Ask an active marketer what numbers they check every morning and you’ll get the same list: CPM, CTR, CVR, ROAS. Ask them their match rate on Meta or Google, the share of the audience they’ve uploaded that the platform can see and target, and you’ll usually get a moment. Many groups do not follow. Nala doesn’t know it’s a number at all.
That rest is expensive. Build an audience of 100,000 customers, upload it to a platform like 55% of them, and the campaign competes with 55,000 people. Another 45,000 are not visible in it, no matter how good the targeting or art. And match rate remains at the top of all metrics tracked by teams.
If the platform sees only a fraction of the audience you’ve built, every subsequent number (reach, frequency, conversion, return spent) is silently calculated against that same small audience. You can redo the creative, redo the bids, and rebuild your conversion model forever, and nothing touches a slice of your audience that the platform has never seen. The part to check is match breaks, why it’s difficult, and how much access groups are losing without seeing it on the dashboard.
The gap between the audience you build and the audience you reach
Here’s a mechanic that most teams haven’t explored. When you push a first-party audience to a paid platform, the platform doesn’t understand your “customers.” It targets a subset of your list that it can resolve to its logged-in users, usually by matching instant emails and phone numbers against identifiers in its accounts. Every record it can’t resolve is dropped, with no error and no warning. The campaign is against anyone who has survived.
Every privacy change of the past few years has widened that gap. Opting out a third-party cookie removes the connective tissue that binds identity to all sites. Apple’s App Tracking Transparency intercepts device identifiers. Walled gardens keep reinforcing their same concept. And the usual failure modes didn’t go away: a customer who signs up for a work email but uses a personal one for communications, a phone number formatted differently on each side, a three-year-old record. Identifiers are dispersed faster than most CRMs and CDPs can compile, so the distance between the audience you build and the audience you can reach grows, not shrinks, no matter how clean your data is.
Latent component: platforms report performance against a matched component. So the campaign looks good. You are measuring the effectiveness of the audience the platform found, not the audience you built, and the difference between the two is not visible in any report you open.
Four places are calling you right now
Most marketers who have thought about match rate lump it under the “retargeting problem.” It’s much broader than that.
- Adoption. Only partially matching seed and output lists make the test less accurate, and field training on partial signals requires more guesswork. That is often seen as increased CAC, and is never traced back to matching.
- Retargeting. Obvious, but state the statistics clearly: if your CRM list is like 45%, more than half of the customers you intended to engage with will never see the campaign again. The program is operating at less than half capacity, and its reported numbers do not say anything about the people it has never reached.
- Oppression. The sneaky one. The suppression list only suppresses the customers the platform sees. Every existing customer that isn’t the same doesn’t show up in your listings, so you pay acquisition prices to re-buy the people you already have, and some of them get a new customer discount that your loyal customers will never see. Low match rates don’t just waste budget; they pay for your own margin erosion.
- It looks like sowing. Models like models grow from the same part of your seed, not from the seed you uploaded. A weak match rate means the model is learning from a skewed small sample of your best customers, and that error compounds as the platform moves to millions of impressions.
Add it, and the level of matching is not a curiosity of the data group. It’s a tax on every dollar of paid consumption, and almost no one has measured how big it is.
What happens when you close the gap
This is not just a theory. CKE Restaurants, the company behind Carl’s Jr. and Hardee’s, served its audience with Rokt mParticle’s Match Boost to enrich the ad platform’s identifiers. Match rates increased to 117% in Google Ads and 29% in Meta.
Note what hasn’t changed: budget, creativity, campaign structure. The same money simply reached a larger audience that the brands had already built, and the ROAS improved on that same spend. That’s the signature of a matching ratio problem. When recognition increases, efficiency follows, because the waste you are removing was never seen in the first place.
This used to be a shopping project. Now the setting.
If the matching rate is ignored, part of the reason is that the adjustment used to be really painful. Developing recognition meant licensing third-party data: vendor evaluations, procurement cycles, legal reviews, integration creation, and months before you measured anything. The cost of maintenance is beyond what most teams can even calculate.
That is no longer the case of the problem. Enrichment happens increasingly in the area where the audience leaves your customer data infrastructure in the ad space, the setting is the connection rather than the system you are building.
Done right, we inherit the management you already have: identifiers you’ve intentionally released for privacy or legal compliance can be excluded, and the enhanced data is used to fine-tune the same on the fly, never being rewritten in your profiles or stored on the destination platform. Closing the gap was a stopgap decision, not a change in data strategy. That doesn’t mean it’s solved for everyone. It means that the excuses for not looking are gone.
How to check your game level
Measure the gap. It takes about thirty minutes.
- Choose your top three paying areas by spending money.
- Alternatively, compare the size of the list you uploaded against what the platform matched. Google Ads reports match rate in Customer Match uploads (bucketed, but closed enough); Meta shows the resulting audience size, which you can hold against your posted list. Most brands only land somewhere in the 40–60% range of email lists, far less than most teams think.
- Do the same check on your main compression list. He is the one who will kill.
If your numbers come back north of 70%, go back to improving the craft. If you are like many brands, you will find that you are paying full price to reach a small part of your audience. Every metric you already track is down to that one number, and most teams never look at it.
Written by: Joseph Rosenberg, Principal Product Manager, Proprietary, Rokt mParticle
