Accessibility Champions Networks: Measuring Whether Internal Advocates Actually Move the Numbers

2026-10-18

Nearly every accessibility program of a certain size starts a champions network: volunteers from product, design, engineering, and content who advocate locally, answer questions, and carry the program into teams it cannot reach directly. The model is sound. The measurement usually is not. Networks report members recruited, sessions held, and messages posted, and leadership is asked to believe the rest.

Belief is not required. A champions network touches real work, and real work leaves traces in defect trackers, review comments, and release data. The measurement task is to follow those traces honestly, including the possibility that the network is a pleasant community and not yet an effectiveness program.

Compare defect rates, not enthusiasm

The core comparison: accessibility defects found after release, per team, split by whether the team has an active champion. Active needs a definition with teeth, such as the champion reviewed work, answered a team question, or ran a check in the period. Membership on a mailing list is not activity, and counting it inflates the network while teaching you nothing.

Give the comparison time and fairness. Champions often volunteer in teams that already care, so their teams may start better. Track the trend within teams before and after a champion becomes active, not only the gap between teams. A falling defect rate in newly championed teams is the signal that survives scrutiny.

Fix time is the second vital sign

Champions shorten the path between a defect being found and being understood. A team with a local advocate does not wait for the central program to triage, explain, and persuade; the explanation is in the room. Median days from report to fix, for accessibility defects, should fall in championed teams even before defect counts do.

Fix time also exposes the network's real constraint. If championed teams understand defects quickly but fix them slowly, the bottleneck is prioritization authority, not knowledge, and the program's next investment is escalation paths rather than more training. Measurement that distinguishes those two failures pays for itself.

Count questions asked before release

The deepest effect of a network is temporal: problems raised in design reviews and office hours instead of audits. Count pre-release consultations per team and track what fraction of championed teams' accessibility issues are first recorded before launch. A rising pre-release share means the network is shifting the work left, which is where defects are cheapest.

This metric needs light instrumentation: a tag or a form, applied consistently. Programs resist instrumenting advocacy because it feels bureaucratic. Keep it to one field in the consultation log. Without it, the network's best work is invisible by design, since a defect prevented leaves no defect record.

Prune and support based on the data

Networks decay predictably: founding enthusiasm fades, champions change teams, and the roster becomes a list of the departed. An annual reconciliation against activity data, who actually reviewed, answered, or consulted in the last two quarters, keeps the roster real and shows where coverage has silently vanished.

Support should follow the same data. Champions in teams with high defect volume need office hours, review backup, and visible wins to cite. Champions in quiet teams may need nothing but recognition. Treating the network as one uniform audience wastes both. The metrics that prove the network works are the same ones that tell you how to keep it working.

Frequently asked questions

How do you measure an accessibility champions network?

By outcomes in champions' teams: accessibility defects shipped per release, median fix time, and the share of issues raised before release, compared over time and against teams without active champions.

What counts as an active champion?

Someone who did the work in the period: reviewed a design or build, answered a team accessibility question, or ran a check. Being on the roster or attending a session is not activity.

How long before results show?

Fix times usually improve first, within a quarter or two. Defect rate trends need two or three quarters of data to separate from noise.

What if the data shows no effect?

Then the network is functioning as a community and communications channel. That has value, but it should be budgeted and improved as that, not credited as a defect reduction program.