Key Takeaways
- Customer operations KPIs should lead with one question: did the customer have to come back? First contact resolution only sees the conversation that just ended.
- The research behind effort metrics recommends tracking repeat contacts within seven to 14 days instead of first contact resolution. It casts a wider net and is easier to measure.
- “Resolved” contacts still come back. In the same study of more than 75,000 B2C and B2B customers, 22% of repeat calls involved downstream issues from a problem that had been fixed.
- Benchmarks rarely transfer. SQM's call-center first contact resolution average is 71%, but tech support averages 64%, and the data comes from inbound call centers.
- Self-service needs its own measure. Customers fully resolve only 14% of issues in self-service, and the top reason is that they can't find relevant content.
- Onboarding KPIs can carry causal weight. In a randomized experiment at a cloud provider, proactive onboarding guidance halved first-week churn.
The monthly operations review has 14 metrics on it. First contact resolution is 78%, well above the industry average printed in small type beneath it. Satisfaction is steady. Average handle time is down. Carmen's CFO asks one question: if we're resolving four in five contacts first time, why did request volume grow faster than customers again? Nobody on the call can answer, because none of the 14 metrics measures whether a customer came back.
Most lists of customer operations KPIs are really customer service KPIs: response time, handle time, first contact resolution, satisfaction. They measure the queue. Customer operations covers support, onboarding, self-service and the knowledge behind them, so its KPIs need to measure what happens across those surfaces. This post covers which measures do that, how to set their denominators and time windows, and which benchmarks transfer to a B2B software team.
Why first contact resolution misleads as a headline KPI
First contact resolution is the most quoted operational metric in service. It's also the one most likely to look healthy while volume grows. The reason is in how it's counted.
The best evidence comes from a three-year CEB study published in Harvard Business Review. The authors surveyed “more than 75,000 B2C and B2B customers about their recent service interactions.” They found that the biggest cause of customer effort was having to call back. They also found that first contact resolution doesn't catch most of the reasons customers do.
That's why Carmen's 78% and her volume growth can both be true. First contact resolution measures the conversation. Repeat contact measures the customer.
The customer operations KPIs that matter, with their denominators
A KPI is only as good as its formula. Two teams can both report “first contact resolution” and mean different things, depending on what they divide by. The card below sets out each measure with its formula, time window, blind spot and the population behind any benchmark.
| KPI | Formula | Window | What it misses | Benchmark and its population |
|---|
| Repeat contact rate | Customers who contact again about the same or a related issue, divided by customers who made contact | 7 to 14 days, all channels | Customers who give up instead of coming back | No B2B benchmark; set your own baseline |
| Net first contact resolution | Contacts resolved on first contact, divided by contacts that could have been resolved on first contact | Per contact | Downstream issues and channel switches | SQM call centers: 71% average, 64% in tech support |
| Self-service full resolution | Customers who report their issue fully solved in self-service, divided by customers who tried it | Per journey | Silent failures that never reach a survey | Gartner consumer study: 14% |
| Content findability | Searches with a clicked result and no follow-up contact, divided by all searches | Per session, then 7 days | Customers who never search | Gartner: 43% of self-service failures were missing relevant content |
| Cost per contact | Total operating expense divided by contact volume | Monthly | Cost moved onto other teams | Gartner 2019 poll: $8.01 live vs $0.10 self-service |
| First-week questions per new account | Questions from new accounts in days 1 to 7, divided by new accounts | First 7 days | Accounts that go quiet | One field experiment; no general benchmark |
Sources: repeat contact window from HBR 2010; net FCR and cost per contact definitions from MetricNet and MetricNet's cost per ticket definition; FCR benchmarks from SQM; self-service figures from Gartner via destinationCRM; cost per contact from Gartner's 2019 poll; first-week measure from a 2016 Manufacturing & Service Operations Management study.
The formulas for repeat contact rate, content findability and first-week questions are our recommendations built from those sources. The benchmarks are what the sources report, with their populations stated. Where no benchmark exists for B2B software, the card says so. That's more honest than borrowing a consumer number.
How to read the benchmarks
Every published customer operations KPI benchmark comes from a population. Most of those populations aren't B2B software teams.
First contact resolution: call centers, customer-reported
SQM runs the most cited benchmark. Its average center reaches 71%, against an 80% world-class target that “only 5% of call centers can achieve.” The method matters. SQM's data comes from inbound customer service call centers only, with at least 400 surveys per center. Tech support averages 64%, and technical support calls specifically come in at 60% in the 2024 results. A B2B software team comparing itself to the 71% average is comparing against the wrong population.
Net versus gross: the denominator changes the number
MetricNet defines net first contact resolution as contacts resolved first time “divided by all calls that are potentially resolvable on first contact.” Gross resolution divides by every contact, including ones no agent could have solved in one touch, such as bugs needing engineering. A team moving from gross to net can gain several points without changing anything. State which one you report.
Self-service: consumer data, and a knowledge problem
Gartner's 2024 survey of 5,728 customers found that customers “resolve only 14 percent of their service and support issues fully in self-service.” It's a consumer study, so treat it as a floor, not a B2B benchmark. The reason matters more than the rate: in 43% of cases, customers couldn't find content relevant to their issue. That's a knowledge KPI hiding inside a channel KPI. Our post on why self-service rates plateau covers the content side.
Cost per contact: useful, but dated
The most quoted figures are from a 2019 Gartner poll: $8.01 per live contact and about 10 cents per self-service contact. They're seven years old and consumer-weighted. Use MetricNet's definition of total operating expense divided by volume on your own numbers instead, and watch the trend rather than the absolute.
The onboarding KPI with causal evidence
Most KPI evidence is correlational. One study isn't. Researchers ran a randomized field experiment at a cloud computing provider in 2011, covering 2,673 new customers. A treated group got proactive onboarding guidance. The result: “the treatment reduces by half the number of customers who churn from the service during the first week.” Treated customers also asked 19.55% fewer questions in that first week.
It's one provider and one experiment, so it isn't a benchmark. It does show that first-week questions and first-week churn move together, and that operations work can move both. That makes first-week questions per new account a KPI worth tracking. It sits at the boundary between onboarding and support, which is exactly where customer operations lives. More on the onboarding side is in customer onboarding at scale.
How to set up customer operations KPIs in 30 days
- Define “same issue” before you count repeats. Use a shared topic taxonomy across support, onboarding and self-service. Without it, repeat contact can't be measured across channels.
- Link every contact to an account. Repeat contact and first-week questions are account measures. Anonymous tickets can't feed them.
- Start the 14-day window from the first contact on a topic. Count any further contact on that topic or a related one, in any channel.
- Log searches that end in a contact. A help-center search followed by a ticket within a day is the clearest findability signal you have.
- Report net and gross first contact resolution side by side for one quarter. Then pick one and label it on every slide.
- Drop the activity metrics nobody acts on. Our guide to metrics that change a decision covers what to stop reporting.
A dashboard built this way has fewer rows than the 14-metric review. Each row answers a question the CFO can ask: are customers coming back, can they find answers themselves, and are new accounts getting stuck?
Measuring across the surfaces, not inside each one
The KPIs above fail for one practical reason in most stacks. Repeat contact needs support tickets, chat, help-center searches and onboarding questions linked to the same account and topic. Those usually live in four tools with four taxonomies.
MatrixFlows puts them on one foundation. Requests from chat, email and forms land in the Conversations Inbox as records linked to the account. Knowledge articles and onboarding projects sit in the same Matrix tables, tagged with one shared taxonomy. App analytics track searches, AI answers and escalations from the help center and in-app assistant, and Trends can slice them by saved segments over time. A repeat contact across channels becomes a query instead of a spreadsheet project. If AI answers part of the load, the same records feed the measures in our guide to AI customer service metrics.
Carmen's next review had six rows. First contact resolution was still there, labelled net. Above it sat repeat contact within 14 days, and the CFO's question had an answer.
Create a Free Workspace → and set up the shared topic taxonomy that repeat-contact measurement depends on.
.svg)
.svg)
.svg)
