The Cohort Your CRM Onboarding Vendor Marks as 'Matched' Includes Records That Will Never Receive an Impression

When a CRM onboarding vendor returns a match report, the number that draws the most attention is the match rate. A file goes in, a percentage comes back, and that percentage is treated as the size of the reachable audience. The problem is that a match and a deliverable impression are two different things, and the distance between them is rarely explained on the summary slide.

Understanding that distance is one of the more useful things a media buyer can do before committing budget to a CRM-based campaign.

What a Match Actually Confirms

A match means that a record in your CRM file, typically an email address or a postal record, was found in the identity graph the onboarding vendor maintains. The vendor links your record to a graph node, and that node may carry one or more device identifiers, cookies, hashed emails, or probabilistic signals attached to it.

The match is a graph-level event. It says nothing about whether the person attached to that node is reachable in the specific inventory you plan to buy. Reachability in a DSP depends on whether a bid request carrying an identifier associated with that node actually appears in an auction during your campaign flight. That depends on device activity, publisher participation, identifier type, and timing. None of those factors are present in the match report.

Why Matched Records Go Dark in Delivery

Several conditions can cause a matched record to produce zero impressions even when everything upstream appears to have worked.

First, the identifier type attached to the matched node may not be supported in your target inventory. A graph node anchored primarily to a third-party cookie has limited reach in mobile app environments. A node with only a mobile advertising identifier may not activate well in connected TV. The match is real, but the identifier travels poorly across environments.

Second, the publisher supplying the inventory you want may not participate in the same identity ecosystem as your onboarding vendor. Activation typically requires the onboarding vendor to translate your audience into a form the DSP can bid against, which then requires a publisher to accept and recognize the translated identifier. If there is a gap anywhere in that chain, the matched record simply does not receive a bid.

Third, device and identifier graphs decay between the time the match is made and the time your campaign actually runs. A person who shared a device identifier signal with a publisher six months ago may not be generating that signal today. The match was accurate when it was created and may not reflect the current state of that node.

Fourth, frequency and supply constraints mean that even a fully addressable matched record will only convert to an impression if supply is available at the moment that person's device generates a bid request within your targeting parameters and budget. Low-volume or niche audiences can have high match rates and still deliver against only a fraction of those records.

How to Pressure-Test a Match Report Before You Commit Budget

The most useful question to ask an onboarding vendor is not what the match rate is. The more useful question is what share of matched records are expected to generate at least one impression across the specific inventory and channels in the plan. Some vendors can provide this as a projected reach estimate broken out by environment. Others cannot, and that inability is itself informative.

A second useful question is which identifier types are attached to the matched nodes, and in what proportion. If most matches resolve to a single identifier type that performs poorly in your priority channel, the effective reach will be much lower than the headline match number implies. Asking for a breakdown by identifier type before activation rather than after gives you something to act on.

It is also worth checking how recently the graph was refreshed for the records in your file. Onboarding vendors update their graphs on different schedules, and some identifiers are refreshed less frequently than others. A match against a stale graph entry carries more uncertainty about current addressability than a match against a recently observed signal.

If your onboarding vendor or DSP can run a small test flight before full budget commitment, the delivery rate against matched records in that test gives you a more reliable projection than the match report alone. A hypothetical example: if a test flight against 100,000 matched records delivers to 40,000 unique devices over two weeks, that observed delivery rate is a more honest planning input than assuming all 100,000 are continuously reachable.

What to Report Internally

Match rate is a reasonable vendor evaluation metric when you are comparing onboarding vendors against the same file. It does not translate cleanly into an audience size for a media plan.

When building internal projections, it is worth separating three numbers: the match count from the onboarding vendor, the estimated addressable reach in the channels you plan to buy, and the actual delivered unique reach after the campaign. Tracking all three across campaigns builds a reference point for how much each layer of translation costs you in effective reach. Over time, that reference point is more useful than any single match report.

Some teams find it helpful to apply a conservative adjustment to match counts when building audience size estimates for planning, based on prior campaign experience. The right adjustment varies by file quality, channel mix, and vendor, so there is no universal percentage to apply. The goal is to avoid planning a campaign around an audience that only exists at the graph-matching stage and has not been tested for delivery.

The Structural Reason This Gap Persists

Vendors report match rates because match rates are what they can measure with certainty at the time of onboarding. Delivery is a downstream event that depends on systems outside the onboarding vendor's direct control. Reporting on something you can measure precisely is easier than reporting on something uncertain, even if the uncertain number is more useful to the buyer.

This is not a criticism of any particular vendor practice. It is a structural feature of how CRM onboarding works across the industry. The match happens at one layer, activation happens at another, and delivery happens at a third. Each layer introduces its own attrition, and the full picture only becomes visible after a campaign runs.

Buyers who understand this structure before the campaign starts are in a better position to set realistic reach expectations, ask sharper questions during vendor conversations, and interpret delivery reports accurately when they arrive.