#383 2012 · Zappos · Retaildrop-the-proxy-metric
Zappos stopped timing how fast its call-center agents hung up and started timing how long they stayed on the phone
the problem
A speed metric was training agents to rush customers off calls
background
Call centers are built around average handle time (AHT): the shorter the average call, the more calls an agent can take per shift, and the metric shows up on nearly every industry benchmarking report. Supervisors coach toward it, agents feel it, and callers feel it too — the push to wrap up and get off the line is audible in how a rushed agent talks, whatever the training manual says about listening to the customer.
Zappos had built its brand on the opposite promise: customer service reps empowered to do whatever it took to make a caller happy, including solving problems that had nothing to do with a Zappos order. But as long as the team's own dashboards still ran on AHT, that promise competed against the one number every call-center manager is trained to protect — a company could say 'take your time' and still measure 'go faster' underneath it.
the move
Zappos's Customer Loyalty Team dropped average handle time from its call-center metrics entirely and replaced it with 'personal service level' — simply the percentage of an agent's shift spent actively on customer-facing calls, with an 80% target. How long any single call took stopped being tracked or coached against; agents meeting the 80% threshold overall were rewarded (including spinning a prize wheel), and those below it were coached on engagement, not speed.
the payoff
Freed from a per-call clock, agents routinely spent as long as a conversation needed — the team's longest recorded call ran more than eight hours, in one case an agent helping a caller find a shoe on a competitor's site when Zappos itself was out of stock, then staying on the line. Operations manager Derek Carder described the trade explicitly: the team optimizes for an emotional connection with the caller over getting them off the phone quickly.
what came after
The personal-service-level metric became one of the most cited examples in call-center management writing of a company deliberately discarding an industry-standard efficiency metric because it was optimizing the wrong behavior, and it is still used as a teaching case for the general problem of a proxy metric quietly training against the outcome it was meant to protect.
references
- [1]Call Centre Metrics the Zappos WayThe Squawk Point, 2012squawkpoint.com
- [2]The Zappos Customer Support Key Performance IndicatorsCallFire, 2012callfire.com