Customer Support Metrics & KPIs: The Ones That Actually Matter

The customer support metrics that matter most for a small software team are median first response time, resolution time, first contact resolution, abandoned conversations and conversations per customer, plus a quality measure such as CSAT or graded conversation quality. Track three or four of them, chosen for your current problem, set targets against your own baseline, and watch the trend each week. A dashboard of twenty charts rarely changes what anyone does on Monday morning.

This guide explains each metric, how to calculate it, how to set a target, which to pick for which problem, and how to connect support to revenue.

Which customer support metrics should a small team track?

A small team should track a handful of metrics across speed, quality and volume, and ignore the rest until they become a problem. A starting set:

Pick the three or four metrics that match your current bottleneck (see the problem-to-metric table below), and add others only when you have a reason to act on them.

How do you measure first response time?

First response time (FRT) is the time from a customer’s first message to the first reply from a person. It is the metric customers feel most, because waiting in silence is what turns a small question into frustration.

First response time = time of the first reply from a person minus time of the customer’s first message. Report the median for the period.

With illustrative figures: five conversations got first replies after 5, 12, 20, 45 and 540 minutes. The median is 20 minutes and the average is 124 minutes, so the median shows what a typical customer saw while the average is dragged up by one slow reply. Exclude automated replies, and if you promise business-hours coverage, count only business hours. Add the 90th percentile (p90) to see the slow tail: for example, if the median is 20 minutes but the p90 is 9 hours, one in ten customers is waiting far too long.

Set a target you can meet consistently, such as a reply within 30 minutes during working hours, and measure the share of conversations that hit it. Convot’s response times report shows median and p90 first response and resolution times, daily trends, and SLA attainment against a target that defaults to 30 minutes. It calculates these per session, the period one agent is assigned, and resolution time counts only sessions that ended as resolved.

How do you measure resolution time?

Resolution time is how long a conversation takes to get from start to resolved. First response time shows you noticed; resolution time shows you fixed it.

Resolution time = time the conversation was marked resolved minus time it started. Report the median and the p90 for the period.

Track the median and the p90 here too. The handful of conversations that drag on for days are usually the ones that end in a bad review or a lost customer, so list them each week and ask why they stalled: waiting on engineering, waiting on the customer, or nobody owning them.

What is first contact resolution, and how do you calculate it?

First contact resolution (FCR) is the share of resolved conversations that stay resolved, without being reopened. It is a quiet driver of both satisfaction and cost: every reopened conversation is a customer explaining the problem again and an agent reloading the context.

First contact resolution = resolved conversations that were not reopened ÷ all resolved conversations.

Its mirror is the reopen rate, so FCR = 100 percent minus the reopen rate. With illustrative figures: if 400 conversations were resolved in a month and 60 were reopened, the reopen rate is 15 percent and FCR is 85 percent. A low FCR usually means answers that miss part of the question, unclear help articles, or conversations closed before the customer confirmed the fix. Convot’s Outcomes report shows FCR, reopen rate and abandoned rate together.

What is the CSAT formula and how do you calculate it?

CSAT (customer satisfaction score) comes from a short rating after a conversation, usually “Was this helpful?” or a 1 to 5 score.

CSAT = (positive responses ÷ all responses) × 100.

With illustrative figures: if 42 of 60 customers rate a conversation 4 or 5, CSAT is 70 percent.

At low volume, CSAT is noisy and biased towards customers who bother to answer, so watch the trend over several weeks rather than any single day. Read the comments on negative ratings; they say more than the number.

Graded quality is a useful complement because it does not depend on customers choosing to rate. AI quality grading scores each conversation your team handles against a rubric and flags the bad ones with a reason. In Convot, AI quality grading, once Auto-QA is turned on, scores conversations your team handles, and CSAT from customer feedback is reported for conversations the Cove AI agent handled.

What is customer effort score?

Customer effort score (CES) measures how hard the customer had to work to get help, usually by asking them to agree or disagree with “The company made it easy to handle my issue.” Lower effort is better.

Effort matters because reducing it tends to do more for loyalty than trying to delight customers; that was the central finding of Matthew Dixon, Karen Freeman and Nicholas Toman’s “Stop Trying to Delight Your Customers” (Harvard Business Review, July-August 2010 issue). In practice, effort shows up as customers repeating themselves, being passed between people and having to follow up. You can reduce it without a survey by fixing those.

What is the abandoned conversation rate?

The abandoned rate is the share of conversations in which a customer wrote at least one message and never got a reply from a person. It is the bluntest coverage metric there is: each abandoned conversation is a customer who asked for help and got nothing from your team.

Abandoned rate = conversations with a customer message and no agent reply ÷ all conversations.

For a small team, the target is close to zero. Abandoned conversations usually come from messages that arrive outside working hours with no away message, or from conversations nobody owned. In Convot, replies from the Cove AI agent do not count as agent replies, so a conversation the bot handled alone shows as abandoned; read the rate with that in mind.

How should you track ticket volume?

Track ticket volume relative to your customer base, not as a raw count. More conversations is not bad in itself; it may mean more customers. Normalise it:

Conversations per customer = conversations in the period ÷ active customers in the period.

Watch the direction. A rise after a release is a product signal, not a staffing problem. A conversation heatmap by day and hour shows when questions arrive, so you can match coverage to demand. Convot includes one.

What is ticket deflection rate?

Ticket deflection rate is the share of customers looking for help who found the answer themselves instead of starting a conversation, usually through a help center or in-product hints.

Deflection rate = help-seeking sessions that ended without a conversation ÷ all help-seeking sessions.

It is what lets support volume grow more slowly than your customer base. The ticket deflection playbook covers how to raise it without making support harder to reach.

What is net promoter score, and should support track it?

Net promoter score (NPS) asks “How likely are you to recommend us?” on a 0 to 10 scale.

NPS = percentage of promoters (9 or 10) minus percentage of detractors (0 to 6).

With illustrative figures: 100 responses with 50 promoters, 30 passives and 20 detractors give an NPS of 50 minus 20 = 30.

NPS is a relationship metric, not a conversation metric, so survey periodically rather than after every conversation. It is most useful as a slow trend and as a prompt to read the comments, where the real product and support insight lives.

Which customer support KPIs should a SaaS or software team track?

A SaaS or software team should track median first response time, median resolution time, first contact resolution, abandoned rate and conversations per customer, then add CSAT or graded quality as the quality check. Choose three or four by the problem you have now:

Revisit the set every quarter. When one number is under control, move your attention to the next constraint.

How do you report support metrics to the rest of the team?

Report support metrics in a short weekly note that pairs each number with what changed and what you will do about it. A number alone invites debate; a number with a cause and an action gets used.

A useful weekly note has four lines:

1. Speed: median and p90 first response time against the target.
2. Outcomes: first contact resolution and abandoned conversations.
3. What drove volume: the top three topics, and any release that caused a spike.
4. One action: the single change for next week, such as a new help article or a change to coverage hours.

Share product-related findings with engineering in the same note, because a cluster of questions about one screen is a product bug report.

How do you set targets for support metrics?

Set targets against your own baseline, not someone else’s benchmark. Published benchmarks mix very different channels, products and team sizes, so they are rarely a fair comparison for a small software team.

A simple method:

1. Measure four weeks of the metric as it is.
2. Set a target slightly better than your current median, one you can hit most weeks.
3. Review weekly, and tighten it once you hit it consistently.

For first response time, a target written as a promise customers can see, such as “we reply within an hour during business hours”, keeps the team honest.

How do you connect support metrics to revenue?

Connect support metrics to revenue by putting each customer’s value beside their conversations and checking, when customers leave, whether support played a part. Most dashboards show support as activity; the business question is what support does to retention and revenue.

For Shopify apps, Convot shows a merchant’s plan, MRR and lifetime revenue to date beside the conversation once your app identifies the shop with a signed shop_hash, with revenue visible to owners by default. With the Shopify Partner API connected, a fresh uninstall reopens that merchant’s conversation if they wrote in during the last 7 days (the default window), and Convot posts an AI verdict on whether support may have contributed: support not the cause, support may have contributed, or unclear. Those verdicts feed a support-attributable churn percentage in the revenue dashboard. The verdict is a signal to investigate, not proof.

The guides to reducing Shopify app churn and getting more Shopify app reviews cover what to do with those signals. The SaaS customer support guide covers the wider picture.

How do you track support metrics without building reports?

Use a help desk that records the timestamps for you. Calculating first response time from spreadsheets is slow and error-prone. Convot’s reports include response times (median, p90 and SLA attainment), outcomes (FCR, reopen and abandoned rates), agent performance, a conversation heatmap, help center analytics and AI quality grading, with a dashboard that pulls the key numbers together.

Common questions about customer support metrics

What is the most important customer support metric? For most small teams, median first response time is the best place to start, because it is the one customers feel first and the easiest to improve with ownership and coverage.

Should you use average or median response time? Use the median, plus the 90th percentile for the slow tail. Averages are distorted by a few very slow conversations.

What is a good first contact resolution rate? A good first contact resolution rate is one that is high and rising for your own team. Compare with your own baseline rather than published benchmarks, which mix very different kinds of support.

How often should you review support metrics? Review speed and coverage weekly, and quality, churn and NPS monthly or quarterly.

Customer support metrics, in seven steps

1. Pick three or four metrics for your current problem.
2. Use medians and p90s, not averages.
3. Measure a four-week baseline before setting targets.
4. Watch the abandoned rate and keep it near zero.
5. Pair speed with a quality measure, such as graded conversations.
6. Normalise volume per customer.
7. Connect support to revenue and review support-related churn.

Start free with response time and outcome reports while you are under $1,000 MRR, with 1 app and 3 seats.

About the author

Tarang Agarwal is the founder of Convot. Convot is built by Sidepanda, a small studio that runs several Shopify apps of its own, including Appointo, Depo and Panda Bundle, and supports all of them from one Convot inbox.