Customer service KPIs
Customer service KPIs — key performance indicators — are the quantitative measures that support organizations use to track quality, efficiency, and outcomes. A well-chosen KPI set tells a leader whether their team is making customers happy, running efficiently, and improving over time; a poorly chosen set optimizes the wrong behaviors and hides the real problems.
The universe of customer service KPIs is large, and no organization tracks every possible metric. The right set is small — usually 6 to 10 core KPIs — chosen so that each measures something specific and no two are so correlated that they duplicate signal. This page walks through the essential categories and the trade-offs among them.
Quality and satisfaction KPIs
CSAT (Customer Satisfaction) measures how satisfied customers felt with a specific interaction, typically via a post-contact survey with a 1-to-5 or 1-to-7 scale. CSAT reflects transactional experience and correlates well with short-term repeat behavior.
NPS (Net Promoter Score) measures a customer's overall likelihood to recommend the brand on a 0-to-10 scale, minus the fraction of detractors. NPS is a relationship-level metric and tracks slower-moving loyalty rather than any specific interaction.
CES (Customer Effort Score) measures how easy it was for the customer to accomplish their goal. Low-effort experiences correlate with retention more strongly than high-satisfaction ones — customers who report an interaction was "very easy" churn less than those who report it was "very delightful."
Efficiency KPIs
AHT (Average Handle Time) is the mean total time an agent spends per contact, including hold time and wrap-up. AHT is the workhorse efficiency metric for voice channels but is easy to misuse: pushing AHT down at the expense of resolution rate creates more repeat contacts and worse outcomes.
ART (Average Response Time) is the average time from when a customer submits a contact to when an agent first responds. Response time is especially important on email and chat, where slow responses drive customers to call anyway.
Cost per contact divides total support operating cost by total contacts handled. It's a useful summary metric but only meaningful when tracked alongside resolution and satisfaction — cheap contacts that never resolve issues generate expensive repeat contacts.
Occupancy is the fraction of an agent's paid time actually spent handling contacts. High occupancy sounds efficient but above 85% correlates with burnout and turnover.
Outcome KPIs
FCR (First Contact Resolution) measures the percentage of contacts resolved without the customer needing to reach out again. FCR is the single best proxy for support quality because it aligns almost every stakeholder: customers prefer to resolve their issue once, and the business prefers not to pay for repeat contacts.
Deflection rate measures the fraction of contacts resolved through automation (chatbots, AI agents, self-service) without human agent involvement. Deflection is increasingly central as AI-driven support automation matures — see customer support automation.
Containment rate is a similar metric specific to AI agent conversations — the fraction of conversations the AI resolves end to end without escalating to a human.
Resolution time tracks the total time from contact opening to full resolution, including any back-and-forth. Unlike AHT, resolution time reflects customer-experienced wait for their issue to actually be fixed, not just the agent's active work.
Volume KPIs
Ticket volume by intent, channel, product area, or customer segment reveals where support demand is concentrated and how it changes over time. Volume analysis often surfaces product bugs, poorly worded UI copy, or documentation gaps — issues that support ops can bring to product teams as reduction opportunities.
Volume forecast accuracy tracks how closely actual contact volume matches the forecast used for staffing. Poor forecast accuracy leads to under-staffing (long queues, high AHT) or over-staffing (idle agents). See forecast accuracy.
Schedule adherence measures whether agents are working the shifts they were scheduled for. Poor adherence undermines forecast accuracy and creates queue-length spikes.
Balancing the KPIs
No single KPI captures customer service performance well because the trade-offs between metrics are real. Aggressively cutting AHT eventually degrades FCR. Optimizing exclusively for deflection can degrade CSAT if the automation isn't good enough for the deflected volume. Chasing a higher CSAT by giving away refunds inflates cost per contact.
The best support organizations track a balanced scorecard with metrics from each category and set targets that make trade-offs explicit: "AHT must not exceed 6 minutes, but not at the cost of FCR falling below 80% or CSAT falling below 4.5." When two metrics move in opposite directions, the KPI set forces a conscious decision rather than a silent shift.
KPIs in AI-augmented support
Adding AI agents to a support organization changes which KPIs matter most. Containment and deflection become primary metrics because they measure the AI's contribution. Cost per contact drops for AI-handled work but must be measured separately from the human-handled residual to avoid mixing the two populations. Quality metrics on AI-handled conversations (CSAT, task completion, sentiment change) need to be tracked with the same rigor as human-handled conversations — the AI's tier-1 performance now bears on the overall brand experience.
A well-designed KPI set for an AI-augmented support organization tracks the AI and human populations both separately and together, keeps quality metrics prominent alongside cost metrics, and sets targets that reflect the leadership's actual priorities.
