WFM guideCustomer experience

Contact centre waiting time communication

Q: Should contact centres give customers a wait time estimate?

Yes — but only if the estimate is accurate. Research consistently shows that customers who receive an accurate wait time estimate abandon less often and report higher CSAT than customers who receive no estimate. However, an inaccurate estimate (telling a customer '2 minutes' when the actual wait is 10 minutes) is worse for CSAT than providing no estimate at all — the customer who was told a short wait and experienced a long wait arrives to the agent more frustrated than a customer who waited in silence. The rule is: provide a wait time estimate only if your system can generate it accurately (within ±30% of the actual wait time). Round up slightly rather than down — a customer told '5 minutes' who connects in 4 minutes has a positive experience; a customer told '4 minutes' who waits 7 minutes does not.

Customers who know how long they will wait abandon less often than customers who do not. But an inaccurate estimate is worse than no estimate — a customer told "2 minutes" who waits 10 minutes arrives angrier than one who waited in silence. Queue communication is an operational precision exercise.

Queue communication elements and their impact

Queue entry acknowledgement

What it is

The initial message confirming the customer is in queue: 'Thank you for calling. You are in a queue and an adviser will be with you shortly.'

Impact on abandonment and CSAT

Without this, customers do not know if their call connected successfully. Immediate acknowledgement reduces early abandonment by confirming the customer is being held.

Best practice

Play within 5 seconds of queue entry. Include the contact type they called about if skill-based routing has confirmed it. Do not play a long menu or marketing message before the acknowledgement.

Wait time estimate

What it is

Communicating an approximate wait time: 'Your expected waiting time is approximately 4 minutes.'

Impact on abandonment and CSAT

Accurate estimates reduce abandonment by allowing customers to make an informed decision to wait or call back. Inaccurate estimates (especially underestimates) significantly worsen CSAT for customers who wait longer than told.

Best practice

Only provide if the system generates the estimate from real-time queue state (not a fixed message). Round up by 20%. Repeat at 90-second intervals with an updated estimate. Never use 'just a moment' — it is not a time estimate.

Queue position announcement

What it is

Telling the customer where they are in the queue: 'You are currently number 3 in the queue.'

Impact on abandonment and CSAT

Position announcements are less effective than time estimates because customers cannot convert queue position to time without knowing the average handle time. They are better than silence but worse than accurate time estimates.

Best practice

Use queue position only when the system cannot generate accurate time estimates. Update at each position change. Do not say 'position 1' until the agent is genuinely about to connect — false 'position 1' announcements followed by continued waiting damage trust severely.

Callback offer

What it is

Offering the customer the option to receive a callback when an agent is available, rather than waiting in queue.

Impact on abandonment and CSAT

Callback offers typically reduce abandonment by 25–40% in high-wait environments. Customers who accept a callback have significantly higher CSAT than customers who waited the same duration in queue — the experience of waiting passively at a phone is significantly worse than doing other things while knowing they will be called.

Best practice

Offer callback at the 90-second mark (after the first wait time estimate). Confirm the callback number. Call back within the estimated time — a callback that arrives late reduces its CSAT advantage.

On-hold music and messages

What it is

The audio the customer hears between queue communication messages: music, comfort messages, self-service deflection offers.

Impact on abandonment and CSAT

The type of on-hold content affects the perceived wait duration. Silence feels longer than music. Repetitive or irritating music raises frustration. Comfort messages ('We are sorry to keep you waiting — your call is important to us') have mixed evidence — frequent repetition can feel dismissive.

Best practice

Use unobtrusive music with a consistent tone that matches the brand. Play comfort messages every 60–90 seconds, not more frequently. Keep self-service deflection offers for genuine alternatives — 'Did you know you can do this online?' when the online channel cannot actually complete the task creates significant frustration.

Wait time estimation: the accuracy problem

Most ACD systems calculate wait time estimates using one of two approaches. Knowing which approach your system uses — and its limitations — determines whether the estimate you are providing is genuinely useful or actively harmful.

Queue-based estimate (simple approach)

Formula: (contacts ahead in queue) × (rolling average AHT)

+Simple to calculate; widely available in standard ACD systems

−Assumes average AHT — contacts ahead in queue may be longer or shorter

−Does not account for agents about to become available from contacts already in progress

−Overestimates wait time when contacts ahead are near completion; underestimates when new contacts arrive simultaneously

Accuracy: typically ±30–50% — acceptable for position 5+; unreliable for positions 1–2

Simulation-based estimate (advanced approach)

Simulates the queue progression using real-time agent state and contact-in-progress timing data

+Accounts for contacts already in progress and their estimated remaining duration

+More accurate for short queue positions (1–4 contacts ahead)

−Requires agent state data and real-time AHT modelling — not all systems support this

−Still cannot predict AHT for contacts that have not started

Accuracy: typically ±15–25% — more reliable across all queue positions

Rule of thumb: Add 20% to the system-generated estimate before broadcasting it to customers. The system assumes average AHT on all remaining contacts — actual contacts will vary. A slightly pessimistic estimate that turns out to be correct (or better) builds trust. A slightly optimistic estimate that turns out to be wrong erodes it.

In-queue self-service deflection: when it helps and when it damages

Deflection offers that work

✓Offering the online account portal when the customer can genuinely complete the task they called about online — account statement, address change, password reset

✓Callback offer for a customer who called about a non-urgent matter and has an alternative number to receive the call

✓IVR self-service completion for simple transactional requests (balance, pin, recent transactions) when the IVR can genuinely complete the task without agent involvement

✓SMS or app link to a status page when the customer is calling about a known outage or delay — gives them the information they need without waiting

Deflection offers that damage CSAT

✗Suggesting the online portal for a task the portal cannot actually complete — customer abandons to try online, fails, calls back more frustrated than when they started

✗Suggesting automated IVR self-service for a complex query that requires agent judgment — customer tries the IVR, fails, is redirected to the queue with a longer wait

✗Repeating the same self-service deflection message every 30 seconds when the customer has already indicated they want to speak to an agent

✗Marketing messages (new products, offers, promotions) played while a customer waits to resolve a complaint or service issue — tone mismatch creates significant frustration

Queue communication questions

Should contact centres give customers a wait time estimate?

Yes — but only if the estimate is accurate. Accurate estimates reduce abandonment and improve CSAT for customers who wait. Inaccurate estimates (especially underestimates) are worse than no estimate — a customer told '2 minutes' who waits 10 minutes arrives to the agent more frustrated than one who waited in silence with no expectation set. Only provide a wait time estimate if your system generates it from real-time queue state. Add 20% to the system-generated estimate to account for AHT variability — a slightly pessimistic estimate that is met or beaten builds more trust than an optimistic estimate that proves wrong.