The customer service metrics that matter for online stores

Black Friday is ten weeks away, on November 27. During peak week your support numbers will look their worst, and the inbox will fill faster than anyone can clear it. That is when the right numbers tell you what to fix within the hour, and why you should set them up now, while things are calm.

The trouble is that most stores either track nothing ("we seem busy") or export every chart their help desk offers and read none of them. Neither helps you decide anything.

This guide covers seven metrics that matter for an online store, how to measure each one with the tools you already have, and what to do when one moves the wrong way.

Why stores measure the wrong things

The usual problems:

• Vanity totals. "We handled 4,000 conversations" says nothing about whether customers were helped.
• Speed in isolation. A fast first reply that doesn't answer the question just produces a second message.
• Bot-flattering numbers. "AI handled 70%" often includes customers who gave up.
• No link to orders. Support numbers that ignore sales miss the fact that chat can make money as well as cost it.
• No routine. Numbers that nobody reviews on a fixed day don't change anything.

A short, balanced set of metrics fixes all five.

The seven metrics

1. Contact rate

What it is: conversations divided by orders, for the same period. Why it matters: it is the best single measure of how well your store answers questions before customers ask. It also normalises for growth: when orders double, volume should double, not triple. When it slips: look at the top reasons for contact. A jump in order status questions points to shipping communication; a jump in sizing questions points to a product page.

2. First response time

What it is: how long a customer waits for the first meaningful reply, measured per channel and in business hours. Why it matters: it's what customers feel most. On chat it decides whether a shopper stays or leaves. When it slips: check staffing at peak hours, let the AI agent answer common questions instantly, and use saved replies for the top situations.

3. Resolution time

What it is: time from first message to a solved issue. Why it matters: a fast first reply means little if the refund takes a week. When it slips: find where conversations wait. It's usually on a decision (approve a refund?), on missing information (order number, photos) or on a third party (a carrier, a warehouse). Clear policies and forms that collect details up front fix the first two.

4. Automated resolution rate

What it is: the share of conversations closed by the AI agent or a flow without a person, where the customer didn't come back within a few days or ask for a human. Why it matters: it shows how much work automation really removes. When it slips: read the transcripts the AI handed over or got wrong, and fix the help content behind them.

5. Customer satisfaction (CSAT)

What it is: the share of customers who rate a conversation as good, usually from a one-question survey after it closes. Track it separately for AI and human conversations. Why it matters: it catches what speed metrics miss, such as wrong answers, cold tone or bad policies. When it slips: read the low-rated conversations. Most fall into a few causes: slow replies, a policy customers find unfair, or a specific product.

6. Repeat contact rate

What it is: the share of customers who write again about the same issue within a few days. Why it matters: it's an honest test of whether answers were correct and complete, for humans and AI alike. When it slips: check which topics come back most and rewrite the saved replies or help content for them.

7. Chat-to-order conversion

What it is: the share of pre-sale chat conversations that end in an order, and the value of those orders. Why it matters: it turns support from a cost line into a revenue channel and justifies the time spent on fast pre-sale answers. When it slips: check response times on product and cart pages, and whether agents and the AI know enough about the products.

Summary: what to track and when

Metric	How to calculate	Review	A warning sign
Contact rate	Conversations ÷ orders	Monthly; daily in peak	Rising faster than orders
First response time	Median time to first reply, per channel	Weekly; daily in peak	Above the times you publish
Resolution time	Median time from first message to solved	Monthly	Growing on one topic, like refunds
Automated resolution rate	AI-closed, not reopened ÷ AI conversations	Monthly	Flat for months, or rising while CSAT falls
CSAT	Good ratings ÷ all ratings	Monthly	AI CSAT well below human CSAT
Repeat contact rate	Customers writing again on the same issue ÷ all	Monthly	Concentrated on one topic
Chat-to-order conversion	Chats ending in an order ÷ pre-sale chats	Monthly	Falling while chat volume rises

Use medians, not averages, for time metrics. One conversation left open over a weekend can wreck an average and hide what most customers experienced.

How to set this up in an afternoon

1. Tag the reason for every conversation, by hand or with automatic tags: order status, returns, product question, payment, discount, complaint, other.
2. Turn on a one-question satisfaction survey after conversations close, on chat and email.
3. Connect your store so conversations link to orders, which you need for contact rate and chat-to-order conversion.
4. Write the seven metrics in a simple sheet, one row per month, filled from your support tool's reports.
5. Book 30 minutes on the first Monday of each month to review them with whoever answers customers.

Reading the numbers during peak week

When Black Friday and Cyber Monday arrive, they distort everything, so change how you read the numbers:

• Look daily, not monthly. Response time and backlog tell you whether to pull in help today.
• Watch the reasons, not just the totals. Ten messages about one discount code is a bug to fix, not a staffing problem.
• Expect contact rate to rise. New customers ask more. Compare against last year's peak, not against October.
• Protect quality. If CSAT drops while speed holds, people are rushing. Slow down on complaints and payment issues.

For the full peak-season playbook, see how to prepare customer service for Black Friday.

Common measurement mistakes

• Counting AI "handled" as resolved. Use resolved and not reopened.
• Mixing channels. Email and chat response times on one chart hide both problems.
• Ignoring business hours. Count overnight waits separately, or the AI agent's instant replies get mixed in with your team's.
• Ranking agents on speed alone. It rewards closing conversations, not solving them.
• Never acting. If a metric hasn't changed a decision in three months, stop tracking it.

Which one should you choose?

• Small or mid-size store, any platform: Tidio. The reports cover response times, AI resolutions and chat-driven sales in the tool your team already works in.
• Established brand with a support team: Tidio Premium if AI handles a large share of your volume. It adds advanced and custom analytics, and guarantees a 50% Lyro resolution rate, so the AI number you report on is one Tidio is accountable for. On Shopify, Gorgias is the alternative for revenue and agent-level reporting.
• Large, multi-brand operation: Zendesk, with someone to own the analytics.
• Small team that wants a classic help desk: Re:amaze.

Whatever you use, the metrics only matter if they lead to changes. Pair this with our e-commerce customer service best practices to decide what to fix first.