Opening a customer experience floor
An engineering company deciding to answer phones sounds like a distraction. It turned out to be the fastest way to find out what our software was doing wrong.
In 2024 we did something that looked, from the outside, like a strategic mistake: an engineering company opened a customer support floor. The reasoning was simpler than it appeared.
We kept watching clients ship good software into an operation that could not support it. The product team measured releases, the support vendor measured handling time, and nobody measured whether a release had created work. Contact volume went up after every launch and everyone treated that as normal.
Running both sides ourselves closes that loop. Every contact is tagged by root cause, and the top five drivers go to the product backlog each month. On one telecom account that removed a fifth of the total volume in two quarters — not by handling calls faster, but by removing the reasons for them.
Operationally it is a different discipline from engineering and we had to learn it properly. Service levels are a forecasting and rostering problem before they are an attitude problem. We staff to an Erlang forecast with intraday re-forecasting; the previous vendor had staffed to a monthly average, which is why Mondays collapsed.
The part we underestimated was hiring. A support agent who knows the product is worth several who do not, and that only works if agents stay on one account. Shared pools are cheaper per hour and worse at everything else, so we do not run them.
Two years on it is our second-largest practice by headcount. The lesson generalises: if you build software for an operation, being inside that operation tells you things no analytics dashboard will.