Shopxare TeamFor an online store with a small support team, seven customer service KPIs are enough: first response time, time to close, first contact resolution, CSAT, contacts per 100 orders, missed reply targets and, if you use AI, the AI resolution rate together with the wait after a handover. Measure them as medians in business hours, look at them once a week on one page, and set targets against your own last quarter. The reference points with a real source: 55 seconds average chat first response in retail (LiveChat, data 2024), a median CSAT of 80% (Survicate, 2025), 69% first contact resolution (SQM Group, 2024) and 20 to 46 support tickets per 100 orders depending on the vertical (Gorgias Ecom Lab, 2026).
Most KPI lists you'll find were written for call centers with 50 agents: 15 to 22 metrics, half of them about phone queues. A store with two or three people answering chat and email needs fewer numbers and a clear idea of what to do when one of them moves.
The 7 KPIs at a glance
| KPI | How to calculate it | Reference point (source, year) | Starting target for a small store |
|---|---|---|---|
| First response time | Median of (first human reply − first customer message), business hours | Chat: 35 s overall, 55 s in retail (LiveChat, 2024) | Chat under 1 min, email under 1 business hour |
| Time to close | Median of (closed − opened), business hours | After an AI handover: chat 8 h, email 32 h, contact form 36 h (Gorgias, 2026) | Chat same day, email within 1 business day |
| First contact resolution | Conversations with no follow-up on the same issue within 7 days ÷ all conversations | 69% average, 80%+ "world class" in call centers (SQM Group, 2024) | 80% |
| CSAT | Ratings of 4–5 ÷ all ratings × 100 | Median 80% (Survicate, 2025) | Hold around 80%, read every 1–3 |
| Contacts per 100 orders | Conversations ÷ orders × 100 | 46 in electronics, 20 in food and beverage (Gorgias, 2026) | Lower than last quarter |
| Missed reply targets | Conversations that broke the first-reply target ÷ all conversations | No reliable public benchmark | Under 5% |
| AI resolution rate and handover wait | Conversations the AI closed without a person ÷ conversations the AI took; median wait from handover to human reply | 45% median, 65% top quartile; 10 h median handover wait (Gorgias, 2026) | Rate going up, handover wait under 1 business hour |
The targets in the last column are our opinion for a team of one to ten people, not industry standards. Change them once you have three months of your own data.
1. First response time
The time between a customer's first message and the first reply from a person. It's the number customers feel most in live chat, because they're sitting in front of the window waiting.
Measure the median in business hours, and leave out automatic replies. If an AI agent answers instantly, report its conversations separately, otherwise your team's speed disappears behind the bot's. Look at the slowest 10% too: a median of 40 seconds can hide a lunch hour where chats wait 20 minutes.
We covered the calculation, the pitfalls and seven ways to bring it down in First response time: benchmarks and how to measure it.
2. Time to close
How long it takes from the first message until the conversation is closed because the problem is solved. First response time tells you how fast you pick up; time to close tells you how long the customer's problem stays open.
Gorgias's 2026 data is useful here because it splits by channel for conversations an AI handed to a person: chat closes in a median of 8 hours, email in 32 and the contact form in 36. The same data shows 13% of handed-over chats abandoned, against 30% for email and 42% for the contact form. If you only fix one thing, move "where is my order" and "can I return this" off the contact form and into chat.
Watch out: teams can lower this number by closing conversations too early. That's why it always sits next to first contact resolution.
3. First contact resolution (FCR)
The share of issues solved in one conversation, without the customer having to come back. Call centers measure it with a survey question ("was your issue solved?"). In chat and email, a simpler rule works:
FCR = conversations where the same customer didn't write again about the same issue within 7 days ÷ all closed conversations × 100
SQM Group's 2024 benchmark found an average of 69% across industries, with retail among the leaders at 73–75% and 80% or more counted as world class. Those are phone figures based on customer surveys, so use them as a rough frame, not as your bar.
What lowers FCR in an online store is almost always the same: a reply that was quick but incomplete ("I'll check and get back to you") or a promise nobody followed up. Better saved replies with the full answer, including the tracking link and the date, fix more than speed does.
4. CSAT
The share of rated conversations where the customer picked 4 or 5 out of 5. Survicate's 2025 data puts the median at 80%, with the middle half of surveys between 64.8% and 90%.
The catch for small teams is sample size. With 50 ratings a month, a drop from 80% to 72% can be pure chance. Under about 30 ratings a month, skip the percentage and read the comments. All the details are in CSAT score: how to calculate it and what a good score is.
5. Contacts per 100 orders
This is the KPI most lists leave out, and the one that tells an online store the most. It measures how many conversations each 100 orders create:
Contacts per 100 orders = conversations in the period ÷ orders in the period × 100
It's the only number here that measures the store rather than the support team. When it rises, something on the site, in the shipping emails or in the product is creating questions. In Gorgias's 2026 data, electronics brands get about 46 tickets per 100 orders and food and beverage brands about 20, so compare yourself with your own past, not with another category.
Break it down by reason once a month. If 40% of your conversations are "where is my order", the fix is a tracking link in the shipping confirmation, not a faster team.
6. Missed reply targets
Set a promise, for example a first reply in chat within 2 minutes and in email within 4 business hours, and count how often you break it. It's the KPI that shows gaps the median hides, and it's easy to act on: you can see the hours and days when targets are missed and move a break or a shift.
There's no reliable public benchmark for this one. Under 5% is a reasonable start for a small team.
7. AI resolution rate and handover wait
If an AI agent answers part of your chats, two numbers matter, and only together:
- AI resolution rate: conversations the AI closed without a person ÷ conversations it took. Gorgias reports a median of 45% of AI-touched tickets resolved end to end and 65% in the top quartile (2026).
- Handover wait: the time between the AI handing over and a person replying. In the same data the median is 10 hours and the 90th percentile 71 hours.
A high resolution rate with a long handover wait is a bad trade: the customers with the hardest questions wait the longest. We'd watch the handover wait first. Our article on what AI can and can't do in customer service covers which questions to give an AI and which to keep.
KPIs you can skip with a small team
- Average handle time. Useful when you pay agents per hour on a phone queue. In chat, it punishes people for taking the time to solve the problem.
- Occupancy and utilisation. They need shift planning and many agents to mean anything.
- Agent rankings. With a few dozen conversations per person a week, you're ranking luck. Use the numbers for coaching, not bonuses.
- NPS every month. It measures the brand, not the conversation. Twice a year by email is plenty.
- Anything you won't act on. If nobody would change anything when a number moves, stop reporting it.
A weekly scorecard you can copy
Paste this into a doc or spreadsheet and fill it in every Monday. Five minutes, one page.
Week: ____ Orders: ____ Conversations: ____
KPI This week Last week Target Note
First response, chat (med) ____ ____ < 1 min
First response, email (med) ____ ____ < 1 h
Time to close (median) ____ ____ same day
First contact resolution ____ % ____ % 80 %
CSAT (ratings: ___) ____ % ____ % ~80 %
Contacts per 100 orders ____ ____ lower
Missed reply targets ____ % ____ % < 5 %
AI resolved / handover wait ____ % / __ ____ rising / < 1 h
Top 3 reasons for contact: 1. ______ 2. ______ 3. ______
One thing we change this week: ____________________
The last line matters most. A scorecard without a decision is a report nobody reads.
Example: a two-person tea shop
A made-up but typical case. An online tea shop with two people sold 900 orders in September and had 270 conversations: 30 contacts per 100 orders. Median chat first response was 45 seconds, CSAT 82% from 41 ratings, and 31 customers wrote again about the same issue within a week, so FCR = (270 − 31) ÷ 270 = 88.5%.
Everything looked fine except the contact rate. Sorting the conversations by reason showed 108 of the 270 (40%) were "where is my order", mostly two days after the shipping email. The email had no tracking link, only "your order has shipped".
They added the tracking link and an expected delivery date to the shipping email. In October, with 950 orders, they had 200 conversations: 21 per 100 orders. Nobody answered faster. Seventy fewer people had to ask.
How to track these in Shopxare
- Reports → Conversations shows the number of conversations, median first response time, median time to close and CSAT, each compared with the previous period, plus busiest hours, missed reply targets and a table per teammate. You can filter by channel, team or teammate. Conversations handled only by a flow aren't counted. See Reports.
- Settings → Teams & assignment is where you set office hours and Reply targets (SLA) for the first and later replies, so first response and missed targets are measured in business hours. See Teams and assignment.
- Contacts per 100 orders isn't calculated for you: take the conversation count from Reports and the order count from your shop admin.
- Are AI hands a conversation to your team when the customer asks for a person, the topic is sensitive, the customer is angry or it isn't confident. Are AI → Overview lists the questions it couldn't answer, which is your to-do list for raising the resolution rate.
- Turn on the rating after a conversation closes in the Messenger settings to collect CSAT (Customize the Messenger).
Shopxare is free, with no credit card. Install the Messenger, set your office hours and reply targets, and after a week Reports → Conversations fills most of the scorecard above.
FAQ
What are the most important KPIs in customer service?
For a small online store: first response time, time to close, first contact resolution, CSAT and contacts per 100 orders. Add missed reply targets once you've set reply promises, and the AI resolution rate with handover wait if an AI agent answers part of your chats.
What is a good first contact resolution rate?
SQM Group's 2024 benchmark puts the cross-industry average at 69%, with 80% or more considered world class. Those are call center figures; for chat and email in an online store, 80% is a reasonable target.
How do you calculate contact rate in e-commerce?
Divide the number of support conversations by the number of orders in the same period and multiply by 100. 270 conversations and 900 orders is 30 contacts per 100 orders.
What is the difference between a KPI and a metric?
A metric is any number you can measure. A KPI is a metric you've picked as a key goal, with a target and someone who acts when it moves. Average handle time is a metric for most small stores, not a KPI.
How often should you review customer service KPIs?
Weekly for speed (first response, missed targets), monthly for quality (CSAT, first contact resolution, contacts per 100 orders), because those need more conversations before a change means anything.
