Tracking & measurement

Separate business outcomes from diagnostic conversions in your scorecard

A page visit can confirm that an ad sent someone to the website. A submitted enquiry can indicate interest. A qualified enquiry or paid order supports a different business decision. A useful scorecard keeps these signals connected without adding them together. Start with the business outcome that can justify expenditure, then use diagnostic actions to investigate where the journey might be breaking.

One decision needs one business outcome

Define a business row and a diagnostic row set. The business row contains the selected outcome, eligibility rule, count, associated spend and cost per outcome. Diagnostic rows describe intermediate actions with their own definitions and denominators. A click-through rate belongs to impressions and clicks; a form completion rate belongs to a clearly identified form-start population. Do not silently calculate one stage's rate using another stage's unrelated population.

Educational example, not a reported customer result: advertising costs 800 units and records 400 landing visits, 80 form starts, 20 submitted forms and 8 accepted enquiries. Cost per submitted form is 40; cost per accepted enquiry is 100. Adding 400 + 80 + 20 + 8 and reporting 800 / 508 = 1.57 as acquisition cost would count journey steps as independent acquisitions. It would also conceal the distinction between raw volume and acceptance.

Google distinguishes primary and secondary conversion actions in its own reporting and optimization framework. That distinction needs configuration-specific review; your internal scorecard labels alone do not change platform behaviour. A business-important action is not automatically present, measured correctly or configured as an optimization goal. Inspect the actual action settings separately, and distinguish measurement changes from the reporting method proposed here.

An AdAce reporting example: two different result labels

In our own stored campaign data for September 27–October 3, 2026, the BBR Meta account reported USD 271.78 in spend and 15 results labeled lead. The Wonder Line Google account reported USD 194.245689 in spend and 26 results labeled conversions. Both datasets use America/Chicago reporting dates. We verified these aggregates from stored rows on October 4, 2026; no campaign was changed to prepare this example. The owner authorized publication.

The separate ratios are USD 18.12 per reported Meta lead and USD 7.47 per reported Google conversion, rounded to cents. They are not evidence that Google delivered cheaper customers: the result labels and provider attribution defaults differ, and we have not established sale value, unique customers or qualification for these aggregates. Do not add 15 and 26 into a sales KPI. Recent attributed results can still be revised. These are reporting examples, not evidence of uplift caused by AdAce Ads.

What diagnostics should explain

  • Give each action a definition owner and an explicit deduplication rule. A repeated button click should not silently become another new customer.
  • Use diagnostic rows to select investigations: many form starts with few submissions suggests a different question from many submissions with poor acceptance.
  • Keep a lineage note between stages. Counts from separate tools may not represent an identical population, so an apparent funnel percentage can be only a rough indicator.
  • Record the business consequence of the primary row. Accepted leads, orders and fulfilled orders are not interchangeable if cancellation or repayment changes their value.

Design the two-level scorecard

  1. Choose the decision the scorecard must support: acquiring qualified enquiries, paid orders or another defined business outcome. Agree on exclusions such as tests, duplicates and existing-customer service requests.
  2. Map diagnostic actions to specific questions. Keep only actions whose movement would change the investigation; do not add every available event just to make the table look comprehensive.
  3. Document source, date basis, account timezone and counting unit for each row. Mark inferred joins or unavailable stages rather than forcing a perfectly continuous funnel.
  4. Reconcile a small sample using permitted aggregate or pseudonymous records. Check that accepted outcomes belong to the same acquisition cohort before interpreting an acceptance rate.
  5. Publish business results first and diagnostics below. In AdAce Ads, inspect available event breakdowns in client context and request a read-only explanation. Review any proposed settings change separately from this scorecard.

Reporting labels are not platform settings

  • A business metric can also arrive late. Retain a maturity label rather than ranking a recent cohort on incomplete qualified outcomes.
  • Secondary does not mean useless, and primary does not mean causal. Diagnostic signals can reveal a measurement outage; the business row can still contain attributed outcomes.
  • Do not promise improved bidding merely from reorganizing a report. Platform configuration, data availability, permissions and the quality of the measured action determine what optimization can use.

Sources and further reading

Ace, the AdAce Ads mascot

Try it on your own accounts

Create a workspace, connect Google or Meta in a couple of clicks and see your accounts clearly. Changes follow your approvals or the policy you configure.

Create your workspace