Did the customer actually get better service?

The metric that flatters everybody

Response time is easy to measure, improves immediately, and is the first number every vendor reports. It is also nearly useless on its own, because a fast wrong answer is worse than a slow right one.

If you measure only speed, you will optimise for speed, and the system will get faster at producing replies that generate a second message from the same customer asking what you meant.

Four numbers that mean something

Resolution on first contact. Of the enquiries that arrived this week, what proportion were finished in one exchange? This is the honest measure of whether an answer answered. It is harder to count than response time and it is the one worth counting.

Repeat contact within 48 hours. The same customer coming back about the same thing. This rises when replies get faster and worse, and it is the clearest early warning that something has degraded. Watch it in the fortnight after any change.

Enquiry to outcome. Of the people who asked, how many booked, ordered or got what they came for? This is the number that connects the front desk to the business, and it is the one that tells you whether a change was worth making at all.

Complaints about the service itself. Not about the product: about being made to talk to a machine, or not being able to reach anybody. A small number, and each one is worth reading in full.

The instrument that is not a survey

Most small-business satisfaction surveys are answered by almost nobody and by an unrepresentative few of those. If you want a signal, keep it to one question, sent by the channel the customer already used, at the moment the thing was resolved:

Did that sort it out? Yes / no

One tap. Response rates are far higher than a five-question form, and the answer is the one you needed. Add an optional free-text box for the noes and read every one.

Before and after

The comparison only works if you measured before, which is the five-day tally applied to outcomes rather than to your own time.

For a fortnight before you change anything, count: enquiries received, how many were resolved in one exchange, how many came back, how many turned into business. Then run the change and count the same four for a fortnight.

If enquiries turn into business at the same rate but each takes you less time, that is a genuine win. If the rate falls while your time falls further, you have made a trade, and it might be the wrong one: a fast reply that loses one booking a week costs more than the hours it saved in most businesses.

Read the actual conversations

Numbers tell you something changed. Twenty transcripts tell you what.

Once a week, read twenty end to end: a mix of resolved, escalated and abandoned. Look for the sentence where it went wrong, the question you cannot answer, and the point where somebody gave up. Owners who do this find things no dashboard reports, and it takes twenty minutes.

The number to stop reporting

Time saved, on its own, with no outcome attached.

“We save six hours a week” is a statement about you, not about your customers, and it is exactly the number that lets a change that quietly costs you sales look like a success. Report it beside the outcome rate, or do not report it.

The question at the end of this block is the same one it opened with. The system got faster. Did anybody get served better?

Counting without a dashboard

None of this needs software. A spreadsheet with one row per enquiry and four columns (date, channel, resolved first time, outcome) takes about fifteen seconds a row and gives you every number in this lesson.

Do it for a fortnight before a change and a fortnight after, then stop. This is a measurement exercise, not a permanent process, and businesses that try to track everything forever end up tracking nothing.

Who reads the numbers

In a business of two or three, the person who made the change is the worst reader of whether it worked, for the same reason they are the worst marker of their own prompt.

Show the four numbers to somebody else with the before-and-after side by side and no commentary. If they reach the same conclusion you did, it is probably in the data. If you find yourself explaining why one of the numbers does not count, that is the number to look at hardest.