The order of external and website intent matters when the decision occurs before all signals have arrived. Compare externally flagged accounts at the same decision point and use website activity visible by that time. That answers whether website context adds information to the external signal. A different design can ask whether external-first and website-first sequences correspond to different outcomes once both events have occurred.
Start with the buyer decision
Imagine a procurement software seller tracking evaluation of purchase approval workflows. Some accounts first show relevant external research and then use the company's implementation guide. Other accounts visit the guide weeks before their external activity becomes visible. If you use whichever event happened later as time zero for one group and the first event for another, the follow-up clocks no longer start from an equivalent decision.
A workable protocol sets a weekly decision index for all eligible externally flagged accounts. Take the same lookback for site activity, capture site matching eligibility, and monitor a fixed outcome horizon. Treat missing observation as a separate condition. A team cannot confidently label an account external-only if the account had no usable website observation in the first place.
A source to check before implementation
IAB Tech Lab's ECAPI clean room guidance focuses on sharing event data. A common format helps compare event dates across collaborators. The analyst still needs the time an observation became available to the actual decision process. See the IAB Tech Lab ECAPI clean room guidance.
A fictional buyer walkthrough
External event occurrence is Monday, arrival is Thursday, and the weekly decision cutoff is Wednesday. A site event arrived Tuesday. That account should not join the externally flagged Wednesday cohort. The next eligible index can include it, subject to the published lookback and eligibility rules.
A clean-looking external-only record may actually lack approved website observation. Carry its uncertainty into a separate count. If many cases sit in that category, the study may need a different population before it can answer whether website context adds information.
Arrival order
Specify the contribution of the arrival order check to the study design before calculating outcomes. Keep occurrence timestamps and ingestion timestamps for both event families. For a real-time decision test, analysts can use only evidence that was actually available then.
The review record for the arrival order check should retain this disposition. Run a second descriptive chronology with occurrence times and explain what ingestion lag changes. For the arrival order 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. Save the unresolved question beside the decision so the next reviewer can reproduce the result.
Choose whether the unit is an account or an account-week
A recurring review can enroll an account once at its first eligible cutoff, or repeatedly after a defined cooldown. These choices affect both business meaning and inference. A first eligibility design asks what helps at the first routing decision. A weekly design may address continuing prioritization, but accounts can contribute several related observations. Document the selection rule and how later events are attributed. Do not report the raw number of rows as the number of independent accounts when the same company appears several times.
External-only definition
Require a defined opportunity for website observation and no qualifying visit inside the lookback. Distinguish genuinely unobserved site engagement from match gaps.
Turn the external-only definition check into an auditable row-level rule. When reviewing the external-only definition check, record which cases it moves into or out of the primary denominator. Recalculate accepted evaluation 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. For this case, the expected next step is clear. Present a separate unknown coverage group before reporting an external-only rate.
Try an adjacent case that differs on one important fact.
Before treating the external-only definition 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.
Follow-up window
Pick a horizon that fits an accepted evaluation, then use it for each group. Retain dates even if the CRM status is later revised. To close the review, write down what the decision owner is expected to do. Wait until the final index cohort has had the full horizon.
Apply the follow-up window 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. Predefine the primary view so the alternative can illuminate a limitation without becoming a search for the most favorable result. A documented exception can remain useful evidence while it waits for an approved decision.
When comparing externally flagged with known prior site engagement with externally flagged with usable observation and no prior qualifying site event, verify how the follow-up window 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.
Existing pipeline
Ask whether the existing pipeline check is measured equally well in both groups. An account in procurement may visit security pages because a seller asked it to. That is useful stage context but different from finding new opportunities.
Apply the rule while the source evidence is still available. Separate already-open opportunities from new-account qualification. For the existing pipeline 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. Where a previous result remains relevant, keep it as a separate snapshot rather than restating it under the new rule.
Consider how the existing pipeline check behaves when the buyer or operator supplies incomplete information. Define a useful next action for that state. That might be gathering a missing input, holding the case, or keeping an aggregate observation. The approved action should match the evidence that exists at the time.
Explain what the comparison can and cannot trigger
Combined evidence may point a seller toward a discussion about implementation, security or rollout. A campaign action requires its own destination eligibility and appropriate buyer context. If the study shows a higher accepted evaluation rate, a team could test an internal review queue before changing its ad or email policy. Use the observed event chronology to choose a helpful question without claiming which employee viewed a page. A test of a seller or campaign intervention should independently record assignment, delivery and qualification.
Action inference
A higher outcome rate in the combined group can support prioritizing a review. It cannot by itself prove that an additional ad or sales email will create incremental demand.
Write down how the action inference check affects the eventual claim. When reviewing the action inference check, record which cases it moves into or out of the primary denominator. Recalculate accepted evaluation 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. Put the next step in the operating notes. Use an appropriately randomized follow-up test if intervention value is the question.
The minimum practical audit is an example a colleague can replay.
Connect the action inference check to the report on accepted evaluation. Establish whether an exception changes eligibility, leaves the result unknown, or just produces a note. Removing a record and counting it as a failure have different effects on the reported percentage. Describe the handling clearly enough for a second reviewer to reconstruct it.
Build the review worksheet
Build a cutoff table by eligible account and week. Keep external observation time, external receipt time, eligible site observation status and a website event flag from the same lookback. Site-unobservable accounts belong in their own group. An external account that appears in multiple weeks needs a predefined indexing and dependence policy.
Separate a buyer who visits a security guide after a seller sends it from someone who discovers the guide independently. Both can be valid site engagement. The seller-directed event reflects a different pathway, especially if the account already had an evaluation. Show results separately if the decision depends on identifying new demand.
A worked comparison of accepted evaluation
In the fictional worksheet below, the denominator consists of accounts externally eligible at the same weekly cutoff. The numerator counts cases that entered a documented evaluation by the end of follow-up. 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 |
|---|---|---|---|
| Externally flagged with known prior site engagement | 17 | 130 | 13.08% |
| Externally flagged with usable observation and no prior qualifying site event | 16 | 210 | 7.62% |
The calculation gives 13.08 percent for externally flagged with known prior site engagement and 7.62 percent for externally flagged with usable observation and no prior qualifying site event. The first is 5.46 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 17 and 16. A report that drops the underlying counts can hide that relationship.
This comparison can inform whether the second evidence layer is useful at the cutoff. Site observability, current pipeline and company size can still influence the rates. It does not establish a benefit from sending an ad or email.
Record the snapshot date beside the accepted evaluation result. For this hypothetical comparison, use the same written review standard for both groups. If a reviewer changes the comparison rule, keep the original accepted evaluation calculation and clearly identify the revised version.
For a real accepted evaluation 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.
Check whether campaign activity creates part of the measured signal
A seller's link can lead to visits on the very page used to define website engagement. If one group already had more seller contact, it may have more observed site behavior and more opportunity to qualify. Record approved outreach timing and any opportunity stage before the cutoff. Include a pre-specified view that excludes sales-directed visits when the business wants an independent web signal. The comparison should explain exactly which kind of engagement it includes.
Before reusing the fictional accepted evaluation comparison, give each group an observation date and a reason it belongs in the report. The 130 and 210 denominators are the eligible groups in this example. A report restricted to the 17 and 16 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.
For a real implementation, check whether an account that appears website-first was recognizable as the same account on both dates. A later CRM merge can recover a relationship retrospectively without making the earlier observation available to sales. Keep two timelines when necessary. The retrospective timeline can describe behavior. The contemporaneous timeline is the evidence for testing a routing decision at the cutoff. Keep the assignment rule separate from the matching method.
For more on the signals behind this decision, read the Cortex intent data insights.
Decide what would change your mind
Test whether prior website activity adds actionable evidence among externally eligible accounts at the same cutoff. Sketch the event timeline for an external-first account and a website-first account. Include both the event and ingestion dates. This is a quick way to discover whether the team can use the signals for live triage or can only study the sequence retrospectively.
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.
