Four numbers that tell you whether your IVR is working

Most voice dashboards lead with containment: the share of calls that ended without
reaching a human. It is the easiest number to improve and the easiest to
misinterpret, because a caller who gives up is contained too.

Containment, paired with callback rate. Count how many contained callers call
back within 24 hours. Containment rising while callbacks rise means you are not
resolving calls, you are ending them. This single pairing catches the most common
way voice deployments fool themselves.

Answer rate during your worst hour, not your average one. The value of
automation is that it does not have a queue. Averages spread that across quiet
afternoons and hide the thing you bought it for. Measure the busiest hour of your
busiest day: the Monday after a holiday, the morning after an outage, the day a
campaign lands. If the automated line holds there, it works.

Time to resolution, from the caller’s first attempt. Not handling time, which
improves every time a call is transferred away sooner. Measure from the caller’s
first call to their problem being resolved, across however many calls and
transfers that took. It is harder to compute and it is the only one that matches
what the caller experienced.

Handover quality, not handover rate. Some calls should reach a person, and
driving that number to zero is a bad goal. The useful question is what the agent
arrives with. Does the transfer carry the transcript, the caller’s identity, and
what was already attempted — or does the caller repeat everything from the top? A
handover that starts with “can I have your number again” cancels out whatever the
automation saved.

Two habits make all four honest. Segment by call reason, because an agent that is
excellent at order status and poor at complaints averages out to mediocre and the
average tells you nothing about what to fix. And read transcripts — thirty a week,
chosen at random, including the ones that went well. Every team that does this finds
something the dashboard could not have shown them.

Set a baseline before you automate anything. Without it you will be comparing this
quarter’s numbers to a memory, and memory is generous.