WFM guideQuality assurance

Contact centre quality framework

Q: How many contacts should be monitored per agent per month for QA?

The minimum QA sample size for statistical validity depends on how the data will be used. For individual performance management: a minimum of 8–10 contacts per agent per month is required before individual scores are meaningful for development purposes. Below 8 contacts, sample noise means a single poor contact can move a score by 10–15 percentage points — making individual score fluctuation misleading. FCA regulated firms typically require a minimum of 10% of contacts monitored for certificated functions. For team-level insight: 3–5 contacts per agent per month provides sufficient data for team trend analysis and knowledge gap identification. For operational improvement analysis: 5–10% of total contact volume should be sampled across contact types and agent cohorts. Smaller contact centres (under 25 agents) should aim for 100% of agents monitored at least monthly, even if only 2–3 contacts per agent.

Q: What is a QA calibration session in a contact centre?

A QA calibration session is a structured meeting where quality assessors independently score the same contact and then compare results. The purpose is to identify and reduce assessor variation — ensuring that the same contact receives the same score regardless of which assessor evaluates it. Without calibration, assessor drift means an agent's QA score depends partly on who assessed them, creating unfairness and undermining the validity of quality data for performance management. Best practice calibration: assess 3–5 contacts independently, compare scores in a facilitated session, discuss disagreements, update the scoring guide with clarifications. Run calibration monthly at minimum; more frequently when new assessors join or when the framework is updated. Calibration scores are not used for performance purposes — they are for assessor alignment only.

A QA framework is not a scoring form — it is an operational standard. The form is how the standard is measured; the framework is what is decided about what good looks like, how it is weighted, and how the data is used. Get the framework design wrong and QA becomes a compliance exercise that agents resent and coaches cannot use. Get it right and it becomes the primary insight engine for agent development, knowledge gap identification, and operational improvement.

The five quality categories

Regulatory compliance

Binary fail — not percentage-weighted

What is scored

Mandatory disclosures, FCA COBS obligations (regulated products), GDPR data handling, consumer credit act requirements, anti-money laundering script requirements, vulnerability flags. Any failure in this category = contact fail regardless of score in other categories.

Why this weight

Regulatory requirements are not a matter of degree — partial compliance is non-compliance. A contact that scored 92% on all other dimensions but failed to deliver a mandatory FCA disclosure has failed absolutely, not scored 92%.

Examples

FCA: providing regulated advice without appropriate disclosure. GDPR: sharing account information without identity verification. Consumer credit: omitting required credit information in a phone sale. Safeguarding: failing to flag a distress signal to a supervisor.

Process compliance

20–25% of quality score

What is scored

Following the correct call structure (opening, identification and verification, handling, closing), adhering to transfer protocols, correct wrap code selection, accurate case notes, SLA adherence within the call (escalation timing).

Why this weight

Process compliance is not customer experience — customers often prefer agents who deviate from the script to be more human. However, process compliance drives data quality, enables accurate reporting, and protects the business operationally. Weight it separately from customer experience criteria.

Examples

Correct ID&V before accessing account. Accurate wrap code selection (affects reporting, billing, root-cause analysis). Case notes completed before next contact. Escalation protocol followed within defined timeframe.

Accuracy

25–30% of quality score

What is scored

Correct information provided. Product details accurate. Policy information correct. Amounts and dates stated correctly. Next steps communicated accurately.

Why this weight

Accuracy directly drives FCR. An inaccurate response does not resolve the contact — it generates a repeat contact (sometimes with a different agent who now has the additional overhead of correcting the prior information). Accuracy is the primary driver of repeat-contact volume and should be weighted accordingly.

Examples

Correct premium quoted. Right renewal date communicated. Correct excess amount stated. Correct eligibility decision. Accurate delivery timeframe.

Customer experience

25–30% of quality score

What is scored

Empathy and rapport, active listening, de-escalation capability, communication clarity, avoiding jargon, managing difficult conversations, personalisation (using the customer's name appropriately), signposting next steps.

Why this weight

Customer experience drives CSAT and loyalty. It is a distinct dimension from accuracy and process compliance — an agent can be accurate and process-compliant but cold and robotic. A separate customer experience category prevents process-heavy frameworks from ignoring the interpersonal dimension of customer service.

Examples

Acknowledges the customer's frustration before moving to resolution. Uses plain language rather than policy jargon. Confirms understanding before acting. Closes the call with a clear summary of actions and next steps.

Resolution and ownership

15–20% of quality score

What is scored

Did the agent take ownership of the contact? Did they attempt to resolve it rather than transferring unnecessarily? Did they use available tools and knowledge before escalating? Was the contact resolved in one interaction (FCR)?

Why this weight

Ownership and resolution are distinct from accuracy (what was said) and customer experience (how it was said) — this category measures whether the agent tried to resolve. An agent who gives accurate, empathetic information but always transfers rather than resolving scores low here. This dimension incentivises FCR behaviour.

Examples

Attempted resolution before escalation. Used knowledge base rather than holding for supervisor. Confirmed resolution with the customer before closing. Avoided unnecessary transfer to another skill group.

Sample size and calibration

Sample size requirements

Individual performance management

8–10 contacts/agent/month

Below 8, score noise from a single poor contact makes individual scores misleading for development or appraisal.

Team trend analysis / knowledge gap identification

3–5 contacts/agent/month

Sufficient for team-level patterns. Not reliable enough for individual performance decisions.

FCA certificated function minimum (regulatory)

10% of contacts monitored

Regulatory minimum for FCA-regulated customer-facing roles. Some FCA permissions specify higher requirements.

Small team (<25 agents)

100% of agents monitored monthly

Even if only 2–3 contacts per agent, ensuring every agent receives at least some monthly feedback is more important than statistical validity at this scale.

Calibration session design

Without calibration, assessor variation creates unfairness — the same contact receives different scores from different assessors. Calibration reduces assessor drift.

Select

Choose 3–5 contacts from the last monitoring period. Include at least one ambiguous or borderline case.

Score independently

All QA assessors score the selected contacts without seeing others' scores.

Compare and discuss

Facilitated discussion of disagreements. Where do assessors give different scores? Why?

Update the scoring guide

Ambiguous situations that produce disagreement should be documented in the scoring guide as worked examples.

Frequency

Monthly minimum. More frequently when: new assessors join, framework is updated, team score variance is above 5pp between assessors.

The QA-to-coaching workflow

What happens after a contact is assessed?

Assessment completed and scored

QA assessment submitted in QA platform (or spreadsheet). Score by category recorded. Specific positive and development feedback noted at contact level.

Agent receives feedback

Agent reviews the assessment, can listen to the contact. In automated QA platforms, this is immediate. In manual systems, within 24–48 hours. Agent can acknowledge or raise a dispute.

TL reviews and prioritises

TL receives weekly QA summary for their team. Identifies agents below threshold (typically below 70% on any category). Schedules 1:1 coaching sessions for that week.

Coaching session (weekly or bi-weekly)

1:1 session between TL and agent. Review 1–2 assessed contacts together. Focus on one or two development areas per session, not everything at once. Agree one specific action the agent will take before next session.

Aggregate QA data for operations

Weekly QA trends by team, by category, and by contact type reviewed by the Operations Manager. Patterns (multiple agents scoring low on the same category) escalated to knowledge management or training, not individual performance management.

Quality framework questions

How many contacts should be monitored per agent per month for QA?

For individual performance management: 8–10 contacts/agent/month minimum (below 8, sample noise makes individual scores misleading). For team trend analysis: 3–5/agent/month. FCA certificated functions: 10% of contacts (regulatory minimum). Small teams under 25 agents: 100% of agents monitored monthly, even if only 2–3 contacts per agent.

What is a QA calibration session in a contact centre?

A structured meeting where all QA assessors independently score the same 3–5 contacts, then compare and discuss results. The goal is to identify and reduce assessor variation (ensuring the same contact receives the same score from different assessors). Run monthly minimum. Calibration scores are not used for agent performance — they are for assessor alignment. Disagreements are documented in the scoring guide as worked examples.