Plan an Intent Data Holdout When Accounts Share a Parent Company

November 20, 2026•11 min read

When account branches share buyers or a procurement team, randomize the group likely to receive the same treatment. That might be a parent company, a shared owner or a sales territory. Otherwise an ad, referral or seller task can reach a branch assigned to the comparison group. Decide which result you want before choosing the assignment unit.

Where the decision gets complicated

A fictional industrial services company targets plants owned by multi-site manufacturers. A corporate engineer compares vendors and forwards a scope template to several plant managers. If one plant is assigned campaign follow-up and another sits in the holdout, the shared engineer can carry the intervention across both groups. Randomizing the parent can reduce this kind of spillover, but it leaves fewer independent units for analysis.

Preserve routine inbound service. A holdout for additional campaign activity still needs to respond to direct buyer inquiries under the existing policy. A design that withholds requested service asks a different question and changes the buyer experience. Keep assignment, delivered exposure and actual contact separate so later analysis can describe deviations honestly.

What to record at handoff

Before assignment, map parent relationships using approved evidence and freeze the list. Mark plants with shared buyers and seller territories that could lead to spillover. Simulate assignment counts at the cluster level and review whether one large parent could dominate the outcome measure.

Record which campaign actions were available to each assignment and which actually occurred. Include direct inbound replies in ordinary service. If marketing reaches an unintended cluster, keep assignment unchanged for the primary analysis and document delivery deviations rather than quietly moving the cluster.

Cluster map

Write down how the cluster map check affects the eventual claim. Identify ties likely to transmit the intervention, such as common ownership, seller assignment or shared purchasing. Document unresolved ties before randomizing.

To close the review, write down what the decision owner is expected to do. Use the group that fits the spillover mechanism and keep cluster creation independent of observed outcomes. For the cluster map 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.

Before treating the cluster map 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.

Decide how to analyze unequal clusters

A large parent with many eligible sites creates more chances for an opportunity than a single-site parent. If each parent has equal weight, the estimated outcome addresses the average parent. A site-weighted analysis asks another question and can be dominated by one big group. Define the primary estimand in plain language before looking at results. If the design stratifies by cluster size or territory, preserve the assignment strata for analysis. Use uncertainty methods compatible with the number of independent clusters, particularly when the trial includes few large parents.

Assignment procedure

Balance known design factors such as cluster size, geographic coverage and prior activity when appropriate. Save the randomization rule and timestamp.

Specify the contribution of the assignment procedure check to the study design before calculating outcomes. When reviewing the assignment procedure check, record which cases it moves into or out of the primary denominator. Recalculate qualified parent accounts 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. Apply the rule while the source evidence is still available. Keep all branches under their original cluster assignment even if a buyer contacts you directly.

When reviewing a batch of examples, keep cases with unknown outcomes visible.

When comparing assigned additional intent activation with assigned standard process, verify how the assignment procedure 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.

Follow two fictional records

A hypothetical parent has six eligible sites and two shared buyers. Assigning locations separately risks exposing the same buying group to both approaches. Assigning the parent gives a cleaner operational boundary, but a portfolio of only a few parents may lack the precision needed for a decision.

If the parent has one qualified opportunity that covers all six sites, count it once for a parent-level success measure. A site-level measure would need a separate rule for projects spanning sites and an analysis that respects clustering.

Outcome unit

A cluster with many sites may have a greater chance of producing at least one opportunity. An opportunity-per-site analysis answers a different question. Put the next step in the operating notes. Specify whether the target is impact per parent, per eligible site, or total pipeline.

Apply the outcome unit 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 decisions that require sensitive detail with authorized reviewers and export only the flags needed for analysis. Ask whether the resulting next step answers the question the buyer may reasonably have.

Consider how the outcome unit 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.

Contamination log

For the contamination log check, plan a sensitivity check that changes one defensible assumption at a time. Record unusual outreach and shared exposures within approved evidence limits. Avoid removing inconvenient clusters after seeing outcomes.

The review record for the contamination log check should retain this disposition. Analyze assignment as the primary view and discuss observed spillover as a limitation. For the contamination log 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.

Connect the contamination log check to the report on qualified parent accounts. 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.

Protect the usual buying experience in the comparison group

A prospect should still get service for an unsolicited inquiry under the normal process. Define what additional treatment is withheld, who knows the assignment, and what happens when the same account shares an approved opportunity with another site. Make the intervention understandable to sales, including what counts as a deviation. Report realized outreach and exposures as supporting evidence while preserving assigned groups for the primary comparison. This helps a client distinguish a test of offering a campaign from a comparison of people selected by actual exposure.

Maturity review

Give clusters equal opportunity to develop outcomes and disclose unresolved deals. If only a few large groups are present, uncertainty can be substantial.

Ask whether the maturity review check is measured equally well in both groups. When reviewing the maturity review check, record which cases it moves into or out of the primary denominator. Recalculate qualified parent accounts 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. Use a method appropriate for cluster assignment and avoid treating every contact as independent.

Try an adjacent case that differs on one important fact.

Put a failure example beside the standard example for the maturity review check. For the first example, require the reviewer to explain why the rule should stop. For the second, ask why it should advance. This exposes definitions that sound precise in a meeting but do not lead to consistent decisions under ordinary conditions.

A source to check before implementation

The IAB cross-channel guide opened for public comment October 20, 2026. Its distinction between attribution and incrementality fits a holdout: assignment supports a different inference than reviewing contacts that happened to receive an ad. See the IAB cross-channel measurement consultation.

A fictional qualified parent accounts check

The following example is hypothetical and uses parent clusters assigned to each fictional group as the denominator. It counts cases that produced at least one opportunity meeting the test rule. Save the underlying counts together with the label used to report qualified parent accounts. Totals and rates will answer different questions when group sizes differ.

Illustrative groupCount meeting ruleEligible countRate
Assigned additional intent activation94221.43%
Assigned standard process74415.91%

The calculation gives 21.43 percent for assigned additional intent activation and 15.91 percent for assigned standard process. The first is 5.52 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 9 and 7. A report that drops the underlying counts can hide that relationship.

This fictional difference is per assigned parent, not per branch or contact. Without cluster-level uncertainty and design checks it would be premature to scale based on the difference.

Record the snapshot date beside the qualified parent accounts result. For this hypothetical comparison, use the same written review standard for both groups. If a reviewer changes the comparison rule, keep the original qualified parent accounts calculation and clearly identify the revised version.

A publishable comparison of qualified parent accounts needs dated cohort membership, an approved aggregation policy and a reproducible calculation. Use a method that fits the assignment and overlap. The study owner should be able to give a reviewer a permitted replication path before a public number is written.

Plan for differences in seller behavior

Sellers may work with several sites under the same parent, or they may support both assignment groups. Write down which tasks each group should receive and preserve a limited log of deviations. If sellers learn which clusters have additional signals, they may change their normal handling. That is part of the real operating environment, but it should be documented. Consider whether assignment at the seller level better matches the spillover mechanism if enough independent groups exist.

Before reusing the fictional qualified parent accounts comparison, give each group an observation date and a reason it belongs in the report. The 42 and 44 denominators are the eligible groups in this example. A report restricted to the 9 and 7 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.

A research owner reviewing qualified parent accounts needs a way to reproduce the comparison and a documented decision about what is publishable. Confirm that every released aggregate has the right unit, denominator and approved scope. Keep the hypothesis and method available even if the actual outcome data are not sufficient to support a result.

Keep holdout assignments out of model training

An experiment can change its meaning if an updated model learns from outcomes that should remain separate for evaluation. Document which feedback is eligible and what the baseline knows. Preserve each parent's assignment even when authorized customer information from outside the initial list later contributes to learning.

For more on the signals behind this decision, read the Cortex guide to buyer intent data.

Give a reviewer enough evidence to challenge the design

Use parent-level assignment when it matches the way campaign effects can spill across sites. Map who actually shares buying decisions before picking a randomization unit. That discussion can reveal whether a parent, seller or territory is the better assignment boundary. Take one expected exception through the normal inbound process so the study does not interrupt service.

See how Cortex Marketing AI uses customer feedback to refine the signals a team can test.

Bring your anonymized data inventory, a proposed comparison, and the approvals available for aggregate research. For help designing a test around that decision, call Cortex Marketing AI to book an intent consultation.

Back to Blog