Reach counts unique people. Impressions count exposures. The same user seen three times is three impressions but one in the reach number — reach is the deduplicated complement, sized over a defined period. What “unique” resolves to depends on the platform’s identity layer: it can be a device, a logged-in account, or a modeled household.
The arithmetic identity that ties the two together is:
Reach–Frequency Identity
Reach × Frequency = Impressions
A campaign that delivered 1,000,000 impressions to 200,000 unique users has a frequency of 5.0. This is the single most useful decomposition operators forget when reading platform reports, because the reports lead with impressions and spend and hide the unique-user denominator a level down. Rearranging to frequency = impressions / reach is usually the first thing to do when a campaign’s efficiency drifts and the cause isn’t obvious.
The reason that decomposition matters is audience saturation. Incremental return per impression tapers as frequency climbs against a fixed audience, and later impressions of a creative to the same person convert at materially lower rates than the first — the curve shape varies by category and creative, but the direction does not. That decay is what creative fatigue tracks, and frequency caps are the lever operators use to bound the frequency side of the identity. On brand-building objectives reach is the headline goal; on direct-response campaigns it’s a diagnostic — a flat ROAS on an exhausted audience supply looks identical to a flat ROAS on a creative problem until the reach number tells you which one you have.
The measurement caveat is that platform-reported reach is deduplicated within one ad account on one platform, and cross-platform reach is not additive. Adding Meta + TikTok + YouTube reported reach overstates true unique reach, sometimes materially, because a person present on all three is counted once inside each platform’s number rather than once globally. True cross-platform unique reach requires a measurement partner (Nielsen, Comscore) or a probabilistic identity stitch. ATT and ad-blocker effects also mean platform-reported reach can understate true exposure on iOS in particular, where the identity signal the dedup runs on has narrowed.