Customer support KPIs for a small team

Five measures that cover most of it

1. First reply time: how long a customer waits for the first human answer, during the hours you cover.
2. Resolution time: from the first message to the moment the issue is solved.
3. Customer satisfaction (CSAT): the share of customers who rate a solved conversation as good.
4. Reopen rate: how often a solved ticket comes back.
5. Contacts per order: support conversations divided by orders in the same period; for a store, it shows whether problems are being fixed at the source.

Help desks such as Zendesk, Gorgias, Help Scout or Freshdesk report the first four in their standard reports. The fifth needs the order count from your store.

Define them before you measure

The same number can mean very different things. Write down, once:

• Whether times count business hours only or the whole clock. A message sent at night and answered in the morning looks slow in clock hours and fine in business hours.
• Whether you report the median or the average. One stuck ticket can move an average a lot; the median shows the usual experience.
• Which channels are included: email, chat, social messages, marketplace messages.
• When a ticket counts as solved, and whether automatic replies count as a first reply. They should not.

Keep the definitions in your support handbook, so the numbers stay comparable when the team or the tool changes.

Set targets from your own baseline

Published benchmarks rarely fit a small business: your channels, products and customers are your own. A better way:

1. Measure the five numbers for a month without targets.
2. Pick the one or two that matter most to your customers right now, for example first reply time in a busy season.
3. Set a target a little better than the baseline, for the hours your agent covers.
4. Review monthly, and change targets when the business changes, not every week.

Share the targets with the agent and explain why each one matters. Targets the agent never sees do not change anything.

What the agent controls, and what they do not

Some numbers depend on the person and some on you. Read them that way:

• Mostly the agent: tone, accuracy, following the policy, first reply time during their shift, tagging.
• Partly the agent: resolution time and reopen rate, which also depend on warehouses, carriers and your policies.
• Mostly the business: contacts per order, which falls when product pages, delivery estimates and packaging improve.
• Shared: satisfaction, which reflects both the answer and the product.

A good agent reports causes, not just numbers: the five most common reasons customers wrote this month and what would remove them. Ask for that in the weekly update.

Review conversations, not just numbers

Numbers show speed; only conversations show whether customers were treated well. Each month, read a random sample of solved conversations together with the agent and score them on the same sheet:

Agree the scores together. The aim is a shared standard rather than a grade: the agent should leave the review knowing exactly what good looks like.

Measures to use with care

• Tickets per hour: rewards quick, shallow answers and picking the easy tickets.
• Handle time per ticket on its own: complex problems take longer, and they are the ones that matter most.
• Satisfaction from a handful of ratings: a few votes swing it; read the comments instead.
• A single number tied to a bonus: people optimise what they are paid for, including the parts you did not intend.

Use volume figures to plan staffing, not to judge one person. Our guide to covering US hours shows how volume by hour shapes the shifts, and our comparison of outsourcing and hiring applies the same measures to both models.