Skip to content
BPO November 6, 2026 12 min read

BPO KPIs and Metrics: Formulas, SLA Metrics and a Weekly Client Scorecard

BPO KPIs and metrics with formulas: SLA, productivity, outbound, commercial and people KPIs, the trap in each, and a weekly per-client scorecard.

D
November 6, 2026
Last updated

Quick answer

A BPO runs on five groups of KPIs. Client SLA metrics: service level, average speed of answer (ASA), abandon rate, first contact resolution (FCR), CSAT or NPS, and QA score. Productivity: average handle time (AHT), occupancy, utilisation, schedule adherence, shrinkage and contacts per hour. Outbound: connect rate, contact rate, right-party contact, conversion and abandoned-call rate. Commercial: billable versus paid hours, revenue per seat, cost per contact and margin per client. People: attrition and time to proficiency. Measure each one per client, write its definition into the contract, and set targets per client contract rather than from an industry average.

An in-house contact centre answers to one boss. A BPO answers to every client on the floor, and each reads a different scorecard. The client watches service level, quality and outcomes; finance watches billable hours and margin. A BPO that tracks only the client's KPIs can meet every SLA and still lose money on the account.

Below, each KPI gets a definition, the formula, what moves it, and the trap that makes it lie, followed by per-client reporting, a sample weekly scorecard, and getting the numbers without spreadsheets. For general formulas and a calculator, see call center KPIs and formulas; for a collections book, see collections calling metrics.

One rule first: no benchmark outranks the contract. Set targets per client contract, baseline for a few weeks, and track your own trend.

1. Client SLA metrics

The numbers in the statement of work. Get each definition signed off before go-live: most SLA disputes are two parties dividing by different things.

Service level

The share of calls answered within a set time. Formula: calls answered within threshold ÷ calls offered × 100. Moves it: interval staffing, AHT, routing, shrinkage. Trap: the contract must say how short abandons are treated and whether the figure is per interval, day or month. A monthly figure can pass while every Monday morning fails. See service level.

Average speed of answer (ASA)

How long answered callers waited. Formula: total queue wait of answered calls ÷ answered calls. Moves it: the same levers as service level. Trap: an average hides a few very long waits; add the 90th percentile wait. See ASA.

Abandon rate (inbound)

The share of callers who hung up before an agent answered. Formula: abandoned calls ÷ calls offered × 100. Moves it: wait time, queue messaging, callback offers. Trap: agree whether calls abandoned in the first few seconds count, and never compare it with outbound abandon rate, which uses a different denominator. See abandon rate.

First contact resolution (FCR)

The share of issues solved on the first contact. Formula: cases with no repeat contact inside the window ÷ eligible cases × 100. Moves it: agent knowledge, authority to act, fewer transfers. Trap: without a stated repeat-contact window (7, 14 or 30 days) FCR is a guess, and self-reported resolution reads higher than repeat matching. See FCR and how to improve it.

CSAT and NPS

CSAT rates one interaction; NPS rates the relationship. Formulas: CSAT = positive responses ÷ valid responses × 100; NPS = % promoters − % detractors. Moves them: resolution, wait and manner; much of NPS sits outside the BPO's control. Trap: always show response count and rate, and do not use NPS as a per-agent grade. See CSAT and NPS.

QA score

How well calls meet the client's quality standard. Formula: points earned ÷ points possible × 100 across scored calls. Moves it: coaching, calibration, script clarity. Trap: an average hides a critical failure such as a missed disclosure, so report critical fails as a separate count, and calibrate with the client's own scorers. See the QA scorecard template.

2. Productivity metrics

Clients rarely see these, but they decide whether the SLA is affordable.

Average handle time (AHT)

Formula: (talk + hold + after-call work) ÷ handled contacts. Moves it: complexity, system speed, knowledge access, wrap-up design. Trap: push AHT down alone and agents rush, FCR falls and repeat calls rise. Read it beside FCR or conversion. See AHT, after-call work and reducing AHT.

Occupancy

Of available time, the share spent handling contacts. Formula: handle time ÷ (handle time + available idle time) × 100. Moves it: volume against staffing, pooling agents across clients. Trap: sustained high occupancy feeds burnout and attrition. See occupancy.

Utilisation

Of paid time, the share spent on productive work. Formula: productive time ÷ paid time × 100. Moves it: shrinkage, scheduling, idle time. Trap: it is often confused with occupancy. Occupancy divides by available time, utilisation by paid time; write down which each report uses.

Schedule adherence

Whether agents worked the schedule they were given. Formula: minutes in the scheduled state ÷ scheduled minutes × 100. Moves it: break discipline, team-lead follow-up. Trap: perfect adherence to a bad schedule still misses service level; adherence measures the agent, not the forecast. See the workforce management guide.

Shrinkage

The share of paid time in which agents cannot take contacts. Formula: (scheduled hours − available hours) ÷ scheduled hours × 100. Moves it: training, coaching, meetings, leave, sickness, downtime. Trap: client-mandated training and calibration are shrinkage too, and are often left out of the price. Split planned from unplanned. See shrinkage, and size staffing with the Erlang C calculator.

Contacts per hour

Formula: handled contacts ÷ productive hours. Moves it: AHT, dialing mode, idle time. Trap: it rewards short, easy contacts. Compare only within one client and queue.

3. Outbound metrics

On outbound programmes the metrics follow the list down a funnel: dialed, connected, reached, right person, outcome. For depth, see connect rate, abandon rate and handle time.

Connect rate

Attempts that become a connected call, person or machine. Formula: connected calls ÷ dial attempts × 100. Moves it: caller-ID reputation, list quality, time of day, routing. Trap: some teams count voicemail as connected and some do not; state which. See connect rate.

Contact rate

Dialed records that reach a live person. Formula: live contacts ÷ records dialed × 100. Moves it: list freshness, retry strategy, answering machine detection. Trap: per-attempt and per-record denominators give very different numbers; the contract should name one. See contact rate.

Right-party contact (RPC) rate

Contacts that reached and verified the intended person. Formula: verified right-party contacts ÷ records dialed (or ÷ live contacts) × 100. Moves it: data quality, verification scripts, call timing. Trap: unverified pickups marked as RPC inflate it and create compliance risk; audit dispositions against recordings. See RPC.

Conversion rate

Formula: qualifying outcomes (sales, appointments, promises to pay) ÷ right-party contacts × 100. Moves it: offer, script, agent skill, list fit. Trap: agree what counts and when it is confirmed; a sale cancelled in a cooling-off period comes back out.

Abandoned-call rate (outbound)

Live answers where no agent was free, so the person heard silence or a message. Formula: abandoned calls ÷ calls answered by a live person × 100. Moves it: pacing, above all the predictive dialer ratio you set. Trap: it is a regulatory number too. Many markets set a ceiling, a measurement period and a required message, and the rules differ by country and change; run your own ceiling below the rule. Model pacing with the abandon-rate estimator.

4. Commercial metrics

Which of these matters most depends on how the client pays: per hour, per FTE, per transaction or per outcome.

Billable hours versus paid hours

Formula: billable hours ÷ paid hours × 100. Moves it: shrinkage, unbillable idle time, unbilled training. Trap: the contract decides what is billable; payroll counts every paid hour. The gap is your margin leak.

Revenue per seat

Formula: client revenue for the period ÷ production seats (or FTE) on the account. Moves it: price, billable ratio, multi-shift seat use. Trap: a seat used on two shifts earns twice; pick physical seats or FTE and keep it.

Cost per contact

Formula: fully loaded cost (wages, supervision, QA, platform, telecom, facilities) ÷ handled contacts. Moves it: AHT, occupancy, shrinkage, telecom spend. Trap: leaving out supervision and QA makes high-touch accounts look cheap. Model it with the call center cost calculator.

Margin per client

Formula: (client revenue − direct cost − allocated shared cost) ÷ client revenue × 100. Moves it: everything above. Trap: shared costs (managers, planners, IT, platform) need an agreed allocation basis such as seat-hours, or the client that absorbs the most overhead looks the most profitable.

5. People metrics

Attrition

Formula: leavers in the period ÷ average headcount × 100, often annualised from monthly figures. Moves it: occupancy, schedule fairness, team leads, pay. Trap: one annual figure hides early attrition; report first-90-day leavers separately.

Time to proficiency

Formula: days from start date to the first date an agent meets the written standard, for example QA score and AHT inside target for a set number of consecutive weeks. Moves it: training design, nesting, coaching. Trap: without a written standard it becomes a team lead's opinion; define it per client programme.

Reporting per client on a multi-client floor

A floor-wide service level is a number no client signed up to. Four rules keep per-client reporting honest:

  • Separate the data, not only the dashboard. If clients are split only by a report filter, one wrong filter leaks numbers. A multi-tenant platform scopes reports and exports to the tenant below the reporting layer; the BPO scaling guide explains how.
  • One definition sheet per client. Two clients may define the same metric differently, so never add their numbers together.
  • Attribute shared agents by time. Split an agent's hours by time logged on each client's queues and campaigns.
  • Keep a client view and an internal view. Clients see SLA, quality and outcomes; your internal pack adds billable ratio, cost per contact and margin.

A sample weekly client scorecard

Send one per client, weekly, adding columns for target, this week, last week, trend and status (met, at risk, missed). It is a layout, not a set of targets; targets come from each contract.

SectionMetricDefinition to agree with the clientRead it with
AccessService levelThreshold in seconds; short-abandon rule; interval or dailyAbandon rate, worst interval
AccessASAAnswered calls only90th percentile wait
AccessAbandon rateInbound: ÷ offered. Outbound: ÷ live answersService level, pacing
QualityQA score and critical failsScorecard version; sample sizeCalibration result
QualityFCRRepeat-contact window; customer matchingAHT, transfer rate
CustomerCSAT / NPSQuestion scale; response count and rateFCR
OutboundContact and RPC ratePer record or per attemptConnect rate, list age
OutboundConversionWhat counts, and when it is confirmedQA score
DeliveryHours deliveredBillable hours as the contract defines themForecast hours
ActionsCommentaryEach miss: cause, action, owner, dateLast week's actions

The "read it with" column matters most: every metric can be improved by damaging its partner, and the pairing stops that showing up as a win. A missed week is forgiven far more easily when the commentary already names the cause and the fix.

Getting the numbers from a platform, not a spreadsheet

A hand-built scorecard takes hours each week and breaks silently when a definition drifts. The operational half should come from your platform, scoped per client. Here is what DialerBee provides, and what it does not.

  • Per-client separation. Multi-tenant architecture enforces tenant isolation with PostgreSQL row-level security, scopes every report, dashboard and export to the tenant, and tracks usage, minutes and seats per tenant.
  • Fifteen named reports. Reports and Insights includes Queue Performance (wait time and service level per queue), Abandoned Calls, Agent Performance (talk time, wrap time, outcomes and volume), Agent Utilization (how much of a paid hour went to productive states), Campaign Summary, Lead Conversion, Quality Metrics, Cost Analysis (spend by campaign, destination and trunk) and CDR. Reports export as CSV, Excel or PDF.
  • Live floor. An Insights centre with Call Outcomes, Service Level, Dispositions and AMD tabs; a wallboard with live KPIs, agent state grid, leaderboard and MOS card; per-agent disposition events behind every agent number; and supervisor tools to listen, whisper and barge.
  • Workforce, quality, surveys. Workforce management with shifts, time-off, adherence, forecasting and scheduling. QA scorecards with per-campaign weighted templates, calibration and trends by agent, team and campaign. Surveys with CSAT on a 1 to 5 scale and NPS on 0 to 10.

What stays outside: revenue, so revenue per seat and margin per client join your billing data to the platform's cost and usage exports; attrition and time to proficiency need HR records. None of the 15 reports is an FCR report; build FCR from CDR and disposition exports with your agreed window.

Looking at pricing, Starter ($69 per seat per month) includes standard reports; Professional ($129) adds all 15 reports plus Insights, and surveys with CSAT and NPS. Branded PDF reports, multi-tenant management and white-label are listed on Enterprise. Request a demo to see the reports on a demo tenant, or start the 14-day free trial: free AI credit, no card, and self-serve accounts connect your own SIP trunk or PBX. More on DialerBee for BPOs, the BPO dialer and choosing dialer software for a BPO.

General operational guidance, not legal advice. Abandoned-call limits, calling windows and recording rules vary by jurisdiction; confirm them for each client with qualified counsel. DialerBee does not provide legal advice or guarantee regulatory compliance.

Frequently Asked Questions

What are the most important KPIs for a BPO?

The most important BPO KPIs are the ones in the client contract, usually service level, abandon rate, QA score and an outcome such as FCR or conversion. Beside them, track the internal numbers that decide profit: billable versus paid hours, shrinkage, occupancy and margin per client.

What is the difference between BPO SLA metrics and KPIs?

An SLA metric is a KPI the client has written into the contract, with a target and often a penalty or bonus. Every SLA metric is a KPI, but most KPIs stay internal, such as occupancy, shrinkage and cost per contact.

What is a good service level or AHT for a BPO?

There is no universal good number. It depends on the client, channel, complexity and what the client pays for. Set targets per client contract, baseline for a few weeks, and judge the trend against that baseline.

What is the difference between occupancy and utilisation?

Occupancy divides handling time by available time; utilisation divides productive time by paid time. Occupancy shows how busy ready agents are. Utilisation shows how much of what you pay for becomes productive work, so it drives margin.

How do you calculate margin per client in a BPO?

Subtract the client's direct costs and its share of shared costs from its revenue, then divide by that revenue. Allocate shared costs such as supervision, planning, IT and the platform on an agreed basis like seat-hours.

How should a BPO report KPIs to multiple clients?

Give each client its own weekly scorecard, built from data scoped to that client, with its agreed definitions at the top. A multi-tenant platform limits each report and export to one client's data; keep margin and cost in a separate internal pack.

Related reading: Running a BPO Campaign for a Foreign Client

Ready to see DialerBee in action?

Request a 15-minute live demo, or start a 14-day free trial — no slides, no commitment. Calling starts once your own SIP trunk or PBX is connected and approved.

14-day free trial · free AI credit · no card · 11 languages · BYOC · compliance-supporting controls