--- title: "Customer Service Quality Assurance Scorecard | Ant Assist" description: "Customer service quality assurance for a small UK team: a scorecard you can run by hand, how to score and sample, and the UK rules on recording calls." url: "https://ant-assist.com/resources/customer-service-quality-assurance/" --- RESOURCES · CUSTOMER SERVICE · QUALITY ASSURANCE # How to run customer service quality assurance For a small team, quality assurance works by hand: score a weekly sample of real conversations against a written scorecard and fix the cause of each miss. Tools our VAs use for this - Updated 23 October 2026 - 12 min read - By Ant Assist [Calculate your savings](https://ant-assist.com/calculator/) · [Book a free 15-min call](https://ant-assist.com/book/) In short Customer service quality assurance means reviewing a sample of real conversations each week against a written scorecard: was the answer correct, complete, on policy, in your tone, and did it solve the problem. Score each conversation, talk through the results with the person who handled it, and fix the cause, whether that is training, a help article or a policy. A small team can run customer service quality assurance by hand with the template below and a spreadsheet. If you record calls to review them, tell callers and follow UK data protection law. On this page Customer Service Virtual Assistant [Book a free 15-minute discovery call](https://ant-assist.com/book/) On this page [Book a call](https://ant-assist.com/book/) · [Calculate your savings](https://ant-assist.com/calculator/) Customer service ## What QA in customer service means QA looks at how a conversation was handled, not just how fast. A reply can arrive in five minutes and still be wrong, rude or against your refund policy. QA catches that; speed numbers do not. It matters just as much once you have [handed support to a virtual assistant](https://ant-assist.com/services/customer-service/). ### QA versus QC Quality control (QC) checks individual pieces of work and corrects them, for example a supervisor approving a refund email before it is sent. Quality assurance is the system around it: the standard, the scorecard, the regular review and the changes that stop the same mistake happening again. A small business usually needs a little of both: QC on high-risk replies while someone is new, and QA on everything after that. ### QA versus KPIs KPIs such as first response time and customer satisfaction tell you that something is going well or badly across all conversations. QA tells you why, by looking closely at a few. Our guide to [customer support KPIs](https://ant-assist.com/resources/customer-support-kpis/) covers which numbers to track; this guide covers checking the work behind them. Part 3 ## Why QA matters Customer service QA matters because speed numbers only show how quickly customers were answered. Only reading the conversations shows whether they were answered well, and that matters at any size of business. ### Benefits for a small business - **Fewer repeat contacts.**: Incomplete answers bring customers back; QA finds the pattern. - **Consistent policy.**: Refunds, returns and goodwill gestures are handled the same way whoever answers. - **Faster training.**: New staff learn from scored examples of their own work. - **Safe delegation.**: You can hand support to someone else and still know the answers are right. - **Early warning.**: One poorly handled complaint is found in the review, not on a review site. ### The link with customer satisfaction Customer satisfaction surveys tell you how the customer felt; QA tells you whether the conversation met your standard. When the two disagree, look closer. A high QA score with low satisfaction often means your policy is the problem, not the person. A low QA score with high satisfaction can mean someone is giving away more than your policy allows. Part 4 ## Building a QA scorecard A scorecard turns "was that a good reply?" into a few questions anyone can answer the same way. Keep it to one page. ### What goes on a scorecard Most scorecards cover the same ground: getting it right, following the process, solving the problem, and how the customer was treated. Add one or two criteria for whatever matters most to your business, such as upselling or a regulated script. ### A scorecard template The weights below are an example for a small business answering by email, chat and phone. Change them to suit your priorities; they should add up to 100. A scorecard template | Criterion | What "met" looks like | Example weight | Score (0, 1 or 2) | Weighted score | |---|---|---|---|---| | **Accuracy** | Every fact, price, date and policy point in the reply is correct | 25 | 2 | 25 | | **Resolution** | The customer's problem is solved, or the next step and timing are clear | 20 | 2 | 20 | | **Policy and process** | Refund, return, escalation and data rules followed; record updated | 20 | 2 | 20 | | **Understanding** | The real question was identified, including anything not asked directly | 15 | 2 | 15 | | **Tone and brand voice** | Polite, warm and in your house style; the customer is addressed by name | 10 | 1 | 5 | | **Clarity** | Easy to read or follow: short, ordered, no jargon, no typos | 10 | 2 | 10 | | **Total** | Every criterion above | **100** | **11 of 12** | **95 out of 100** | **Auto-fail (example):** any reply that shares personal data with the wrong person, promises something your policy forbids, or is rude scores zero overall, whatever the other rows say, and is discussed the same day. ### How to weight and score each item Score each criterion 0 (not met), 1 (partly met) or 2 (fully met). Multiply the score by the weight and divide by 2 to get the weighted score; add the rows for a total out of 100. In the example weights, a reply fully met on everything except tone, which was partly met, scores 95. Weight most heavily what costs you most when it goes wrong: accuracy and policy for most businesses, tone for a premium brand. Keep the scale short. A 0 to 2 scale is quicker to agree on than a 1 to 10 scale, and two reviewers are more likely to give the same score. How one reply scores on the example weights. Set your own. ### What counts as a good score There is no official or industry pass mark that fits every business, so set your own. Score a few weeks of conversations first, then pick a pass mark a little above your current average. Watch the trend and the auto-fails more closely than a single week's score. Act on a low score by reading the conversations behind it: one person's low score usually points to training; everyone's low score on the same criterion points to a process, a policy or a missing help article. Part 5 ## Setting up a QA programme A weekly QA loop, with a monthly check that reviewers agree. Set up a QA programme by choosing a weekly sample you can review properly, deciding who reviews whom, and coaching from the scored conversations. ### Sample size and frequency The right number is the most you can review properly every week without it slipping. As an example, a small team might review five conversations per person per week, picked across channels and days, plus every complaint and every escalation. Choose some at random and some on purpose: low satisfaction scores, long threads, refunds. Review weekly so feedback arrives while the conversation is still fresh. If chat is one of those channels, our overview explains [what live chat customer service is](https://ant-assist.com/resources/what-is-live-chat-customer-service/). ### Reviewers and reviewees In a small business the reviewer is usually the owner or the support lead. Everyone who answers customers is reviewed, including the owner when they step in. Once a month, two people score the same few conversations separately and compare. If the scores differ, tighten the wording on the scorecard. ### Coaching from scores Share the scored conversation, not just the number. Start with what went well, then one thing to change, with a rewritten example. Turn repeated fixes into a reply template or a help article so the improvement lasts; our [customer service email templates](https://ant-assist.com/resources/customer-service-email-templates/) are a starting point. Keep a short log of coaching points and check them in the next review. ### Common mistakes - Scoring only bad conversations, which makes QA feel like punishment. - A scorecard so long that reviews stop happening. - Scoring tone and ignoring accuracy. - Never changing the policy, even when it is the cause. - Reviewing in a batch once a quarter, long after anyone remembers the conversation. Scored on your terms A support VA measured on your own scorecard Your dedicated customer service VA works to the scorecard you share at onboarding, with samples reviewed as often as you choose. [See customer service VAs](https://ant-assist.com/services/customer-service/) · [See your savings](https://ant-assist.com/calculator/) Part 6 ## UK rules on recording and monitoring calls Recording calls for QA in the UK means having a clear reason, telling callers that calls may be recorded and why, keeping recordings secure and keeping them no longer than you need. Phone QA usually means recording calls. Email and chat are already written down, but reviewing them still involves personal data. This is general information, not legal advice. For calls that end as messages, see the [message taking service](https://ant-assist.com/resources/message-taking-service/) guide. ### UK GDPR and informing callers A recorded call or a saved chat with a named customer is personal data, so UK data protection law applies. In practice that means having a clear reason for recording, telling callers that calls may be recorded and why (usually a short message at the start of the call and a line in your privacy notice), keeping recordings secure, and keeping them no longer than you need. If staff conversations are monitored, the ICO's guidance on [monitoring workers](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/employment/monitoring-workers/) says employers must follow data protection law and tell workers about the monitoring. The ICO notes that guidance is under review after the Data (Use and Access) Act, so check the current version. ### Lawful business practice for monitoring The [Telecommunications (Lawful Business Practice) (Interception of Communications) Regulations 2000, regulation 3](https://www.legislation.gov.uk/uksi/2000/2699/regulation/3) allow businesses to monitor or record communications on their own systems for specific purposes, including establishing facts and checking compliance with regulatory or self-regulatory practices. Read the regulation for the full conditions, and do not treat it as a reason to skip telling callers under data protection law. Part 7 ## Manual sampling versus automated QA ### Sampling versus scoring every conversation Reviewing by hand means reviewing a sample. That is fine for a small team, as long as the sample includes the conversations most likely to go wrong. QA software can score every conversation automatically against rules you set, and is usually aimed at teams large enough that a person cannot read a meaningful share. ### When automated QA is and is not enough Automated scoring is good at spotting patterns: a missing greeting, a phrase you have banned, a long silence on a call. It is weaker at judging whether an answer was correct for your business, whether a goodwill gesture was right, or whether a complaint was handled with care. For a small team, a person reading a sample is usually the better start. If you add software later, keep a human review alongside it and check that its scores agree with yours. Part 8 ## Running QA on an outsourced or VA team On an outsourced or VA team, the provider should run its own QA checks, but you still own the standard: you approve the scorecard, review a sample yourself and decide what changes. When a virtual assistant or an outsourced team answers your customers, QA becomes the main way you see the quality of their work. For what this work costs, see [outsourcing customer service cost in the UK](https://ant-assist.com/resources/outsourcing-customer-service-cost/). ### Who owns QA Agree it before they start. The provider should run its own checks and supervision, and tell you how. You should still own the standard: you write or approve the scorecard, you review a sample yourself each week or month, and you decide what changes when a pattern appears. Give the team access to the conversations, the policies and the help articles they are scored against. ### Linking QA to the service promise QA checks the answer; your service level agreement sets how fast it must arrive; KPIs track both over time. Keep them consistent: a reply time promise that forces rushed answers will show up as falling QA scores. Our guide to the [customer service SLA](https://ant-assist.com/resources/customer-service-sla/) covers how to word the promise. If you are still choosing who should handle support, our guide on [how to hire a VA for customer support](https://ant-assist.com/resources/how-to-hire-a-va-for-customer-support/) covers the hiring steps. Customer service ## Quality checks on a customer service virtual assistant An Ant Assist [customer service VA](https://ant-assist.com/services/customer-service/) is one dedicated person, never shared, who learns your policies and tone. VAs are recruited, managed and supervised in Ant Assist's own office in Cairo, not freelancers and not AI, and the service is UK-managed. Share your scorecard during onboarding and agree how often you will review a sample. Access is set up with password managers and can be revoked at any time, and NDAs and data-processing agreements are in place. If quality is not right, the VA can be replaced without restart cost. Cover is UK business hours, 9am to 5pm Monday to Friday, with extended hours and weekend cover at additional cost, from £5 per hour on the introductory rate; the [calculator](https://ant-assist.com/calculator/) shows the monthly cost for your hours. Rated 5.0 on Google. · Trusted by our clients FAQ ## Questions about QA in customer service ### What are the principles of quality assurance in customer service? There is no single official list, and different guides use different principles. The ones most lists share are a clear standard, regular review of real work, consistent scoring, feedback to the person who did the work, and changes that fix the cause. The scorecard and routine in this guide cover all of them. ### How often should QA evaluations be done? Weekly works for most small teams, because feedback lands while the conversation is still fresh and patterns show up quickly. New starters benefit from more frequent review in their first weeks. Complaints and escalations are best reviewed as they happen. ### Can customer service QA be automated? Partly. Software can score every conversation for patterns you define, such as greetings, banned phrases or long waits. Judging whether an answer was correct and fair for your business still needs a person, so most small teams start by hand and add software only when volume makes sampling too thin. ### Does automated QA replace QA reviewers? Automated scoring changes what reviewers spend their time on rather than removing them. The software flags conversations; a person decides what went wrong and coaches the fix. Someone also needs to check that the automated scores match human judgement. ### Should you review email and chat as well as calls? Yes. Written replies carry the same risks as calls, and they are easier to review because the whole conversation is already on the page. Use the same scorecard across channels and adjust only the tone row where your style differs by channel. Didn't find your answer? Browse [more common questions](https://ant-assist.com/help/) in the Help Center. Get Started Answers you can check, every week. A dedicated customer service VA, from £5 per hour. Book a free 15-minute call. [Book a free 15-min call](https://ant-assist.com/book/) · [Calculate your savings](https://ant-assist.com/calculator/) Discovery call · Role brief · VA matched · Onboarded into your tools