: [ ]. Test reference and timestamp with time zone: [ ].
  • Business record: Invalid submission created [ ] records. Valid submission created [ ] records. CRM status: [ ].
  • Tag evidence: Expected action and source: [ ]. Fires on invalid submit [ ], valid submit [ ], reload without resubmit [ ].
  • Reporting check: Campaign goal [ ], primary or secondary [ ], column and date basis [ ], action counts [ ]. Ad attribution verified, pending, or not applicable: [ ].
  • Decision: First mismatch [ ]. One change to test [ ]. Owner [ ]. Retest result [ ].

A clean test tells you the measurement path worked for that case. It doesn’t tell you whether the inquiries are useful. Once you trust the count, check what happened to those leads. If the problem is broader than one action, use the PPC audit checklist. If you want another set of eyes on what your campaigns are actually optimizing for, that’s part of my Google Ads management work.

Platform guidance checked November 12, 2026 against the Google Ads Help pages linked above.