How many responses do you need?

The deadline passes, five of your twelve board members have answered, and the report tells you it's being withheld. That is the system working as designed, but it's a lot easier to accept if you understand it before you run into it rather than after.

Two questions are tangled together here. One is statistical: how many answers before a number means anything. The other is about confidentiality: how few answers before a number identifies the people who gave it. In this assessment the second question sets the rules, because it's the one where getting it wrong is irreversible.

The thresholds, exactly

These are published, they apply to everyone, and there's no override for administrators or for us.

Board report: at least 4 responses, and at least 60% of the roster. Both conditions, not either. On a 12-member board that means 8 responses, because 60% of 12 rounds up. On a 6-member board it means 4, since the flat minimum binds before the share does. On a 20-member board it means 12.

The chief executive is excluded from both the count and the roster figure, so the numerator and the denominator describe the same group of people. Their responses appear as their own column and are never averaged into the board, which also means an executive's answer can't substitute for a board member's in clearing the threshold.

Board breakouts by tenure or committee: at least 5 responses within that group. The breakout is withheld on its own; the rest of the report is unaffected.

Community report: at least 15 responses. A flat count, with no share requirement, because there's no roster. You're often asking a mailing list, and there's no meaningful denominator to take a percentage of.

Community breakouts by donor, volunteer, or participant: at least 10 within that group.

Staff report: at least 4 responses, and at least 60% of the roster. Same formula as the board rather than the community's flat count, because a staff team is a small, known, bounded group in the same way a board is. A flat 15 would be unreachable for most organizations, and a flat 4 would be far too loose for a team of six.

Written comments follow the section they belong to. Released with it, withheld with it, shown exactly as written or not at all.

The executive director's column follows the board report. If the board report is withheld, the executive's answers are withheld too, since a column of one alongside a suppressed report just exposes that one person.

Why a count and a share

Either alone breaks in a predictable way.

A count alone is too weak for large groups. Four responses from a 30-person board is not a picture of that board, it's a picture of four people who felt strongly enough to answer, and there's no reason to assume they're typical. The share requirement is what keeps a motivated minority from speaking for everyone.

A share alone is too weak for small groups. Sixty percent of a five-person board is three people, and three responses to a five-point scale can be reconstructed by anyone who knows the group. The flat minimum of four is what stops a very small board from producing a report that reads as three named opinions.

Requiring both means each one covers the other's failure. It also means the remedy differs depending on which is binding, and the withheld notice says which. A small board needs more people to answer at all. A large one needs a bigger share of the room.

Why a section is withheld rather than shown

The obvious alternative is to show the numbers with a warning label. Based on 3 responses, interpret with caution. It reads as more generous, and it's worse in three ways.

First, confidentiality. A mean of three responses, next to a distribution, next to three written comments, is not anonymous in any way the people who answered would recognize. They were told the threshold before they answered question one, and they decided how honest to be based on it. Releasing a thin section anyway would falsify that promise retroactively, for responses already collected, and there's no way to undo it.

Second, caveats don't travel. The label survives about as long as it takes for someone to screenshot the number into a board packet, where it becomes "our financial oversight scored 2.9" with no mention of how many people said so. A withheld section can't be misquoted. A caveated one is designed to be.

Third, it wouldn't be worth much anyway. A mean of three answers on a five-point scale moves by a third of a point when one person shifts one notch. There's nothing there to act on.

So below the threshold, the section is withheld, the report says exactly what's missing and what would release it, and nothing partial renders. If a cycle closes short, no findings are stored at all.

What to do when you are short

Check the roster is accurate first. Rosters go stale, and a member who resigned in March is still inflating your denominator in September. Keeping the roster current is legitimate. Trimming it to clear a threshold isn't, and you'd only be lying to yourself, since a report built on four of eleven real members tells you about four people either way.

Put 15 minutes on a meeting agenda and have everyone answer in the room. Laptops or phones out, chair leaves the questions on screen and stops talking. This is far and away the most effective thing a board chair can do, and it routinely takes a cycle from five responses to eleven in one sitting. It doesn't compromise anonymity, because there's no link between a submission and a person regardless of where someone sits while they answer.

For community, widen the ask rather than pushing harder. Fifteen is a low bar in absolute terms and a real one if you're only emailing major donors. Send it to the whole list, put it in the newsletter, and expect a single-digit response rate. The community questions take about three minutes, which is worth saying in the email.

For staff, watch the share, not the count. With eight staff you need five responses, and the fifth is usually the hardest. The staff questions are short by design for exactly this reason.

If you can't get there, close the cycle and run it again. A short cycle isn't wasted. It tells you something real about engagement, and the second attempt with a meeting-room fifteen minutes almost always clears.

What the numbers do not tell you

Clearing a threshold makes a report releasable. It doesn't make every difference in it meaningful.

With eight responses, a gap of a tenth of a point between two domains is noise. A gap of a full point is worth a conversation. And no single score means anything without cohort context, which is why a domain result should never be described as good or bad on its own. When there aren't enough comparable organizations to say, the honest thing is to say that rather than to imply a verdict.

This is also the strongest practical argument for treating an assessment as an annual habit, discussed further in how often a board should evaluate itself. Your own board a year ago is a better comparison than almost anything else available, and it requires no cohort at all. For outside reference points on board size and composition, BoardSource publishes periodic governance research worth reading alongside your own results.

The thresholds aren't a hurdle the product puts between you and your data. They're the reason your board members answered honestly in the first place. A report you can't see because four people answered is the same rule that made those four people willing to say what they actually thought.