Customer Service Metrics for Online Stores: What to Track and How to Measure It

Most online stores don't have a support team big enough to justify a dashboard. They have a founder, a part-time teammate and an inbox. That is exactly when a few good numbers help most: they tell you whether customers are waiting too long, whether problems actually get solved, and whether your store is creating problems it could prevent.

This guide defines the customer service metrics that matter for an online store, shows how to calculate each from your own data, and explains what each one can't tell you. You won't find industry benchmarks here. Benchmarks mix stores of every size, channel and product, and the best comparison for your store is your store last month.

Start with three questions

Every useful support metric answers one of three questions. If a number doesn't help answer one of them, you probably don't need it yet.

• Are customers waiting?: First response time and resolution time.
• Are problems actually solved?: Resolution rate, repeat contacts and customer satisfaction.
• Why are customers contacting us at all?: Contact rate per order and contact reasons.

First response time

First response time is how long a customer waits between their first message and the first real reply, from a person or an AI agent.

• How to calculate it: For each conversation, subtract the time of the customer's first message from the time of the first reply. Then take the average, and also the median, across conversations.
• Why the median matters: One email answered after a long weekend can drag the average up for the whole month. The median shows what a typical customer experienced.
• What to exclude: Automatic "we got your message" replies. They are not a response; they are a receipt.
• What it misses: A fast reply that doesn't help. Always read it next to resolution.

Split it by channel. Customers expect a chat or a text to be answered much faster than an email, so one blended number can hide a slow chat queue.

Resolution time

Resolution time is how long it takes from the customer's first message until their problem is solved.

• How to calculate it: Time of resolution minus time of first message, averaged (and as a median) across resolved conversations.
• Define "resolved" first: Pick one rule and keep it: the conversation was closed, the requested change was made, or the customer confirmed. Changing the definition changes the number.
• Watch the waiting: A refund that waits two days for a warehouse to confirm the return is a process problem, not a support problem. Note where the time goes.

Resolution rate and repeat contacts

Resolution rate is the share of conversations where the customer's problem was actually solved. It is the hardest metric to measure honestly and the most important.

• How to calculate it: Resolved conversations ÷ all conversations in the period × 100.
• Check with repeat contacts: Count customers who wrote again about the same issue within a few days, say seven. A high repeat rate means some "resolved" conversations weren't.
• First contact resolution: A stricter version: the share solved in one conversation, with no follow-up needed. It is the number customers feel most.

AI resolution rate and deflection

If an AI agent answers some of your conversations, two more numbers matter, and they are easy to confuse.

Report both, and read conversations to check them. See the glossary entries for ticket deflection and AI resolution rate. Also track the handoff rate, the share of AI conversations passed to a person, and which topics they cluster in. Those topics are your list of knowledge or rules to add.

Customer satisfaction (CSAT)

CSAT asks customers to rate a conversation, usually on a 1 to 5 scale, right after it ends.

• How to calculate it: Ratings of 4 or 5 ÷ all ratings × 100.
• Watch the response rate: If only a small share of customers answer, the score mostly reflects the very happy and the very angry.
• Read the comments: A score tells you something is wrong. The comment tells you what.
• Don't confuse it with NPS: NPS asks whether someone would recommend your brand. It reflects the product and price as much as support.

Contacts per order

This is the metric most stores skip, and the one that points at fixes outside the support team.

• How to calculate it: Customer conversations in a period ÷ orders in the same period. Multiply by 100 to read it as contacts per 100 orders.
• Why it matters: It grows as your store gets busier only if something is causing contacts. Flat contact volume while orders double is a good sign.
• Break it down by reason: Tag or group conversations by topic: order status, address changes, returns, subscriptions, product questions, discount codes. The biggest group is your best fix.

If "where is my order" is your top reason, the fix is usually better shipping notifications, not faster replies. Our guide to shipping update texts and the glossary entry for WISMO cover that.

Backlog and unanswered customers

A simple count, checked daily: how many customers are waiting for a reply right now, and how long the oldest has waited. It is less elegant than an average and more useful on a Monday morning. Anyone waiting more than a day deserves a look before anything else.

How to measure all this without a data team

• 1. Pick one source of truth: Your help desk or inbox, not a mix of email, a phone app and someone's texts. Conversations outside it are invisible to every metric.
• 2. Write your definitions down: What counts as a reply, what counts as resolved, how many days counts as a repeat contact. One paragraph is enough.
• 3. Use the same window every time: Last 30 days, or each calendar month. Compare like with like.
• 4. Track five numbers, not twenty: Median first response time, resolution rate, repeat contact rate, contacts per 100 orders and your top three contact reasons.
• 5. Read conversations every week: Numbers tell you where to look. Reading tells you what is wrong. Our QA checklist for AI customer service has a review routine.

A worked example

Here is how the numbers fit together for an imaginary store, with made-up figures to show the math. Last month it had 2,000 orders and 260 customer conversations. 230 were resolved, and 20 customers wrote back about the same issue within seven days.

• Contacts per 100 orders: 260 ÷ 2,000 × 100 = 13.
• Resolution rate: 230 ÷ 260 × 100 = about 88%.
• Repeat contact rate: 20 ÷ 260 × 100 = about 8%.
• Top reason: 110 of the 260 were order status. Better shipping updates could remove a large part of the month's work.

None of these numbers is good or bad on its own. They become useful when you compare them with next month, after you change something.

Mistakes to avoid

• Chasing fast first replies with canned messages that don't help.
• Counting closed conversations as resolved without checking repeat contacts.
• Comparing yourself with industry averages that don't match your channels or products.
• Measuring the support team for problems the store creates, like slow shipping or unclear subscription terms.
• Changing definitions mid-year, then reading the change as progress.

How Chargezen helps

Concierge puts every customer conversation, from chat, text, WhatsApp, email and phone calls, in one team inbox, so your numbers come from one place. Its overview shows three numbers for the last 30 days:

• Conversations: Customer conversations started in the window.
• Tickets deflected: Conversations an agent answered with no handoff and no reply from a teammate.
• Average response time: From the customer's first message to the first reply, by an agent or a teammate.

For anything else, ask. Concierge can sum up last week's conversations by channel and topic, list customers still waiting on a reply, or group this month's complaints by theme, and it reads your orders to count contacts per order. On answers built from your data, Show work lists what it read and the commands it ran. See data analysis and Concierge for customer support.

Concierge Starter is $20 a month with 20,000 credits. Pro is $199 a month with 199,000 credits, or $1,990 a year. You can also connect your own OpenAI or Anthropic account. See pricing.

Related terms: first response time, resolution time, CSAT and AI resolution rate. Related guides: a QA checklist for AI customer service, AI escalation and human handoff and subscription metrics.

Frequently asked questions

What is the most important customer service metric for an online store?

Resolution rate, checked against repeat contacts. Fast replies matter, but customers remember whether their problem was solved.

What is a good first response time?

Faster than last month, on every channel. Chat and text customers expect minutes; email customers will usually wait longer. Set your own target per channel and track the median.

How do I calculate contacts per order?

Divide customer conversations in a period by orders in the same period. Multiply by 100 to get contacts per 100 orders, which is easier to read.

Is ticket deflection the same as resolution?

No. Deflection counts conversations that never reached a person, including customers who gave up. Resolution counts problems actually solved. Track both.