You are standing in the back of a poster session at your annual broader impacts showcase. Forty posters line the walls. Students are presenting to a mix of faculty, community partners, and program administrators. The energy is good. You can see real conversations happening: a community college instructor leaning in to study a student's data, a nonprofit director pulling out her phone to photograph a methods section, two graduate students from different institutions exchanging contact information.

Then you look at your evaluation plan. You have a rubric for three external judges who are scoring each poster on a five-point scale. You have a post-event survey queued up in Qualtrics. And you have a head count at the door.

None of that captures what you are watching right now.

The rubric tells you whether posters met technical standards. The survey will tell you whether attendees were satisfied. The head count tells you people showed up. But the community college instructor who just realized her students could partner with a nearby research university? The connections forming between two programs that did not know each other existed? The moment a first-generation student heard a faculty member describe her data analysis as "exactly the kind of work we need"? None of that will appear in your annual report.

You are generating impact you cannot document. And when your funder asks what happened, you will have numbers that tell the easy story and miss the important one.

The poster session problem is actually a design problem, not a measurement problem. Most poster sessions are designed for one purpose (presentation) and then evaluated with instruments designed for a different purpose (satisfaction measurement). The activity and the evaluation live in separate worlds, connected only by the assumption that if people present and people attend, something good probably happened.

This separation creates three specific failures.

First, timing failure. Surveys sent after an event depend on memory and willingness to respond. What people remember the next day is not what they experienced in the moment. The response card a participant writes while standing in front of a poster captures real-time resonance. A survey response three days later captures reconstructed impressions.

Second, data type failure. Rubric scores and satisfaction ratings are thin data. They tell you whether something met a standard but not what actually happened. The richest evaluation evidence (what resonated, what surprised, what connections formed) lives in the conversations that poster sessions generate but nobody captures.

Third, resilience failure. When logistics change (and they always do), bolt-on evaluation breaks. If the post-event survey window disappears because of a scheduling conflict, you lose your data. But if the evaluation is embedded in the activity itself, it survives logistical disruption because the activity still happens.

The deeper issue is philosophical. When evaluation is separate from the activity, it communicates that the activity serves participants and the evaluation serves the funder. Embedded evaluation dissolves that distinction. The evaluation activities serve both purposes simultaneously.