Average frequency can hide the exposure pattern that matters most.

We set the frequency cap from the campaign's funnel stage, then break the average into an exposure distribution. That shows the difference between someone who saw the ad once and someone who saw it twenty times. We review reach and frequency together across placements and devices because a change to one reshapes the other. For teams whose only frequency number is a single average nobody has ever broken down further.

A frequency distribution you can explain, showing who is underexposed, who is overexposed, and what changed in response.

A frequency distribution chart revealing an over-exposed cluster hidden behind a normal-looking average

Some of the 500+ brands we've worked with

See all references
  • EY
  • Akakçe
  • İstikbal
  • Axa Sigorta
  • Sina Pırlanta
  • Bundle
  • Karel

We begin with an agreed cap, build the exposure distribution and then check whether it moved after an adjustment. AI can compile and deduplicate large reach and frequency exports. The media lead sets the cap, and only the authorized owner can change the budget.

How we hold ourselves to it

  • Set the frequency cap around the funnel stage
  • Read the full distribution behind the average
  • Deduplicate reach across placements and devices before reporting it
  • Treat reach and frequency as one tradeoff
  1. Set the cap to the funnel stage

    Decide the frequency cap and reach goal from the campaign's role, whether it is meant to introduce, remind, or convert. Leave the previous campaign's default behind.

    Frequency cap and reach goal set from the campaign's role and recorded with the reasoning.

    AI assist
    Drafts a recommended cap range from the stated objective and prior campaign data.
    Human gate
    Media lead and client owner approve the cap and reach goal.
    Owners
    Media lead + client owner
  2. Deduplicate reach across placements and devices

    Reconcile exposure where placement and device reporting can be compared. Keep every gap visible before drawing a conclusion from the deduplicated reach estimate.

    Deduplicated reach estimate with every unreconciled gap left visible.

    AI assist
    Runs the cross-placement and cross-device deduplication and flags any gap in the data.
    Human gate
    Media lead confirms the deduplicated number before it's used in reporting.
    Owners
    Media lead + analytics lead
  3. Build the supported frequency distribution

    Break deduplicated reach into a distribution that shows how many people saw the ad once, a few times, or well beyond the cap.

    Frequency distribution showing exposure across the deduplicated reach, not only the average.

    AI assist
    Builds the distribution chart from the deduplicated exposure data.
    Human gate
    Media lead reviews the distribution for anything the average was hiding.
    Owners
    Media trader + analytics lead
  4. Flag over- and under-exposed segments

    Identify which placements, audiences, or devices are driving excess frequency, and which parts of the intended audience are barely being reached at all.

    Over- and under-exposure register naming the placements, audiences or devices involved.

    AI assist
    Clusters the distribution by placement and audience to surface the outliers.
    Human gate
    Media lead decides which segments need a real adjustment.
    Owners
    Media lead + media trader
  5. Rebalance budget and caps

    Adjust budget, caps, or placement mix to correct the over- and under-exposure found, and record what changed and why.

    Applied budget, cap or placement-mix change with the evidence recorded.

    AI assist
    Drafts the rebalancing recommendation with the distribution evidence behind it.
    Human gate
    Client owner approves any change beyond the pre-agreed adjustment range.
    Owners
    Media trader + client owner
  6. Re-check the distribution after the change

    Confirm that the rebalance moved the distribution in the intended direction. The average alone cannot show that.

    Post-change distribution read confirming, or failing to confirm, the intended shift.

    AI assist
    Compares the new distribution against the prior one and flags whether the intended shift happened.
    Human gate
    Media lead signs off on the result or escalates if it didn't work as expected.
    Owners
    Media lead + analytics lead

These artifacts show who was reached, how often and where the reporting has gaps. Each figure links back to the distribution and its coverage.

  • Frequency distribution report

    The full spread of exposure across the campaign's deduplicated reach, with the average kept in context.

    Accepted when

    The full spread of exposure is reported, and the average is presented in its context rather than on its own.

    Cadence: Each optimization cycle

  • Cross-device reach reconciliation

    How exposure across placements and devices was deduplicated into one reach number, with any unreconciled gaps stated.

    Accepted when

    The deduplication method is stated and every unreconciled gap is named before the reach figure is used.

    Cadence: Each distribution report

  • Over/under-exposure register

    Which segments, placements, or audiences were driving excess frequency or getting under-reached, and what was done about it.

    Accepted when

    Each segment, placement or audience driving excess or missing exposure is named with the action taken.

    Cadence: Each rebalance

  • Rebalance decision log

    What budget, cap, or placement changes were made, the evidence behind them, and how the distribution moved afterward.

    Accepted when

    Every cap or budget change records its evidence and how the distribution moved afterwards.

    Cadence: Every rebalance

The average leaves the exposure pattern out. A campaign can report an average frequency of four while the same person receives twenty impressions and someone else receives none. The distribution reveals that imbalance. If reach isn't deduplicated across placements and devices first, even the average may be wrong.

A good fit when

  • The campaign spans several placements or devices, but no deduplicated reach figure shows how much exposure is being counted twice.
  • Average frequency sits below the cap, yet nobody can see who saw the ad once and who saw it twenty times.
  • Over- and under-served segments are visible in the distribution, so the media owner can tie a cap or budget change to a named exposure problem.
  • Frequency gets reported as one average number, with no breakdown of the distribution behind it.
  • The same person's exposure across desktop, mobile, and in-app inventory never gets reconciled into one number.
  • A frequency cap gets set once at launch from a rule of thumb, without reference to the funnel stage or objective.
  • Reach and frequency get optimized separately, so a fix to one quietly makes the other worse.

Better handled as other work when

  • One small, single-placement buy has little exposure spread to inspect, so a full frequency distribution would add no useful decision evidence.
  • You want a single frequency number proven optimal. No universal rule provides one. The right number depends on the funnel stage and creative.
  • Walled or unreported inventory hides placement-level reach and frequency data, so the method cannot build a supported distribution for the whole buy.

Paid Search, Paid Social, CRO, and Programmatic each run under a named owner at Zeo. The consultants below are matched to the channel this page is about, so you can see who you'd actually work with.

  • Nielsen ONE

    the cross-media currency a flight's deduplicated reach and frequency get checked against

The current buy gives us the starting point. We'll estimate deduplicated reach, build its frequency distribution and show where the reporting coverage runs out.
Map reach and frequency

No. An awareness push and a retargeting campaign ask different things of the audience, so the same cap may work badly for one of them. We set a range from the funnel stage, creative and campaign objective, then revisit it once the distribution is visible. The result still applies only to that campaign.