Your B2B incidence rate is probably wrong

A supplier quotes 18% incidence on IT decision-makers at companies with 500 or more staff. Field comes back at 7%. Both numbers were computed honestly from real data, and the gap between them is where most sample disputes actually live.

Incidence rate is a ratio, and a ratio is only as well defined as its denominator. In consumer work the denominator is usually obvious enough that nobody argues. In B2B it is almost never obvious, and the argument is worth having before the study fields rather than after.

The denominator problem

Ask three people what the incidence of "IT decision-makers at 500+ employee firms" is and you can get three defensible answers, because they are dividing by different things.

Denominator	Reads as	Typical result
Everyone in the panel	What share of all members qualify	Very low, and useless for planning — most of the panel was never a candidate
Everyone profiled as working full-time	What share of the employed base qualifies	The number most suppliers quote
Everyone who entered the screener	What share of people we actually invited qualified	The number field produces, and usually much lower
Everyone who entered and completed the screener	Qualification rate excluding drop-off	Higher than the previous row, and the one that flatters a report

Rows two and three are both correct and routinely differ by more than a factor of two, because the people invited into a study are not a random draw from the profiled base. They are the ones who were available, who had not hit a recent participation cap, and who had not already been used on a similar study that month. Every one of those filters correlates with seniority.

When a supplier quotes incidence, the question worth asking is not "how confident are you" but "what did you divide by". A supplier who cannot answer that immediately has not built the report themselves.

Claimed against verified

The second gap is larger and less discussed. Profile data is self-reported, and B2B profile data is self-reported about things people have an incentive to round upward.

"Director" is a title with no fixed meaning across markets — in some sectors it denotes board membership and in others it is three years in. "Involved in purchase decisions" covers everyone from the person who signs to the person who was once asked their opinion. Company size is answered from memory, and people who joined a firm of 400 that has since grown will answer with the number they learned on their first day.

None of this is dishonesty. It is what self-report does, and a feasibility model that treats claimed seniority as verified seniority will miss consistently in the same direction.

Three things reduce the gap, in descending order of how much they help and ascending order of how much they cost:

1. Ask the same thing twice, differently, months apart. Somebody who answered "VP" in March and "individual contributor" in July has an answer history worth reading before they are routed into a $40 study. This costs nothing beyond keeping the history, which is why the history is kept.
2. Ask for the decision, not the title. "Which of these have you personally signed off in the last twelve months" separates authority from job title better than any seniority scale, and it is much harder to answer aspirationally.
3. Verify against something external for the narrowest audiences. Expensive, slow, and the only thing that actually works above a certain price per complete.

Profile decay, which nobody budgets for

A profile answer is a photograph, not a fact. People change jobs. Companies get acquired, and 5,000 staff becomes 40,000 overnight without anyone updating anything. Departments get renamed.

The practical consequence is that the usable incidence of a B2B audience declines with the age of the profile data underneath it, and it declines faster for exactly the segments that are worth the most. Senior people change roles more often than junior ones.

So a feasibility number carries an implicit date. An audience defined on answers collected eighteen months ago is not the same audience today, and the difference is not small. Re-profiling is unglamorous work that nothing on a dashboard rewards, and it is the single highest-return maintenance activity a B2B panel has.

A feasibility estimate, taken apart

Here is the arithmetic for 200 completes among IT decision-makers at 500+ employee firms in the United States, over a fourteen-day field window. Every line is a place the estimate can go wrong.

Step	Factor	Remaining
Profiled members, US, employed full-time	—	Base
Claim IT or engineering function	× 0.11	11.0% of base
Claim 500+ employees	× 0.28	3.1%
Claim purchase authority or strong influence	× 0.41	1.3%
Survives in-survey verification	× 0.62	0.8%
Active in the last 90 days	× 0.55	0.44%
Not capped by recent participation	× 0.80	0.35%
Responds inside a fourteen-day window	× 0.22	0.077%

Four of those eight factors are properties of the panel rather than of the audience, and they are the four a buyer never sees. The step that usually breaks an estimate is the second to last: a supplier models the audience correctly and then assumes everyone in it is reachable in the window, which is the same mistake as assuming everyone who owns a phone will answer it this week.

What we will not quote

We do not quote feasibility on an audience we have not measured, and we say so rather than producing a number with an apology attached. A quote of "we think this is reachable but the profile coverage is thin in two of your four markets" is less satisfying than a number and considerably more useful, because it tells you which part of the plan to change.

We also do not quote a single figure where the honest answer is a range. A point estimate implies a precision that a self-reported B2B audience cannot support, and the buyer who builds a timeline on it is the one who pays for the false precision.