Our customer service team is measured closely on how long each call takes. Last winter the average rose by about forty seconds, and over three months twelve agents were placed on coaching plans for handling time. On the slow calls, the recordings were mostly silence while the agent waited.
What they were waiting for was the order history screen. After an upgrade in October it took between eight and fifteen seconds to open for any customer with a long history, which describes most of our trade accounts, and a typical call opens it three or four times. Our monitoring had recorded this. It was one line on a dashboard of application response times, sitting in amber, logged as a known issue with a fix planned for the spring release.
Nobody connected the two. My team saw a slow screen, rated it a moderate problem, and scheduled it. Customer services saw slow agents, rated it a performance problem, and managed it. Each department handled its half competently, and the only people who experienced both halves were the agents, who were being coached for our delay.
I found out from the head of customer services, who asked me in a corridor whether anything had changed in October.
The fix took three weeks once it had a proper priority, and average call length went back to where it had been. The coaching plans were withdrawn, and I apologised to the team in person.
What has changed since is how we decide what a slow system costs. For the screens that customer services, dispatch and finance use most, we now express response time in the measures those departments manage their people by: seconds per call, orders per hour, invoices per day. A delay of ten seconds on a screen used forty thousand times a month is about a hundred and ten hours of staff time, and it is scheduled on that basis instead of on our own rating. The heads of those departments see the same figures in their monthly reports, so a change in their team's numbers can be checked against ours before anybody is coached.
When a system slows down, the cost does not usually arrive at IT. It turns up in somebody else's performance figures, and it tends to get treated as a problem with the people involved.
– Serguey Shinder
Top comments (0)