A small online store needs about seven ecommerce customer support metrics: first response time, resolution time, the age of your oldest open conversation, CSAT, contact reasons, contacts per order, and how many pre-sale chats turn into orders. If an AI agent answers some chats, add its handoff rate. That list fits on one screen and takes twenty minutes a week to review. Everything else can wait until you have a bigger team.
This guide explains what each number tells you, how to measure it without a data team, what a sensible target looks like for a team of one to ten people, and which popular metrics you can safely skip.
Why metrics matter even for a two-person team
When two people answer every message, it feels like you already know how support is going. You usually know how it felt on the last busy day. Numbers catch the slow drifts that memory misses.
Take a hypothetical store that switches carriers in March. Nobody notices anything wrong, but order status questions rise from 8 to 21 per 100 orders over four weeks, and replies slow down because the inbox is fuller. A weekly look at two numbers (contact reasons and first response time) shows the problem in week two. Without them, the store finds out from a wave of one-star reviews.
Metrics also settle arguments. "We need a second person" or "we should turn on chat at weekends" are much easier decisions with three months of response times in front of you.
The ecommerce customer support metrics at a glance
| Metric | The question it answers | How to calculate it | Review |
|---|---|---|---|
| First response time | How long do customers wait for a person? | Median time from a customer's first message to your first human reply | Weekly |
| Resolution time | How long until the problem is solved? | Median time from first message to closed | Weekly |
| Oldest open conversation | Is anything falling through the cracks? | Age of the oldest conversation still waiting on you | Daily |
| CSAT | Did customers feel helped? | Good ratings ÷ all ratings × 100 | Weekly, read every comment |
| Contact reasons | What is causing the work? | Share of conversations per tag (order status, returns, pre-sale...) | Weekly |
| Contacts per 100 orders | Is the store creating more questions than it should? | Conversations ÷ orders × 100, same period | Monthly |
| Chat to sale rate | Does support help sell? | Pre-sale chats followed by an order ÷ all pre-sale chats | Monthly |
| AI handoff rate | How often does the bot pass to a person? | AI conversations handed to a human ÷ all AI conversations | Weekly while you tune it |
The free support KPI calculator works out most of these from numbers you can export from any help desk or inbox.
First response time: the first one to fix
First response time is how long a customer waits before a person replies. For an online store it matters more than almost anything else, because a shopper with a question about sizing or delivery is deciding whether to buy right now. In chat, a slow first reply usually means the visitor has already left.

Three rules make the number honest:
- Use the median, not the average. One email that sat over a long weekend can double an average and hide that most replies were quick.
- Do not count auto-replies. "Thanks, we got your message" is not a response, and counting it makes every team look instant.
- Split it by channel. Chat and email have different expectations, and a blended number hides a slow channel behind a fast one.
For targets, Shopify's guide to customer service metrics suggests replying to live chat in under a minute and to email within one business day. For context, Gorgias's benchmark data, drawn from ecommerce brands of around $10 million in sales, puts the median first response time at 6.3 hours, so a small team that answers email the same day is already ahead of many larger stores.
If your number is too high, the fix is rarely hiring first. Check that notifications reach the person on duty, publish your chat hours so nobody waits in an empty chat at 11 pm, and save replies for your five most common questions. The first response time SLA estimator shows what target your current staffing can actually hold.
Resolution time and backlog age
Resolution time runs from the first message until the problem is solved. For a store, many conversations close in one reply (a delivery estimate, a size recommendation). Others wait on a carrier, a warehouse or a refund, and those drag the number up.
Track it as a median and split it by contact reason. A median of four hours for order status questions and three days for returns is normal. A sudden jump in one reason tells you which process broke.
The number most small teams should check daily is simpler: the age of the oldest conversation waiting on you. A customer whose email has been sitting for six days is a chargeback or a bad review in the making, and averages will never surface them. Sort your inbox by oldest first each morning and clear anything over one business day.
Keep conversations that are waiting on the customer out of the count. If you asked for a photo of the damaged item, the wait is on their side.
CSAT: how to ask and how to read it
Customer satisfaction (CSAT) asks one question after a conversation closes, usually "How would you rate the help you received?" on a scale of 1 to 5 or with a thumbs up or down. Your score is the share of good ratings (4 and 5, or thumbs up) out of all ratings.
How you ask changes what you get:
- Ask right after resolution, in the same channel. A survey emailed two days later gets fewer and grumpier answers.
- Ask one question, with an optional comment box. Every extra question cuts responses.
- Ask after every closed conversation, not only the ones that went well. Hand-picked surveys produce a score that means nothing.
Read the score with its sample size in mind. With 20 ratings in a week, a single unhappy customer moves the score by five points. Look at the four-week trend and read every comment, especially the bad ones. A low rating on a returns conversation is often about the returns policy itself, which is worth knowing too.
On targets, the same Shopify guide puts very small businesses above 95% and larger ones around 80%. Gorgias's ecommerce data shows averages between 4.45 and 4.66 out of 5 depending on the category. If you are a small team well below 90%, read the comments before changing anything else. The CSAT calculator converts your raw ratings and shows how much a few responses can swing the result.
Contact reasons: spotting product and shipping problems
Contact reasons are the most useful number a store can track and the one most teams skip. Every conversation gets one tag for why the customer got in touch. Keep the list short, eight tags at most:

- Order status (where is my order)
- Returns and exchanges
- Damaged or wrong item
- Pre-sale question (sizing, stock, materials, delivery time)
- Discount or payment problem
- Account or login
- Other
Then turn the counts into contacts per 100 orders, so growth does not look like a problem. Gorgias's benchmark data ranges from 19 tickets per 100 orders in toys and games to 46 in electronics, which gives you a rough sense of scale. Your own trend matters more than anyone's average.
The value is in what the tags point to. If sizing questions cluster on one product, its size chart is wrong. If order status questions climb after a carrier change, the tracking emails are not doing their job. Each fix removes contacts permanently, which no amount of faster replying can do.
Chat to sale: connecting support to revenue
Support in a store is also a sales channel. A shopper who asks whether a jacket runs small and gets a quick, useful answer often buys. Measuring that keeps support from being treated as a pure cost.
You do not need attribution software for a rough number. Tag pre-sale chats, then once a month check how many of those customers placed an order within seven days, using the email address they gave in the chat. Divide orders by pre-sale chats and you have a chat to sale rate. If you hand out discount codes in chat, give support its own code and the revenue counts itself.
The number is approximate and will undercount shoppers who never left an email. That is acceptable for a trend line, and the trend is what shows whether quick pre-sale answers pay for themselves.
AI deflection and handoff rates
If an AI agent answers some of your chats, two numbers keep it honest. The deflection (or resolution) rate is the share of AI conversations that ended without a person. The handoff rate is the share it passed to your team.
Neither should be pushed to an extreme. A very high deflection rate can mean the bot is good, or that customers gave up. A handoff rate near zero is a warning sign when the bot makes it hard to reach a human. Check three things each week while you tune it:
- Read ten AI conversations end to end, including at least two that were not handed off.
- Compare CSAT on AI conversations with CSAT on human ones.
- List the questions it handed off and add the missing answers to the pages it learns from.
Spend matters too. Many tools bill AI per resolution or per conversation, so a rising deflection rate can raise your bill. Sellio's AI agent, for example, answers on website chat from your own pages and documents, hands off to a person when it is unsure and can be capped so the spend cannot run away. The chatbot ROI calculator helps you check whether the savings beat the cost.
Setting realistic targets by team size
Targets should match the hours you can actually staff. These are sensible starting points, not industry benchmarks. Adjust them after a month of real data.
| Team | Chat first response | Email first response | CSAT to aim for | Review rhythm |
|---|---|---|---|---|
| Solo owner | Under 5 minutes when chat shows as online | Same business day | 90%+ | 15 minutes on Monday |
| 2 to 3 people | Under 2 minutes in staffed hours | Within 4 business hours | 90%+ | 20 minutes weekly, together |
| 4 to 10 people | Under 1 minute in staffed hours | Within 2 business hours | 90%+, tracked per channel | Weekly review plus a monthly look at reasons |
Two habits make any target easier to hit. Show chat as online only when someone is watching it, and set the offline message to promise a reply time you know you can keep. The agent capacity planner estimates how many people a given chat volume needs, and the after-hours coverage calculator shows what evenings and weekends would cost to staff.
A simple weekly review routine
Twenty minutes on the same day each week is enough. Keep a shared sheet with one row per week so trends are visible.

- Write down the median first response time per channel and the median resolution time.
- Note the oldest open conversation and why it is still open.
- Record CSAT and the number of ratings, then read every comment below 4.
- Look at the top three contact reasons and their change from last week.
- Pick one fix for the coming week: a saved reply, a product page edit, a policy clarification or a change to chat hours.
- Once a month, add contacts per 100 orders and the chat to sale rate.
If your tool already reports response time, resolution and CSAT over the same conversations as the inbox, as Sellio's support analytics do, most of this is reading numbers rather than calculating them. The support KPI scorecard is a free template if you would rather keep it in a sheet.
What to ignore for now
Large support teams track dozens of numbers. Most of them answer questions a small store does not have yet.
- Net Promoter Score after every ticket. NPS measures loyalty to the brand. Sending it after a support chat mixes up the product, the delivery and the agent.
- Average handle time and tickets per hour. These are staffing metrics for big teams. On a small team they push people to rush customers.
- Agent utilization. With two people who also pack orders and run ads, the number is meaningless.
- Total conversation count on its own. It rises with sales. Contacts per 100 orders tells you what you actually want to know.
- Benchmarks from other industries. A software help desk and a clothing store get different questions at different speeds.
Mistakes to avoid
- Measuring time on a 24-hour clock when you staff 9 to 5. Report business hours, or every Monday looks like a disaster.
- Changing tag definitions halfway through a quarter. Decide what "order status" covers and write it down.
- Chasing a number instead of the problem behind it. Closing conversations early improves resolution time and annoys customers.
- Reviewing numbers without making a change. Each review should end with one fix and a name next to it.
For a side-by-side look at how two common help desks report these numbers, see Freshdesk vs Help Scout for a 5-person team.
FAQ
What is a good first response time for an online store?
For live chat, under a minute or two while chat shows as online. For email, the same business day. Shopify's guide suggests under a minute for chat and one business day for email, and Gorgias's benchmark data puts the median across ecommerce brands at 6.3 hours.
What is a good CSAT score for ecommerce?
There is no single figure. Shopify's guide puts very small businesses above 95% and larger businesses around 80%, and Gorgias's data shows ecommerce averages of 4.45 to 4.66 out of 5. With fewer than 30 ratings a month, watch the trend and the comments rather than the exact number.
How many support metrics should a small store track?
Five to seven. First response time, resolution time, oldest open conversation, CSAT and contact reasons cover most of what matters. Add contacts per 100 orders and chat to sale once a month, and AI handoff rate if you use a bot.
How do I measure support metrics without a help desk?
Export your conversations or emails to a sheet once a week with the time received, the time of the first reply, the time closed and a reason tag. Medians and counts take a few formulas. Most chat and help desk tools calculate the same numbers for you.
Should AI-resolved chats count in my response time?
Report them separately. An AI reply arrives in seconds, so mixing it in makes your human response time look better than it is. Track human first response time on its own and the AI handoff rate beside it.
What is a good contacts per order ratio?
It depends on what you sell. Gorgias's data ranges from 19 tickets per 100 orders in toys and games to 46 in electronics. Use your own trend: if the ratio rises while orders are flat, something in the store or the delivery has changed.




