Most support teams track too many metrics and understand too few of them. Dashboards fill up with ticket counts, agent leaderboards, and tag clouds while the numbers that actually predict customer loyalty go unexamined. This guide covers the six KPIs worth your attention, the 2026 benchmark ranges that keep them honest, the formulas behind each, and the vanity metrics you can safely demote. Benchmarks below come from published 2026 industry analyses (wow24-7, Ferndesk, Ringly — all observed September 2026); treat them as orientation, not targets carved in stone, because the right target depends on your segment and channel mix. This is research-based analysis, not first-hand product trials.
- Track six: first response time, resolution time, CSAT, first contact resolution, SLA compliance, and reopen rate. Everything else is secondary.
- Always split FRT by channel — a blended average across email and chat is meaningless because customer expectations differ by an order of magnitude.
- 2026 benchmarks (published analyses, Sept 2026): chat FRT under ~2 min typical, email FRT 4–12 hours typical; CSAT 85–95% on post-ticket surveys; FCR 55–75% typical.
- CSAT without response-rate context is a vanity metric — a 95% score from 8% of customers tells you nothing.
- Your help desk either measures these natively or it doesn't; "export to a spreadsheet and pivot" is not a reporting strategy.
KPI 1: First response time (FRT)
What it is: the elapsed time between a customer submitting a ticket and an agent's first substantive reply. Auto-acknowledgements don't count — "we got your email" is not a response.
Formula: sum of (first reply timestamp − ticket created timestamp) ÷ number of tickets.
Why it matters: response speed is the single most visible signal of whether you care. Published 2026 analyses put typical chat FRT around 2 minutes (top-tier under 40 seconds) and typical email FRT at 7–12 hours, with top-tier teams under 4 hours (wow24-7's 2026 contact-center benchmarks; Ringly's ecommerce KPI guide). Sixty percent of customers define "immediate" as ten minutes or less — yet only about a third of companies meet response expectations across channels.
The trap: blended FRT. One number averaging chat (seconds) and email (hours) tells you nothing about either. Split by channel, and split further by priority tier — your VIP queue and your general queue are different services with different promises.
How to move it: auto-assignment rules, SLA timers with escalation, and macros for the common first replies. AI triage that drafts or sends first responses is the 2026 lever — but measure whether it improves resolution, not just the timestamp.
KPI 2: Average resolution time
What it is: total elapsed time from ticket creation to resolution — the customer's whole experience of waiting.
Formula: sum of (resolved timestamp − created timestamp) ÷ resolved tickets. Decide whether you count business hours or calendar hours and never change the definition mid-year.
Why it matters: FRT measures courtesy; resolution time measures competence. A fast first reply followed by a nine-day resolution is worse than a slow reply with a same-day fix — the customer remembers the total. Ferndesk's 2026 metrics roundup cites common ranges of 4–24 hours for account/billing issues and 1–3 business days for technical B2B cases (ferndesk.com).
The trap: agents gaming the metric by closing tickets prematurely — which is exactly why you track reopen rate (KPI 6) alongside it. Any resolution-time target without a reopen-rate guardrail manufactures fake speed.
How to move it: knowledge-base articles linked in replies (agents who link instead of explain resolve faster), skills-based routing so tickets land with the right specialist first, and escalation paths that don't require the customer to re-explain.
KPI 3: CSAT
What it is: customer satisfaction, usually a post-resolution survey asking "how satisfied were you?" on a 1–5 or 1–7 scale. Reported as the percentage of 4–5 (or top-2-box) responses.
Formula: (satisfied responses ÷ total responses) × 100.
Why it matters: it's the closest thing to a direct measurement of the customer's feeling about the interaction. Published 2026 ranges put post-ticket CSAT at roughly 85–95%, with ecommerce averages around 75–82% and top performers above 85% (Ferndesk; Ringly).
The trap: response rate. A 96% CSAT from 8% of customers is a survey of your happiest customers, not a measurement of your service. Always report CSAT with the response rate, and watch for the pattern where only angry or delighted customers bother — the silent middle is where churn brews. Also: survey timing moves the number several points on its own, so keep the methodology constant.
How to move it: fix the drivers, not the score. CSAT correlates most with resolution speed and first-contact resolution — chasing the survey number directly (begging for 5 stars, gaming the send list) corrupts the instrument.
KPI 4: First contact resolution (FCR)
What it is: the percentage of tickets resolved in the first interaction, with no follow-up needed.
Formula: (tickets resolved on first contact ÷ total tickets) × 100. Define "first contact" precisely — same conversation thread vs. same day matters.
Why it matters: FCR is the efficiency metric that doesn't punish agents for being thorough. High FCR means routing works, knowledge is accessible, and agents are empowered. Common 2026 ranges run 55–75%, with 70–79% described as "good" and 80%+ as world-class (Ferndesk; wow24-7).
The trap: FCR varies enormously by ticket complexity — a password-reset desk and a B2B API-support desk aren't comparable. Benchmark against your own history and your segment, not the internet's average. Also, AI-resolved conversations complicate the definition: decide whether an AI resolution counts as "first contact" and document it.
How to move it: better routing (the right agent first time), empowered agents (authority to issue refunds/credits without escalation), and a knowledge base current enough that agents link articles instead of writing essays.
KPI 5: SLA compliance
What it is: the percentage of tickets handled within your contracted or promised service levels — e.g., first response within 2 hours for priority tickets.
Formula: (tickets meeting SLA ÷ total tickets under SLA) × 100.
Why it matters: SLAs are the only KPI with contractual teeth — enterprise contracts attach penalties to misses. Common 2026 ranges run 90–98%, tightening sharply where penalties apply (Ferndesk). Even without contracts, SLA compliance is the operational heartbeat: it tells you whether the queue is under control.
The trap: SLAs you can't enforce. Promising a 2-hour first response on a shared Gmail inbox is fiction — there's no timer, no escalation, no measurement. The day you sign an SLA is the day you need a real help desk with SLA policies, which is why this metric appears in every serious migration plan.
How to move it: realistic SLA tiers (don't promise chat-speed on email), automated escalation before breach rather than after, and staffing to the SLA — tighter promises cost more headcount, a trade-off to make explicitly.
KPI 6: Reopen rate
What it is: the percentage of resolved tickets that customers reopen — the "actually, this isn't fixed" rate.
Formula: (reopened tickets ÷ resolved tickets) × 100.
Why it matters: it's the quality control on every other metric. Fast resolution times mean nothing if 15% of tickets bounce back. Published 2026 ranges put reopen rates at roughly 3–10% (Ferndesk). A climbing reopen rate usually means documentation has lagged the product, or agents are being pressured to close quickly.
The trap: treating reopens as agent failure. Reopens are a system signal — stale KB articles, missing macros, or resolution-time targets set too aggressively. Fix the system, not the person.
How to move it: confirmation step before closing ("does this resolve your issue?"), KB articles updated with each product release, and post-reopen review — every reopened ticket is a free bug report on your process.
The vanity metrics hall of shame
Total ticket count. Up, down, sideways — it tells you about volume, not quality. Useful for staffing; useless for judging performance. (Falling ticket count can mean great self-service or customers giving up — the number alone can't tell you which.)
Average handle time as an agent target. Genuinely useful for capacity planning — multiply by expected volume to estimate staffing needs. Actively harmful as an individual target, because the fastest way to lower it is to end conversations early. If you must track it, pair it with CSAT and reopen rate.
NPS per ticket. NPS measures relationship sentiment with your company, not one support interaction. A detractor score after a chat usually reflects a pricing or roadmap grievance the agent couldn't touch. Read NPS quarterly alongside churn; never use it to grade tickets.
CSAT without response rate. Covered above — a score without its denominator is marketing, not measurement.
AI deflection rate without a cost line. "Our AI deflects 40% of tickets" means nothing until you see the per-outcome or per-resolution invoice. Deflection is a cost story; tell it with costs attached.
How your help desk choice affects your KPIs
Your platform decision determines which of these you can even measure. The practical differences:
- Native reporting vs. exports. Zendesk, Freshdesk, Intercom, and HubSpot all report FRT, resolution time, and CSAT natively with per-channel splits. Cheaper or free tiers often gate the good reporting behind upgrades — Freshdesk's deeper analytics sit on Pro ($55/agent/mo) and up, which is a real cost of the $19 Growth tier. Check what's behind the paywall before you buy, not after.
- SLA engines. Real SLA policies with timers, escalation, and breach alerts are a paid-tier feature almost everywhere. If SLAs matter to your contracts, verify the SLA feature exists on the exact tier you're buying.
- CSAT tooling. Native post-resolution surveys with response-rate reporting beat bolted-on survey tools — the integration determines whether you get the denominator.
- AI and the metric definitions. AI first responses, AI resolutions, and AI-assisted replies each need a documented definition in your KPI methodology, or your year-over-year trends become meaningless the quarter you turn the AI on.
The through-line: pick the desk whose reporting matches the KPIs you've committed to, especially the ones in customer contracts. Our help desk comparison hub evaluates the major platforms on exactly these dimensions.
—
FAQs
What is a good CSAT score for a help desk?
Published 2026 analyses put post-ticket CSAT at roughly 85–95%, with 90%+ considered top-tier. But the number is meaningless without the response rate — always report both, and benchmark against your own trend and segment rather than a universal target.
What's more important: first response time or resolution time?
Resolution time predicts satisfaction better, but first response time is what the customer feels first. Track both, split FRT by channel, and never set a resolution-time target without a reopen-rate guardrail — otherwise you manufacture fake speed.
How many KPIs should a support team track?
Six, covered in this guide. Teams that track twenty metrics understand none of them; teams that track three miss the quality-control signals. Six is the sweet spot: two speed metrics, two quality metrics, one contract metric, one guardrail.
Should AI-resolved tickets count in our KPIs?
Yes, but with documented definitions. Decide up front whether an AI resolution counts as first contact, how AI first-responses affect FRT, and report AI-handled and human-handled cohorts separately for at least the first two quarters. Otherwise your trends break the month you deploy.
Our CSAT is 97% but churn is rising — what gives?
Your survey is measuring the wrong population. Check the response rate (it's probably single digits), check who responds (probably the delighted and the furious), and look at the silent middle. Also verify the survey goes to a representative sample — some teams accidentally survey only resolved tickets, excluding the customers who gave up.