Segment Sizing Errors That Quietly Distort Every Forecast

Most segment sizes are wrong in the optimistic direction. Overlapping definitions double count people, stale rules keep the departed, and test audiences leak into reporting. Every forecast built on those numbers inherits the error, usually discovered after budget is spent.

Where phantom members come from

A segment defined by behavior in the last ninety days still contains people whose last qualifying act was on day ninety, long cooled. Another defined by traits contains people who once matched and no longer do. Neither list is lying exactly, both are aging in place while reports treat them as current.

For more on this, see this related guide and our earlier explainer on the same theme.

Overlap is the bigger sin. If three campaign segments each claim the same high value customers, the sum of segment forecasts can exceed the entire customer base. Plans built per segment in isolation rarely get reconciled until results expose the double count painfully.

Count exclusively before you forecast

Build a counting pass that assigns each person to their primary segment for planning purposes, even if they qualify for several for targeting. Report both numbers: qualified for, and counted in. Forecasts use counted in. Targeting may use qualified for, with eyes open about overlap.

Date every count. A segment size without an as of date is a rumor. Sizes drift weekly in active businesses, and a forecast meeting using last quarter's count for this quarter's plan has built error in from the first slide.

Leaks between test and production

Test audiences, staff accounts, seed lists, and partner test orders creep into segments defined loosely. Each is small, together they pollute response rates and inflate reachable counts, especially in small premium segments where a few test records bend percentages hard.

Readers also found this companion piece from our archive useful.

Keep a standing exclusion list applied at segment build time, not at send time. Exclusions applied late get forgotten under deadline pressure. Build time exclusion is automatic, auditable, and boring, which is what data hygiene should be.

A monthly sizing audit

Once a month, reconcile three numbers per key segment: definition count, deduplicated counted in number, and reachable number after consent and contact quality filters. The gaps between them are the story. A segment of fifty thousand with twelve thousand reachable is a twelve thousand segment for planning.

Publish the audit where planners actually look. Buried data quality reports change nothing. A single honest line in the forecast template, reachable audience as of date, forces every plan to start from the truth.

Frequently asked questions

Why do segment counts from different tools disagree?

Because definitions, refresh timing, and identity matching differ per tool. Small gaps are normal. Large gaps mean definitions have drifted, and reconciling them is a definitions meeting, not a tooling argument.

Should overlapping segments be merged?

Not always for targeting, since different campaigns legitimately reach the same person. For forecasting and budgeting, yes, use an exclusive primary assignment so totals cannot exceed real people.

How stale can a behavior based segment get?

It depends on purchase cycle, but any behavior window should be reviewed against actual repeat timing. A ninety day window in an annual purchase category mostly captures people who will not buy again soon.

What is reachable audience?

The people in a segment you can actually contact: valid address or number, consent in place, not suppressed. It is almost always smaller than the definition count, and it is the only honest base for forecasts.