The right customer service KPIs: what should a small business measure?

Start with the question, not the number

A small business doesn't need a dashboard with twenty charts. It needs answers to four questions:

1. Are customers helped on time? For that you look at first response time.
2. Is anything being left undone? For that you look at the percentage resolved, the callback rate and open follow-up.
3. Does contact lead to an appointment? For that you look at the booking rate.
4. Can the team handle the questions themselves? For that you look at the escalation rate.

A useful KPI has a fixed definition, a clear numerator and denominator, and leads to a decision. If you can't say what you will do differently when a number goes up or down, you don't need to track it.

KPI 1: count unique requests, not messages

Almost every other KPI divides by the number of unique requests. A unique request is one customer with one question, however many messages or calls go with it. Someone who calls three times and then sends two WhatsApp messages about the same repair is one request, not five.

Decide in advance what you count as one request and stick to it. For example:

• The same customer with the same question within three days is one request, whatever the channel.
• A new question from an existing customer, such as a second appointment, is a new request.
• A missed call from a number that never gets back in touch counts as an unanswered request.
• Advertising, spam and wrong numbers don't count.

KPIs 2 to 7: the formulas

For each metric, here is what goes into it: the numerator (above the line) and the denominator (below it). You get the numbers from your contact list or appointments list.

2. % resolved

3. First response time and % within target time

First response time is the time between a request coming in and the first real answer from someone on your team. An automatic away message doesn't count. Decide together whether the clock keeps running outside business hours, or whether you look at evening messages separately. Choose one way and write it down.

4. Callback rate

5. Booking rate

A low booking rate doesn't tell you why. Look at the notes: was there no slot available, was the price the problem, or did your response come too late?

6. Open follow-up

7. Escalation rate

A high or low escalation rate is not good or bad in itself. What matters is the reason: if the same question keeps coming back, a standard answer or a clear rule can make sure the team resolves it themselves.

Average or median? A simple explanation

The average adds up all the response times and divides by how many there are. The median puts all the times in order and takes the middle one. If there is one outlier among them, the two can be far apart.

Suppose seven customers got a first response after 5, 8, 10, 12, 15, 20 and 600 minutes (fictional). The last one sent a message on Saturday evening that was not seen until Monday.

• Average: 670 ÷ 7 = about 96 minutes.
• Median: the middle one of the seven in order, so 12 minutes.

The average suggests that customers wait an hour and a half, while six of the seven had an answer within twenty minutes. The median shows how it went for a typical customer. You see the outlier in the % within target time. So look at the two together.

How to set your own target

There is no percentage that is “good” for every business. A garage doing emergency repairs has different customers than a salon that books weeks ahead. So set your own target.

1. Measure first without changing anything

   Track for two to four weeks how things go now. That is your baseline. The customer contact audit from lesson 1 helps with this.

2. Choose one metric

   Start with the metric where you lose the most, for example the callback rate if a lot of calls are missed.

3. See what is achievable

   Take your staffing and business hours into account. A short target time makes no sense if nobody is free to respond at busy times.

4. Think from the customer's side

   Someone whose car broke down on the side of the road wants an answer sooner than someone asking for a quote. You can agree on a different target time for each type of question.

5. Aim just above your baseline

   Example: if 6 out of 10 customers got a response on time in your baseline, aim for 7 out of 10 first. After four weeks, you adjust.

6. Write down the definitions

   Record what a unique request is, what resolved means and what your target time is. Don't change that halfway through, or you will be comparing apples to oranges.

Pitfalls that make your numbers unreliable

• Counting messages instead of people. More messages can mean that customers had to repeat themselves because they got no answer. Then it looks busier, while things are actually getting worse.
• Vanity metrics. The number of followers, chats or messages sent looks good, but doesn't say whether customers were helped. If a number reassures you without you doing anything with it, leave it out.
• Comparing weeks that include holidays or vacation periods. A week with a day off or during school vacation is not a normal week. Note such exceptions and look at the trend over four weeks instead.
• Percentages without numbers. With eight appointment requests, one customer is already 12.5 percentage points. Always put the number next to the percentage and don't draw conclusions from one week with small numbers.
• Attribution. The channel through which someone gets in touch is not always the channel through which they found you: someone sees your post on Facebook and then calls. If you want to know, ask (“How did you find us?”), note it the same way every time and treat the result as a pointer, not as proof.
• Only counting what was answered. If you only measure answered requests, you miss exactly the customers who never got an answer.

The fifteen-minute weekly meeting

Numbers don't improve anything if nobody looks at them. Set aside the same fifteen minutes every week with everyone who handles customer contact, for example on Monday morning before opening. One person fills out the scorecard beforehand, so the meeting isn't spent counting.

Keep it to one improvement per week. If you try three things at once, next week you won't know what helped. And talk about how you work, not about people: the question is why something was left undone, not who left it.

Step-by-step example: two weeks at a garage

Checklist and next step

• Written down what counts as one unique request for you
• Recorded the numerator and denominator for each KPI
• Chosen your own target time, possibly per type of question
• Measured for two to four weeks before setting a target
• Put the number next to every percentage
• Noted unusual weeks (holiday, vacation, illness)
• Scheduled the same fifteen minutes each week, with one person who fills out the scorecard
• One improvement each week, with a name and a date

With the weekly scorecard you don't have to do the math yourself: you enter the numbers for each week and the percentages are calculated for you. In lesson 5 of the free course you will read how to use these numbers to choose an improvement every week and stick with it.

Does the work itself take too much time? Bits Please can set up Claudia to take over suitable, recurring tasks, such as answering phone calls and WhatsApp questions and recording requests, depending on the package. The KPIs in this article work without it too.

Frequently asked questions