Most prospecting budgets have a quiet leak: impressions served to people who already bought, unsubscribed, or work for the company. A suppression audience is the fix. It is a list of known people that an advertiser tells a platform not to show a campaign to.
What it actually does
Suppression began in direct mail, where names were removed from a mailing file before print. Platforms rarely use the same word. Google Ads and Meta call the function an exclusion, Amazon calls it an exclude group, and OpenAI calls it an exclusion audience. The process is simple: define a segment, bring in the records, match them to platform accounts, and apply the list as an exclusion on a campaign or ad set.
The records usually come from a CRM or from behaviour captured by a tag. Uploads are normalised and hashed with SHA-256 before transfer. Since April 1, 2026, Google’s Data Manager API is the only route for programmatic Customer Match uploads, with a maximum of 10,000 members per request and up to 10 identifiers per member.
The 540-day trap
Membership duration is the setting teams overlook. Google Customer Match membership is capped at 540 days, and the clock starts at upload or refresh rather than the last purchase. A suppression list that is never refreshed decays, so a buyer from two years ago can quietly drop back into an acquisition campaign.
- Fast-turnaround brands may suppress recent buyers for 30 days or less.
- Durable goods brands may need suppression windows of years.
- Match rate, not just list size, decides how complete the exclusion is.
Why it matters for ROAS
The commercial case is incrementality. If ads take credit for purchases already under way, reported cost per acquisition looks strong while true return suffers. Field experiments at eBay, published in 2015, found most search spend reached frequent buyers whose behaviour the ads did not change. Excluding those buyers can make in-platform numbers look worse while improving real return.
Automated buying has absorbed suppression into bidding. Google’s Performance Max offers a new customer acquisition goal, and Microsoft added a comparable mode on January 15, 2026 that stops serving ads to uploaded customer lists. Amazon’s DSP introduced simplified audience targeting on November 21, 2025 with a single exclude group across up to 10 include groups.
What goes wrong
Leakage is the first weakness. Unmatched records, logged-out sessions, shared devices and new email addresses slip past an exclusion. Google can take up to 48 hours to process a list, and a 40% match rate on a 50,000-customer file suppresses only 20,000 people. The rest remain eligible to see acquisition ads.
There are also control and comparability issues. Meta stopped allowing detailed targeting exclusions in new ad sets after January 21, 2025, and its existing customer budget cap for Advantage+ shopping was no longer available by March 2025, according to Jon Loomer. Custom Audience exclusions still work, but they are a harder control in fewer places. Over-suppression has a cost too: excluding every past customer removes people who are likely to buy again.
Do not confuse suppression with negative keywords, frequency capping, or a holdout group. Negative keywords block queries, frequency capping limits repeat exposure, and a holdout withholds ads at random. A suppression audience withholds delivery from specific known people for a reason.
What to do next
Refresh suppression lists before the 540-day cap, check the match rate rather than just the list size, and apply exclusions at prospecting level instead of across every campaign. A tight suppression strategy saves budget and protects incrementality without turning off likely repeat buyers.
Source: PPC Land
