A School of Continuous Improvement research poster examines what happens after a caller fails a spoken security question, coding a season of a call centre's own outcome tags into a typology of agent workarounds behind an unmoving verification pass-rate KPI.
A season of a call centre's own outcome tags shows what actually happens once a caller fumbles the question
A metric that never seems to fail is either working, or it has quietly stopped measuring the thing it was built for. That is the question the School of Continuous Improvement put to a telecommunications call centre’s spoken security-question step — the brief exchange every caller clears, or doesn’t, before an account can change hands — after noticing the centre’s own verification pass-rate KPI had not moved in a season.
Dr Dagny Okafor and Associate Professor Casimir Beng pulled 214 flagged episodes from a single contact centre’s Mar–Jun 2026 quarter and open-coded what agents did once a caller’s spoken answer failed to clearly match the record. The pass-rate KPI itself told a tidy story throughout: 97.6 per cent logged as passed. The coding told a messier one.
“Waved through” isn’t a category anyone trains an agent into. It’s just where a caller ends up once the clock and the checklist stop agreeing with each other.
— Dr Dagny Okafor, Lecturer and Convenor, Evaluation of Evaluation, School of Continuous Improvement
Four outcomes recur once a spoken answer doesn’t clearly match: an agent waves the caller through anyway, tries the question again in different words, escalates to a supervisor, or switches quietly to a fallback question. Escalation, the only one of the four that leaves an independent trace in the centre’s own systems, was also the rarest, logged in barely one flagged call in eight.
“An agent watching their own handle-time target doesn’t get to treat a fumbled answer as a research question,” said Associate Professor Casimir Beng, Lead of the Adaptive Metrics Lab. “They get a few seconds to pick one of these outcomes, and the KPI can’t see which one they picked.”
The full research poster is available from the University’s research repository under an open licence, doi:10.5555/slop.ptj0ny.