A buyer intent event dictionary defines what was observed, which entity it refers to, and which version of the tracking rule created it. Start with business meaning rather than a page name. A website redesign can preserve the same buying action across multiple URLs. It can also change the action while leaving the URL intact. Analytics needs a record of both possibilities.
Start with the buyer decision
Imagine a vendor combining a pricing overview and product configuration page into one interactive quote builder. Under the old design a visitor could read a general price range. Under the new design the site logs a quote start after a buyer selects equipment type and quantity. Counting both as a pricing visit hides the effort difference. Requiring the exact old URL would make the former event vanish even for behavior the site still supports.
Name events as observations and keep interpretations in a separate field. A selected quantity is observable. Calling the visitor ready for a sales call adds an inference. Record fields for expected timing, eligible traffic, duplicate rules, policy context, version and owner. Pair each definition with an example that qualifies and one that looks similar but should fail.
A source to check before implementation
IAB Tech Lab's 2026 ECAPI materials emphasize common event definitions for exchange. An internal event dictionary complements that work by documenting the actual behavior behind the same event label across site versions. See the IAB Tech Lab ECAPI clean room guidance.
A fictional buyer walkthrough
Old version records a guide open. New version records an accepted configuration with quantity. During the overlap, a buyer can produce both events with different effort and meaning. The comparison table should show parallel rows, not one trend spliced at the cutover.
A simpler page URL change may preserve an identical confirmation event. Prove parity with marked test runs before combining periods. Record the approved dictionary entry and the date on which both implementations were checked.
Canonical event
For the canonical event check, plan a sensitivity check that changes one defensible assumption at a time. Choose an observable action such as completing a supported quote configuration. List the required parameters and accepted routes.
For this case, the expected next step is clear. Keep page paths as implementation details and map them to the event where they still mean the same thing. For the canonical event check, a study reviewer should be able to explain the decision time, comparison unit, and evidence that would justify any exclusion. Where those conditions differ between groups, report the difference as part of the method instead of presenting the final rate as self-explanatory. When available evidence disagrees, route the conflict to the accountable owner instead of automatically choosing whichever system returned the more favorable status.
Distinguish the business event from the code event
A developer can fire a tag whenever the page emits a particular action. The dictionary defines when that action is meaningful for the business. A quote start could mean opening a page, choosing a supported configuration or submitting enough information for a review. Each might have a useful role. Give each a separate label when it changes the interpretation. Make the dictionary identify which inputs need to be present. An analyst looking at a drop in quote starts should be able to distinguish a real behavior change from a new validation rule.
Non-event
Opening a quote builder, clicking a help tooltip and submitting a complete configuration are distinct actions.
Ask whether the non-event check is measured equally well in both groups. When reviewing the non-event check, record which cases it moves into or out of the primary denominator. Recalculate version parity using the approved definition, then assess whether unresolved cases could change the interpretation. If they could, report the size of the gap rather than applying a favorable assumption. To close the review, write down what the decision owner is expected to do. Put a near-miss in the specification so the tag engineer can write a meaningful negative test.
Distinguish the rule itself from the effect of applying it.
Agree on the point at which the non-event check becomes a decision someone can act on. Earlier evidence can stay visible in the approved working record. The final disposition should say which requirement was satisfied. A later change can then update the decision without turning earlier uncertainty into a historical fact.
Version cutover
Run old and new definitions in parallel when approved and practicable. Store their separate identifiers and compare account-level coverage. Apply the rule while the source evidence is still available. Document periods that cannot be compared rather than splicing a broken trend together.
Apply the version cutover check rule before seeing the outcome totals. This keeps a retrospective analysis from silently favoring one group. If the definition genuinely needs correction, describe why, keep a dated revision, and show the impact when the prior definition was feasible. Keep those details adjacent to the finding so they survive when it is copied into a decision memo. They often show a gap in timing, coverage, or ownership that a filter would hide.
In the worksheet, attach the expected result of the version cutover check to a dated record. Ask the reviewer to point to the evidence that supports it. For a case that cannot be resolved, record what is missing and leave the substantive decision open until the right owner can respond.
Meaning change
Specify the contribution of the meaning change check to the study design before calculating outcomes. When the new interface removes a required buyer input, the same event label may overstate the strength of the action. Review fit with sales.
Put the next step in the operating notes. Give genuinely different behavior a new event family or version and preserve the earlier definition. For the meaning change check, a study reviewer should be able to explain the decision time, comparison unit, and evidence that would justify any exclusion. Where those conditions differ between groups, report the difference as part of the method instead of presenting the final rate as self-explanatory. If the evidence is too weak for a person-level action, retain it only for the approved account, diagnostic, or aggregate purpose.
Before treating the meaning change check as ready for routine use, agree how to handle evidence that arrives after the initial decision. Identify which current state can change and which original snapshot remains frozen. This preserves the difference between the latest information and the evidence available when the earlier action was approved.
Verify what happens when the same buyer changes paths
Test a reader who uses a legacy URL and is redirected to the new builder. Test a return after saving a configuration and a case where a person stops before submitting. Verify which events are generated and whether duplicates are merged under the documented key. Check the schema version on the output rather than assuming a successful page interaction reached the analysis table. Update the seller's documentation if the new event requires more commitment. A stronger event can justify different routing without representing a measured change in demand.
Reprocessing
Do not assume a historical event can be reconstructed from the fields you kept. Record which versions can be translated and with what loss.
Turn the reprocessing check into an auditable row-level rule. When reviewing the reprocessing check, record which cases it moves into or out of the primary denominator. Recalculate version parity using the approved definition, then assess whether unresolved cases could change the interpretation. If they could, report the size of the gap rather than applying a favorable assumption. The review record for the reprocessing check should retain this disposition. Mark estimated mappings clearly and avoid mixing them into a primary benchmark without disclosure.
To verify the handoff, ask the receiving owner to explain the decision in their own words using only the approved record.
When comparing revised tracking build with initial cutover build, verify how the reprocessing check was handled in both views. Different rules can explain an apparent difference even if the underlying cases behave similarly. Keep the rules visible with the results and show any material exception when the two groups cannot be judged on equivalent terms.
Build the review worksheet
Give each intent event a business name, observed behavior, version, eligible route, required fields, duplicate key and owner. Record dates when the code and the definition changed. Provide a test example for an incomplete configuration and another for a full submission.
When two page layouts differ, look at the actual action and not just the data name. A button that advances to step two may replace an old button that confirmed the request. Verify both against a trace and keep metrics separate when the meaning changed.
A worked comparison of version parity
In the fictional worksheet below, the denominator consists of scripted journeys expected to retain the same business event. The numerator counts cases that produced a comparable event under both schemas. Both figures belong in the report. A reader who has only a percentage cannot tell whether a handful of outcomes drove the apparent difference.
| Illustrative group | Count meeting rule | Eligible count | Rate |
|---|---|---|---|
| Revised tracking build | 28 | 35 | 80.00% |
| Initial cutover build | 21 | 35 | 60.00% |
The calculation gives 80.00 percent for revised tracking build and 60.00 percent for initial cutover build. The first is 20.00 percentage points above the second, using the underlying rates before rounding. The rate gap and the volume gap are separate. In this example the two numerators are 28 and 21. A report that drops the underlying counts can hide that relationship.
These are script-level parity counts, not buyer conversion rates. The failed test cases are the main deliverable. Keep a separate list of events intentionally changed by the redesign.
Record the snapshot date beside the version parity result. For this hypothetical comparison, use the same written review standard for both groups. If a reviewer changes the comparison rule, keep the original version parity calculation and clearly identify the revised version.
For a real version parity study, retain cohort dates, one primary eligibility rule, a fixed outcome cutoff, and a note about any missing observation. Repeated units may introduce dependence and the comparison can be sensitive to outcome maturity. Choose an uncertainty calculation that matches the study rather than treating all rows as independent by default.
Audit an apparently smooth transition
A stable daily volume can hide a change in meaning when the new tracker sends more events on some paths and fewer on others. Compare known journeys and reconcile event-level rules before declaring parity. Record whether the event name maps across versions, whether a feature was added, and what changed in the buyer's effort. When the change cannot be translated, start a new series and explain the boundary in the report.
Before reusing the fictional version parity comparison, give each group an observation date and a reason it belongs in the report. The 35 and 35 denominators are the eligible groups in this example. A report restricted to the 28 and 21 successes would lose the reference needed to interpret the result. Retain a separate ledger for exclusions and date the moment each group completed its observation window or test sequence.
If the next step is a real version parity study, register the population and observation dates before inspecting results. Include which actions may affect the outcome and how those actions are measured. Ask a reviewer to identify an alternative explanation that would remain even if the calculation itself is correct.
Teach the model which event really changed
Customer-specific scoring can incorporate first-party activity and approved sales outcomes. If the event called quote start begins to represent more buyer effort, record that change before the model consumes the new feed. Earlier and later events can remain useful when each carries a definition and a version the reviewer can interpret.
The migration owner should sign off on comparability by action and version rather than approving the aggregate event count alone.
For more on the signals behind this decision, read the Cortex guide to buyer intent data.
Decide what would change your mind
Combine measurement periods only where behavior and version evidence establish comparability. Bring a tracking rule that changed after a website edit. Review whether the new event means the same thing to the buyer, seller and analyst. That determines whether you can combine the periods or need to start a new series.
See how Cortex Marketing AI uses buyer research and customer outcomes to inform its scoring approach.
Bring the unit of analysis, an example outcome definition, and the observation windows you can support. To put the decision into practice, call Cortex Marketing AI to arrange a consultation.
