Call centre service levels in South Africa: the formulas, normal targets, attrition and switching provider

A service level is a promise about waiting time: a share of calls answered within a set number of seconds. The best known target is 80/20, meaning 80% of calls handled within 20 seconds. It is common, not a law or a standard, and the same words can describe very different service depending on the formula, how abandoned calls are counted and the period the figure is averaged over.

Before you agree a target with an outsourced call centre, settle four things in writing: the formula for each measure, the interval it is measured in, what is excluded, and what happens when it is missed. Then check that the provider has staffed for the target, ask how it measures and manages staff turnover, and agree how you would leave before you need to.

The International Customer Management Institute (ICMI), a contact centre training and research body, defines service level as X% of contacts answered in Y seconds, for contacts that have to be handled as they arrive, such as phone calls and live chat. Its calculation:

• Service level = (calls answered within Y seconds + calls abandoned within Y seconds) ÷ (all calls answered + all calls abandoned).

That is one of several versions in use. The difference is how callers who hang up are treated. Call Centre Helper, an industry publication, runs one set of figures through five formulas: 1,000 calls answered, 60 abandoned, 860 answered within 20 seconds, 20 abandoned within 20 seconds and 40 abandoned after. Four of them:

1. Answered and abandoned within the threshold, over everything answered and abandoned: 83%.
2. Answered within the threshold, over answered calls only: 86%. Callers who gave up have vanished from the sum, which is why the publication does not recommend it.
3. Answered within the threshold, over everything answered and abandoned: 81%. Every abandoned call counts as a miss.
4. Answered within the threshold, over answered calls plus calls abandoned after the threshold: 83%.

The same calls score between 81% and 86% depending on the formula. Against an 80/20 target all four above pass, but a centre hovering near its target can pass or fail on the formula alone. Write the formula into the contract, not just the target.

Two more settings move the number:

• Short abandons. It is common practice to exclude calls abandoned in the first five seconds, on the reasoning that the caller misdialled. Agree the cut-off in seconds and check that the phone system is set to it.
• The period. ICMI says service level loses its meaning when averaged across a day or more, because a quiet afternoon hides a morning of long queues. Centres most often track it in half-hour intervals. In ICMI's survey of 2007, the timeframe most often used in external reporting was monthly (35.6% of respondents), with only 12.5% reporting by interval. Ask for the monthly figure and the half-hour detail behind it, and for a count of half-hours that missed.

Email, letters and call-backs are measured differently: ICMI calls the measure response time, 100% of contacts answered within a stated number of hours or days. Agree whether "answered" means acknowledged or resolved.

Yes. 80/20 is a middle setting. In ICMI's 2007 survey of call centres, 80% of calls in 20 seconds was the most common objective, with 80% in 30 seconds, 90% in 20 or 30 seconds and 85% in 20 or 30 seconds also frequently reported. ICMI's metrics guide puts 80/20 in its "moderate" band, beside 80/30 and 90/60; 90/20, 85/15 and 90/15 in its "high" band; and 100/0 for emergency services. Those sources are from the United States, and we found no published South African figure for service level measured that way.

What South African centres do publish is speed to answer. In ContactBabel's survey of 106 senior contact centre professionals, published in November 2025, respondents reported:

• Average speed to answer: a mean of 27 seconds, 18 in in-house centres and 37 in outsourced operations, with a median of 13 seconds.
• Abandonment: a mean of 4.2%, with a median of 3.0%.
• First-contact resolution: a mean of 86%, 88% in-house and 83% in outsourced operations. The report warns that this may reflect simpler calls rather than better work, and that the measure should not be read alone.

These are what respondents said about their own operations, not audited results. They suggest that a provider proposing a slow answer target for routine inbound calls should explain why.

The right target is a cost decision. A tighter target needs more agents waiting between calls, and you pay for that waiting time. Set it by what the call is for: a sales line or a fraud line earns a tighter target than a general query line, and email can wait longer than a phone call.

A common way to work out staffing is a formula named after the Danish mathematician A. K. Erlang. Traffic is measured in erlangs: the number of calls arriving in a period multiplied by the average time each takes, and one erlang is the most that one line, or one agent, can carry. Erlang's C formula then gives the probability that a caller has to wait, for a given number of agents, and the share who wait longer than any number of seconds you choose. It only works when there are more agents than erlangs; with fewer, the queue grows without end.

A worked example, by our arithmetic using the formula in the ITU's teletraffic handbook. Say 100 calls arrive in a half-hour and each takes 180 seconds to handle, talk, hold and wrap-up included. That is 10 erlangs of work.

• With 12 agents, 64% of calls are answered within 20 seconds, and agents are busy 83% of the time.
• With 13 agents, 79.6% are answered within 20 seconds, at 77% occupancy.
• With 14 agents, 88.8% are answered within 20 seconds, at 71% occupancy.

Three lessons follow. Each agent added or removed moves the result by a large step, so a provider that is one person short in a busy half-hour misses badly. Meeting 80/20 here takes 14 agents, not 13. And the busier you keep agents, the worse the service: occupancy and service level pull in opposite directions.

The formula assumes callers arrive at random and wait as long as it takes. Real callers hang up, so treat its answer as an estimate. It also counts agents logged in and taking calls, not people on the payroll. South African respondents to ContactBabel's survey reported short-term absence of 10.2% of contracted days, and talk time took 56.3% of agents' time, with the rest on wrap-up, other channels, training, administration and waiting. Ask each provider to show the half-hour staffing plan behind its quote, and the allowance it makes for leave, absence, training and meetings.

Each measure below needs a written definition, a target, a reporting period and a consequence.

• Abandonment rate. ICMI gives two versions: all abandoned calls over all calls answered and abandoned, or only calls abandoned after the target time over those plus all answered. Call Centre Helper says an abandonment rate of 2% is generally seen as good and 5% as acceptable. ICMI advises reading abandonment with other measures rather than on its own, because callers hang up for reasons the centre does not control.
• Average speed of answer. Total waiting time divided by calls answered. It hides the long waits inside an average, so ICMI says it should not be the main measure. Use it beside service level, not instead of it.
• Occupancy. For a half-hour, calls handled times average handle time in seconds, divided by the number of agents times 1,800 seconds. ICMI says it should not be a target for individual agents. A provider that promises very high occupancy is promising long queues.
• Average handle time: talk, hold and wrap-up. A handle-time target pushes agents to end calls early, so pair it with a resolution measure.
• First-contact resolution: the issue is solved without the customer coming back within an agreed window, such as a week. Agree how a repeat contact is detected.
• Quality: the share of sampled contacts that pass a scorecard. Hold calibration sessions where your team scores the same calls as the provider's.
• Attrition and tenure: covered in the next section.

National Treasury's bid for its own contact centre asked for reports on contact volume, handle time, talk time, call resolution rate, abandonment rate, cost, client satisfaction and quality assurance, which makes a reasonable starting list.

Tie the targets that matter most to a service credit off the fee, cap the credits so the provider can afford to stay, and add a right to end the contract after repeated misses, so credits are not your only remedy. Exclude only events neither side controls, and define them. Agree that you can see the raw half-hour reports and listen to recordings yourself.

Attrition is the share of agents who leave in a year. ContactBabel defines it as agents leaving over 12 months, voluntary and involuntary, divided by the average headcount over the same months. To compare a partial year, ICMI annualises: leavers divided by average headcount, multiplied by 12 and divided by the number of months measured.

South African respondents to ContactBabel's survey reported mean annual agent attrition of 25%: 19% in in-house centres and 31% in outsourced operations. The spread is wide. The best quarter of respondents reported 8% or less, while the worst quarter reported 36% or more, and 46% or more among outsourced operations. Early leaving was low: 75% of respondents said fewer than 10% of new starters had left by 90 days.

Attrition matters to a buyer because every leaver is replaced by someone who knows less about your customers. Ask each provider:

• Its attrition on comparable accounts over the last year, using the formula above, and whether that counts involuntary exits.
• How many agents on your account in a typical month will be in their first weeks, and whether you are billed for them at the full rate.
• Who pays to recruit and train replacements.
• Whether it will report attrition and tenure on your account every month.

A transition plan moves the work, the knowledge, the data and sometimes the people from one team to another while customers keep calling. The order below works for a move from in-house to a provider, from one provider to another, or back in-house.

1. Read the exit terms you already have: the notice period, what the outgoing provider must do while it runs, and what it must hand back. If the contract is silent, negotiate a short exit assistance agreement before giving notice.
2. Agree the new provider's service levels, reports and pilot measures before the move, using the definitions above.
3. Take stock of what you own: scripts, knowledge base articles, training material, call recordings, customer records, reports, telephone numbers and system accounts. Get each one out in a format the new team can use.
4. Plan the numbers. Under ICASA's rules, geographic numbers, mobile numbers and non-geographic numbers in the 080, 086 and 087 codes can be ported. A block of ten or more numbers ports as a complete range, while 080 and 086 numbers port one at a time, so list every number and book the port for a quiet day.
5. Transfer knowledge: the new team listens to live calls, sits beside the outgoing team where it can, and is trained on your top contact reasons first.
6. Start small. Route one queue, one channel or one share of calls to the new team, compare its results with the outgoing team's, and widen in steps. Keep the outgoing provider able to take the calls back until the new one meets its targets.
7. Agree a settling-in period with closer reporting and a named person on each side, rather than a period in which the service levels do not count.
8. Close out: data returned, copies deleted, access removed and a final report signed.

Three legal points apply to most moves.

• Data. A provider handling your customers' information is usually an operator under POPIA, and section 21(1) requires a written contract that makes it keep the security measures in section 19. Section 14(1) says records must not be kept longer than their purpose needs, subject to exceptions such as a legal or contractual duty. Write return, deletion of every copy, and written confirmation into the exit terms.
• People. The Labour Relations Act's transfer rule, section 197, may move staff with the work when it changes hands as a going concern, including from one provider to the next. Ask your attorney before you give notice.
• Regulated buyers. For insurers, Joint Standard 1 of 2024 requires a reasonable termination period in the contract. It counts the time needed to find another provider or bring the work back in-house as a factor in materiality, and requires the Authorities to be told within 7 days of ending a material arrangement, including how the work will be done afterwards.

Plan the handover at the start of a contract, not the end. National Treasury's contact centre bid scheduled a closeout and handover report for month 36 of a three-year contract. This is not legal advice; have an attorney check your own arrangement.

IAMX offers business process outsourcing of all types through BPO partners: customer support, back-office work, collections and recoveries, and managed teams that scale with you, behind one accountable front door to a vetted network of specialist partners. The business process outsourcing page sets out what is included.

• How do you calculate call centre service level?

  Divide the calls answered within your target time by the calls offered in the same interval, and agree how abandoned calls count. ICMI's version is calls answered plus calls abandoned within Y seconds, divided by all calls answered plus all calls abandoned. Measure it by half-hour, not as a daily or monthly average.

• What is the formula for an SLA with BPO partners?

  There is no single formula. For each measure, the SLA should state the calculation, the interval, the exclusions such as calls abandoned in the first five seconds, the target, the reporting and the service credit for a miss. The same calls can score between 81% and 86% depending on how abandoned calls are counted, so write the formula down.

• What service level should I expect from an outsourced call centre?

  80/20 is the most common target and is normal: it sits in the middle of ICMI's range; 90/20 is in its high band. South African respondents to ContactBabel's survey reported a mean speed to answer of 37 seconds in outsourced operations and 18 in in-house centres. Choose the target by what each call type is worth, and check the staffing plan behind it.

• What is the call centre attrition rate in South Africa?

  South African respondents to ContactBabel's survey reported mean annual agent attrition of 25%, with 31% in outsourced operations and 19% in in-house centres. It is agents leaving in a year divided by the average headcount.

• What should a transition plan between BPO partners include?

  What is moving and who owns it, the exit terms with the outgoing provider, data return and deletion, porting of telephone numbers, knowledge transfer and training, a phased move of calls with the old team as a fallback, closer reporting while the new team settles, and a signed closeout.

• How do I switch BPO partners without disrupting customers?

  Agree the new provider's service levels before you give notice, keep the outgoing provider running at the same standard through the notice period, move calls in stages, and keep a way to send them back until the new team meets its targets. Port your numbers on a quiet day, and ask an attorney whether section 197 of the Labour Relations Act moves any staff with the work.

• ICMI (International Customer Management Institute), 2007 Call Center KPI/Performance Metrics Survey Report: In ICMI's 2007 Call Center KPI and Performance Metrics Survey, the most common service level objective for calls handled by live agents was 80% of calls handled in 20 seconds; 80% in 30 seconds, 90% in 30 seconds, 90% in 20 seconds and 85% in 20 or 30 seconds were also frequently reported. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics defines service level as X% of contacts answered in Y seconds, for inbound contacts that must be handled when they arrive, calculated as calls answered plus calls abandoned within Y seconds, divided by calls answered plus calls abandoned, read from ACD reports. Checked 12 October 2026.
• Call Centre Helper (industry publication): Call Centre Helper works one set of figures (1,000 calls answered, 60 abandoned, 860 answered within 20 seconds, 20 abandoned within 20 seconds and 40 after) through five service level formulas, among them: answered plus abandoned within the threshold over all answered and abandoned gives 83%; answered within the threshold over answered calls gives 86%, which it does not recommend because it ignores abandoned calls; answered within the threshold over all answered and abandoned gives 81%; answered within the threshold over answered plus calls abandoned after the threshold gives 83%; and a fifth, which counts calls abandoned before a shorter cut-off as positive, gives 82%. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics defines response time as 100% of contacts answered in N days, hours or minutes, for outbound contacts and inbound contacts that can be handled later, such as email. Checked 12 October 2026.
• Call Centre Helper (industry publication): Call Centre Helper gives the call abandon rate as calls offered minus calls handled, divided by calls offered, times 100; says it is common practice to exclude calls abandoned in the first five seconds; and says an abandonment rate of 2% is generally seen as good and 5% as acceptable. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics says service level and forecast accuracy lose their meaning when averaged across a day or more, and that accessibility by live agent is most often tracked in half-hour intervals (31% of centres), with 15-minute intervals second (29%). Checked 12 October 2026.
• ICMI (International Customer Management Institute), 2007 Call Center KPI/Performance Metrics Survey Report: In ICMI's 2007 survey, asked for the main timeframe of the service level used in external reporting, 35.6% of respondents said monthly, 32.7% daily, 14.9% weekly and 12.5% by interval. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics gives example service level objectives in four bands: emergency services 100/0; high 90/20, 85/15 or 90/15; moderate 80/20, 80/30 or 90/60; modest 70/60, 80/120 or 80/300. Checked 12 October 2026.
• ContactBabel: ContactBabel's 2026 Inner Circle Guide to South African Contact Centres, published November 2025, is based on a questionnaire answered by 106 senior customer contact professionals between August and October 2025: 64 from in-house contact centres and 42 from outsourced or BPO operations. The figures are in the free report offered for download on this page. Checked 12 October 2026.
• ContactBabel: South African respondents to ContactBabel's 2026 survey reported a mean speed to answer of 27 seconds (18 in in-house centres, 37 in outsourced or BPO operations) and a median of 13 seconds (12 in-house, 15 outsourced), against means of 116 seconds in the UK and 99 in the US. The figures are in the free report offered for download on the publisher's page. Checked 12 October 2026.
• ContactBabel: South African respondents to ContactBabel's 2026 survey reported a mean call abandonment rate of 4.2% (4.4% in in-house centres, 4.0% in outsourced or BPO operations) and a median of 3.0%, against means of 5.9% in the UK and 8.9% in the US. The figures are in the free report offered for download on the publisher's page. Checked 12 October 2026.
• ContactBabel: South African respondents to ContactBabel's 2026 survey reported a mean first-contact resolution rate of 86% (88% in in-house centres, 83% in outsourced or BPO operations); the report says South African centres may deal with a greater share of simpler calls rather than necessarily being more effective, and that the metric should not be viewed in isolation. The figures are in the free report offered for download on the publisher's page. Checked 12 October 2026.
• International Telecommunication Union (ITU-D Study Group 2), Teletraffic Engineering Handbook, draft of 20 June 2001: The ITU-D Teletraffic Engineering Handbook says the unit of traffic intensity is the erlang, named in 1946 by the CCIF (a predecessor of ITU-T) after the Danish mathematician A. K. Erlang (1878 to 1929), and that offered traffic A equals the mean number of calls offered in a unit of time multiplied by the mean service time; a line can carry one erlang at most. Checked 12 October 2026.
• International Telecommunication Union (ITU-D Study Group 2), Teletraffic Engineering Handbook, draft of 20 June 2001: The ITU-D Teletraffic Engineering Handbook gives Erlang's C formula for the delay system M/M/n (random Poisson arrivals, exponential service times, n servers and unlimited waiting places): the probability that a caller has to wait, which holds only when the offered traffic A is less than n; and the probability of waiting longer than t as that probability times e to the power of minus (n minus A) times t divided by the mean service time. Checked 12 October 2026.
• International Telecommunication Union (ITU-D Study Group 2), Teletraffic Engineering Handbook, with IAMX's arithmetic: Applying the Erlang C formula in the ITU-D handbook to 100 calls in a half-hour (1,800 seconds) with an average handle time of 180 seconds, which is 10 erlangs: 12 agents answer 64% of calls within 20 seconds, with an average speed of answer of about 40 seconds and occupancy of 83%; 13 agents answer 79.6%, about 17 seconds and 77%; 14 agents answer 88.8%, about 8 seconds and 71% (IAMX's arithmetic). Checked 12 October 2026.
• ContactBabel: South African respondents to ContactBabel's 2026 survey reported a mean short-term agent absence rate of 10.2% of contracted days: 13.6% in in-house centres and 6.1% in outsourced or BPO operations. The figures are in the free report offered for download on this page. Checked 12 October 2026.
• ContactBabel: ContactBabel's 2026 survey found talk time on inbound calls took 56.3% of a South African agent's time, with the rest spread across wrap-up, email and chat, idle time, administration, training and other work. The figures are in the free report offered for download on this page. Checked 12 October 2026.
• National Treasury: National Treasury's bid NT003-2021 for contact centre services, a three-year contract, asks for reports covering contact volume, handle time, talk time, call resolution rate, abandonment rate, cost, client satisfaction and quality assurance, says the successful bidder will enter into a service level agreement, and schedules a closeout and handover report for month 36. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics gives two calculations of abandoned contacts: all calls abandoned divided by all calls abandoned plus all calls answered, or calls abandoned after the objective divided by calls abandoned after the objective plus all calls answered; it says abandonment is best used with other metrics rather than on its own. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics calculates average speed of answer as total delay divided by the total number of answered contacts, and says it shows only the average experience and should not be the primary measure of accessibility. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics calculates occupancy for a half-hour as contact volume times average handle time in seconds, divided by the number of agents times 1,800 seconds, and says occupancy should not be a goal or standard for individual agents. Checked 12 October 2026.
• ContactBabel: South African respondents to ContactBabel's 2026 survey reported mean annual agent attrition of 25%: 19% in in-house centres and 31% in outsourced or BPO operations, counting voluntary and involuntary leavers over 12 months against average headcount. The figures are in the free report offered for download on this page. Checked 12 October 2026.
• ICMI (International Customer Management Institute), Guide to Contact Center Metrics, first edition, 2015: ICMI's Guide to Contact Center Metrics calculates turnover as the number of agents leaving the job divided by the average actual number of agents in the period, multiplied by 12 divided by the number of months in the period, counting voluntary and involuntary leavers. Checked 12 October 2026.
• ContactBabel: In ContactBabel's 2026 South African survey, the first quartile (best performance) of annual agent attrition was 8% and the third quartile 36%, with a third quartile of 46% among outsourced or BPO operations. The figures are in the free report offered for download on the publisher's page. Checked 12 October 2026.
• ContactBabel: In ContactBabel's 2026 South African survey, 75% of respondents reported that fewer than 10% of new starters had left by the 90-day mark, 23% that 11% to 25% had left, and 2% that a higher share had left. The figures are in the free report offered for download on the publisher's page. Checked 12 October 2026.
• Independent Communications Authority of South Africa (Government Gazette 41949): ICASA's Number Portability Regulations, 2018, regulation 4: number portability applies to geographic numbers, to non-geographic numbers in the 086, 080 and 087 codes, and to mobile numbers. Checked 12 October 2026.
• Independent Communications Authority of South Africa (Government Gazette 41949): Number Portability Regulations, 2018, regulations 5 and 6: a block of ten or more numbers is ported as a complete range, with no individual numbers ported out of it and none of its numbers assigned to another subscriber; non-geographic numbers in the 080 and 086 codes may only be ported as an assigned individual number. Checked 12 October 2026.
• Government of South Africa, Government Gazette: Section 21(1) of the Protection of Personal Information Act requires a responsible party, in terms of a written contract, to ensure that its operator establishes and maintains the section 19 security measures. Checked 9 October 2026.
• Government of South Africa, Government Gazette: Section 19 of the Protection of Personal Information Act requires a responsible party to take appropriate, reasonable technical and organisational measures to secure personal information against loss, damage and unlawful access. Checked 9 October 2026.
• Department of Justice and Constitutional Development (Act 4 of 2013): POPIA section 14(1): records of personal information must not be retained any longer than is necessary for achieving the purpose for which the information was collected or subsequently processed, unless retention is required or authorised by law, reasonably required for lawful purposes related to the responsible party's functions or activities, required by a contract, or consented to. Checked 12 October 2026.
• Department of Employment and Labour: Section 197 of the Labour Relations Act, as substituted in 2002, applies to the transfer of the whole or a part of any business, trade, undertaking or service as a going concern, and automatically substitutes the new employer in all existing contracts of employment. Checked 9 October 2026.
• Prudential Authority and Financial Sector Conduct Authority: Joint Standard 1 of 2024, Outsourcing by Insurers, requires a written contract for a material function that covers service levels, performance reviews, access to information, subcontracting, intellectual property, continuity, termination and a reasonable termination period. Checked 9 October 2026.
• Prudential Authority and Financial Sector Conduct Authority: Joint Standard 1 of 2024 (Outsourcing by Insurers): clause 7.1(c) counts the time taken to find an alternative provider or bring the function back in-house as a factor in materiality; clause 9.1(t) requires a reasonable termination period that allows the insurer's contingency plans to be implemented; clauses 10.6 and 10.8 require an insurer to notify the Authorities within 7 days of terminating a material arrangement, explaining among other things how the function will be performed afterwards. Checked 12 October 2026.