Throughput vs. understanding

Career offices at most colleges are designed around volume. The internal dashboards measure first-destination placement, employer events held, advising sessions delivered, percentage of seniors with offers by graduation. These are reasonable things to measure. They are also the wrong things to measure if the goal is development.

A 95% placement rate can describe a class of graduates who landed in jobs they don't understand themselves to want. The throughput is real; the understanding is absent. The metric system can't see the difference.

The tools the office uses reflect throughput design:

• Assessments that produce a one-time, packaged result.
• Advising appointments capped at 30 minutes, structured around getting the student to a next action.
• Job boards optimized for the first-three-result attention attractor.
• Annual surveys that measure placement, not change.

None of these tools is badly built. They are well-designed for what they're for. The question the lab keeps returning to is what happens when you design a parallel set of tools for the work throughput tools were never meant to do, the slow work of figuring out what you actually want, and revising that as your experience accumulates.

This is not an argument against career offices. It is an argument about which problem each design is solving. The two problems coexist. A student needs both a job-application infrastructure and a developmental infrastructure. The lab's projects exist because the second has been historically under-built.

Open thread. The hardest part of running anything in this mode inside an institution is that the work doesn't show up on the same dashboards. Throughput is easy to count. Development isn't. The lab has started keeping its own internal records, mostly instructor field notes that accumulate across cycles. Tracking the same students' artifacts across multiple terms would require a consent and review process the lab has not yet operationalized. It may not amount to a real measurement system. But the alternative, no internal record at all, is worse, so the records keep getting written.