Data & Program Evaluation
Your math is fine. The harm was never in the math.
Data and program evaluation for museums and cultural institutions, built on Unintended: A Data Practice — our own framework. Six critical points, eighteen questions, and one sheet you can hand a funder, a board, or the community you have been counting.
Recognize it · Evaluate it · Respond — or best of all, avoid it
Why you need to read this
Evaluation is the part everyone postpones.
And it is the only part a funder, a board, and a community will all ask you about. Done late it is a report. Done early, it is a decision.
“How do you know?”
Your funder asks. Today the honest answer is an exit survey of the people who stayed to the end.
The number is already deciding
Attendance totals quietly defund the free day. To a spreadsheet, depth and volume look identical.
It opens, then you find out
The gallery lands as something done about a community, not with it. Eight weeks earlier, that was a meeting.
In your name
A number gathered for one reason gets used for another — to justify what you would never have chosen.
What the work catches
One finding. Two framings.
Nothing about the data changed — only what it was compared against, who it hands the work to, and who it treats as the problem.
What you’d have published
“Visitors from 60624 are 62% less likely to convert to membership. Recommend targeted outreach to raise conversion in low-performing ZIP codes.”
The comparison group is never named. The fix is aimed at residents. The community becomes a performance problem — and your board approves it.
What you publish instead
“Our membership offer converts 62% less often in 60624 than in the ZIPs it was designed around. Recommend testing what the offer assumes about price, hours, and travel time.”
The comparison is named. The burden moves back to the institution. The next step is a test you can run, not a campaign you must fund.
Same number. Opposite decision.
The method — Unintended: A Data Practice
Six critical points.
Six patterns to recognize in how data gets framed, before any of them reaches a decision. None require anyone to act in bad faith — that is the entire difficulty.
Data Responsibility
Who decided what this is meant to change?
Implied Comparison
Less likely … than whom?
Data Burden
Who pays the cost of producing this?
Context Matters
What would make this mean the opposite?
Saviorism
Did they express this need, or did we?
Weaponization
What will this be used to justify?
What comes back to you
Not a score. Four things you can use.
A sheet you can hand a funder
Eighteen questions answered in writing, each with the sentence that earned it. This is what IMLS, NEH and accreditation reviewers mean by “how do you know.”
The problem found before it opens
A harm map: where you are exposed, and what it costs to close while closing it is cheap.
Language you can actually use
The exact sentence causing the exposure — and the sentence to use instead.
A team that doesn’t need us again
Your staff leave running the method themselves. A capability, not a retainer.
Remember
Factual does not mean right.
A number can be accurate, current and fully documented — and still be doing something nobody intended, to people who never agreed to it. Accuracy is not harmlessness, and only one of the two is anybody’s job to check.
Contact us for equitable, culturally aware evaluation.
Bring us an exhibit in development, an evaluation design, a grant report, or a master plan — anything still early enough to change.