Monday, before. A support team reports first response time and volume. Both look good and have improved over the year.
What the numbers do not show is that a proportion of those fast first responses resolve nothing, generating a second exchange, sometimes a third, and a customer who is measured as having been answered quickly four times about one problem.
The team is optimising for the thing being counted, which is exactly what anybody would do, and the actual experience is not improving at all.
Monday before: the report is response time and volume
Two numbers, both easy to collect, both improving, and neither describing whether anybody was helped.
Monday after: resolution is counted alongside them
Whether the issue was closed, and whether it stayed closed, which is the outcome the speed was supposed to serve.
Tuesday before: the fast holding reply scores well
Thanks for your message, we are looking into this, sent in four minutes and resolving nothing, counted as a response.
Tuesday after: first contact resolution is visible
How often a single reply actually ended the matter, which is the number that best describes the quality of the answers.
Wednesday before: the repeat contacts are invisible
Four exchanges about one problem counted as four well handled contacts rather than as one poorly handled one.
Wednesday after: repeat contact is the headline
Which immediately identifies the answers that are not working and the topics where the underlying information is wrong.
Thursday before: the customer view is a quarterly survey
With a low response rate, arriving long after the interaction, answered mostly by people who were angry.
Thursday after: the customer is asked at the end of the conversation
Briefly, immediately, which produces considerably more responses and attaches them to specific interactions.
Friday before: the difficult tickets look bad
Because a complex case takes longer and scores worse, which encourages avoiding them.
Friday after: complexity is accounted for
So the team is not penalised for handling the cases that need the most skill, which is where support is actually valuable.
What to measure in total
Response time, resolution rate, repeat contact rate, exchanges per resolution and a brief customer rating. Five numbers.
What not to do with them
Set individual targets on speed, which produces exactly the behaviour described above and is the commonest failure in support management.
What changes when automation is added
Volume falls and complexity per ticket rises, which makes speed look worse and quality look better, as in why average handle time drops when you add a.
What the team should see
Their own numbers, weekly, rather than a management report they never read. Support people act on information about their own work and ignore information about the department.
What a good rating actually indicates
Frequently that somebody was pleasant rather than that a problem was solved. Ratings are worth collecting and should never be the only measure of whether support is working.
Count resolution rather than speed
An assistant handling the straightforward questions instantly leaves a team whose remaining tickets are genuinely complex, which requires measuring resolution and repeat contact rather than response time, or the reporting will suggest the team got slower while the service improved. The baseline discipline sits in measuring whether automation actually worked, the commitments in support tier and response time questions, and the coverage question in weekend and holiday cover questions. See pricing, run the numbers on the chatbot ROI calculator, and the sector view sits under small business.