Google Ads reports more “conversions” than the business can connect to actual inquiries. Inside the total are phone-number clicks, form starts, thank-you page views, duplicate conversion actions, and a smaller number of delivered leads.
Nothing about that total is useful until each action has a job:
- Optimize toward it. It is a completed business outcome that may guide bidding.
- Observe it. It is a supporting signal that helps explain behavior.
- Repair or remove it. It is broken, duplicated, obsolete, or misleading.
Then verify how the campaign actually uses each action. A conversion can look healthy in the account and still be irrelevant to the campaign currently spending money.
Start with an outcome the business can recognize#
Name the customer action the campaign is meant to produce. For a service business, that might be a contact request that reaches the business, a qualified booking, a call meeting an agreed lead standard, or a completed deposit.
“Engaged visitor” is not an outcome. A phone-number click is not a completed call. A booking-widget open is not a booking. A thank-you page view is evidence only when a successful submission caused it and the lead arrived.
Tracking should follow the outcome, not turn whatever is easy to count into an outcome.
| Action being recorded | Likely role | What must be verified |
|---|---|---|
| Completed inquiry delivered to the business | Usually optimize | Success shown, inquiry received, event recorded once |
| Confirmed consultation or appointment | Usually optimize | Booking exists and represents new-business intent |
| Meaningful phone call | Possibly optimize | Call used the intended route and the qualification rule makes sense |
| Phone-number click | Usually observe | A tap does not prove a conversation |
| Form start or booking-widget open | Observe | Useful friction signal, not a completed lead |
| Page view, scroll, or time threshold | Observe or remove | Engagement alone does not establish an inquiry |
| Thank-you page load | Investigate | Direct visits and refreshes cannot create false counts |
| Duplicate native and Analytics events | Repair | One action must not become two bidding signals |
| Support request, job inquiry, spam, or irrelevant lead | Exclude from acquisition optimization | Counted activity is not necessarily desirable work |
These are starting roles, not universal settings. Commercial meaning depends on how the business sells and fulfills the work.
1. Optimize toward completed outcomes#
An action is a candidate for optimization when the customer completed it, the business received it, it has clear commercial meaning, and it is recorded consistently enough to support decisions. A delivered lead form, qualified booking, or completed purchase may qualify.
Calling an action “primary” does not guarantee every campaign uses it. The campaign's selected goals determine which actions are eligible, and its bidding strategy determines whether conversion signals affect bids at all.
2. Observe supporting signals#
Phone-number taps, pricing-page visits, downloads, form starts, and booking-widget opens can expose interest or friction. Keep them for observation when they help diagnose the path, but do not let them stand in for the finished result.
If the account shows twelve form starts and three received inquiries, the starts may be informative. They are not three more leads.
3. Repair or remove misleading actions#
The most dangerous conversion problems are often plausible enough to survive for months.
A native Google Ads conversion and a Google Analytics 4 (GA4) import can both count the same lead. A form start can be mislabeled as a submission. A booking-widget open can be treated as a booking. A thank-you page can remain directly accessible. Site code and Google Tag Manager can both fire the same event. Old conversion actions can also survive a rebuild after the page or form they described has changed.
Rename a useful but misleading event. Repair a broken one. Remove an obsolete action from the reporting and goals that matter. One customer action should not accidentally become several bidding signals because two implementations both recorded it.
Counting rules should match the outcome: a single lead inquiry and repeat purchases may reasonably be counted differently. Define which event represents completion and leave supporting events observational. Advanced deduplication belongs with whoever owns the implementation.
Test event behavior and delivery first#
Choose one intended lead action and complete it directly. Verify that:
- the visitor can complete the action;
- the visitor sees a truthful success state;
- the business actually receives the inquiry, call, booking, or order;
- exactly one intended completion event fires;
- the event's name and meaning match what just happened; and
- refresh, back-button, or reopen behavior does not create another completion.
A success screen confirms only that the browser displayed success. It does not prove delivery to the inbox, CRM, calendar, call record, order system, or responsible person. If the conversion fires while the lead disappears, repair the delivery path before trusting the total.
A direct manual test can prove event and delivery behavior without creating an attributed Google Ads conversion. Record the specific break. “Fix tracking” is not a diagnosis; “the lead arrives, but both the native tag and GA4 import record it” is.
Verify attributed reporting separately#
An attributed Google Ads conversion depends on an eligible prior ad event. Depending on the campaign and reporting type, that may be a click or engagement, or a qualifying impression reported as a view-through conversion. A direct visit may prove the entire event and delivery path works, but it does not by itself prove Google Ads attribution.
Do not click your own ad to force the test. Use direct testing for immediate event and delivery behavior. Then, allowing for the platform's processing delay, verify that eligible ad traffic receives the conversion.
If the event works but attributed reporting does not, inspect whether the intended action is included in the campaign's active goal configuration. Also check for duplicate actions and custom goals that may pull an otherwise observation-only signal into optimization.
Confirm whether the campaign uses the signal#
Keep four separate questions separate:
- Does the conversion action exist and record the intended behavior?
- Is the action marked primary or secondary?
- Is it included in the campaign's selected goals?
- Does the campaign's bidding strategy actually use conversion signals?
Those settings answer different questions. An action can record perfectly and be marked primary while the active campaign excludes it. A campaign can also report conversions while using a bidding strategy that does not bid toward them. A secondary action can still be used for bidding when it is included in a custom goal the campaign uses.
More than one optimized action can be valid when the business genuinely values several completed outcomes. The danger is not plurality. It is treating unequal or duplicate behaviors as interchangeable results.
Keep measurement accuracy separate from lead quality#
Accurate measurement tells you whether the intended action happened and was recorded correctly. It does not tell you whether the inquiry was good.
A campaign can count every delivered form perfectly and still attract wrong-service, wrong-location, spam, or low-value inquiries. Tracking repair makes the numbers trustworthy; traffic, offer, qualification, and follow-up determine what those numbers are worth.
Once inquiries are reliably received, the lead-tracking guide provides a simple record for source confidence, fit, stage, and outcome.
Migrations, landing-page replacements, CRM changes, and booking-tool changes should trigger the same end-to-end retest. Old URLs, tags, imports, and conversion actions can survive long after the behavior they were meant to represent has changed.
Run one practical end-to-end test#
Submit one intended lead action directly. Confirm that the visitor sees a truthful success state, the business receives and owns the result, and exactly one intended completion event fires with the expected meaning. Then verify that the action has the intended optimize, observe, or repair role, that the campaign goal includes it when appropriate, and that the bidding strategy uses it as expected.
If any link in that chain is unclear, the next task is the specific break you found, not another conversion total.