The essentials on what quality monitoring covers
- Quality monitoring is the systematic evaluation of exchanges between advisors and customers, against a written standard, in order to reach a decision. This definition is widely agreed. It is correct, but it leaves out what matters most: which decision, and who makes it.
- Four disciplines circulate under this label: scoring advisors, checking regulatory compliance, listening to the voice of the customer, coaching a sales force. They do not have the same users, the same scorecards or the same standards of proof.
- Four verbs set them apart: score, check, count, correlate. Scoring produces an average per advisor. Compliance judges each call, and a breach cannot be offset. The voice of the customer counts call reasons. Sales coaching looks for what precedes a won sale.
- Speech analytics, AI, analysis of 100% of calls: these are techniques, not disciplines. "Automating quality monitoring" does not say which of the four is being automated.
- A specification that does not name its discipline receives answers that cannot be compared. Writing down the discipline targeted, the unit judged and the proof required is enough to sort the offers.
A definition, and what it leaves unsaid
In early October 2026, we asked five AI assistants, in French, what quality monitoring is in a customer contact centre. All five answers describe the same thing: an evaluation scorecard, a sample of calls listened to, feedback to the advisor, coaching, and often automation as the horizon. Software comparisons published in English say much the same thing.
This definition is not wrong. Quality monitoring is indeed the systematic evaluation of exchanges between advisors and customers, against a written standard, in order to reach a decision. But it leaves aside the two words that matter: which decision, and who makes it.
For behind the label, four different teams are looking for a "quality monitoring tool". The quality manager wants to score their advisors and help them improve. The compliance officer has to prove, call by call, that an obligation was met. The head of customer relations wants to know why customers call and what makes them call back. The sales director wants to understand what wins a sale. All four listen to calls, all four talk about quality. They are not doing the same job.
The confusion is paid for at the point of purchase. A specification headed "quality monitoring solution" receives answers from tools each designed for one of these disciplines, often excellent in their own, and impossible to compare with one another. The misunderstanding often surfaces late: in a demo, or after signing, when the chosen tool answers very well a question nobody had asked.
Four disciplines under one label
| Scoring advisors | Checking compliance | Listening to the voice of the customer | Coaching a sales force | |
|---|---|---|---|---|
| Question asked | Did the advisor handle this exchange well? | Was the obligation met, and can I prove it? | Why do customers call, and what makes them call back? | What wins a sale, and how can it be passed on? |
| Who uses it | Quality manager, supervisors, advisors | Compliance, internal control, audit | Customer relations leadership, product, marketing | Sales leadership, sales managers |
| Unit judged | The advisor, over a period | The call, linked to a contract | The call reason, across a population of customers | The salesperson and their opportunity |
| Operation | Scoring: a weighted average | Checking: yes or no, call by call | Counting: frequencies and how they change | Correlating: behaviours and a sales outcome |
| What carries weight | Fairness | Defensibility | Representativeness | Adoption |
| What comes out of it | A score, a debrief, a training plan | A list of breaches to remedy, an audit trail | Contact drivers, pain points, weak signals | Key moments, advice, a reading of the pipeline |
None of these disciplines is more legitimate than the others, and a single call can serve all four. But each has shaped its own scorecards, its own indicators and the tools that serve it.
Scoring advisors
This is the most common meaning of the term in contact centres, the one English speakers call quality assurance. Exchanges are evaluated to judge how an advisor handled them, and feedback, training, and sometimes a share of variable pay follow from that.
Its unit is the person. A single call score says little; it is the average over a month, compared with the team's, that supports a decision. Hence a weighted scorecard, where a missed criterion costs points that other criteria can offset. This offsetting is not a flaw: a call that is weak on one point can be good overall.
What carries weight is fairness. A score that affects an employee's career must be the same whoever the evaluator, open to contest by the advisor, and announced in advance. In France, no information concerning an employee personally may be collected by a system that has not been brought to their attention beforehand (article L1222-4 of the French Labour Code), and the comité social et économique (CSE), the French works council, is consulted on the means used to monitor employees' activity (article L2312-38). This is the discipline addressed by our article on the procedural and relational approaches: the question there is not whether to score, but what to score.
Checking regulatory compliance
Compliance monitoring resembles scoring: a scorecard, criteria, calls. But it does not answer the same question. It does not seek to know whether the advisor was good. It seeks to know, for each call, whether an obligation was met: the commercial nature of the call announced, consent obtained, the right of withdrawal mentioned, the customer's demands and needs collected before recommending a contract.
Its unit is the call, and often the contract that results from it. Its operation is binary: the obligation is either met or it is not. Above all, a breach cannot be offset. A sale with no mention of the right of withdrawal is not redeemed by a faultless greeting. A team with a 92% average score can have one sale in ten without that mention: the average dilutes those sales instead of pointing them out. COPC, which publishes performance standards for contact centres, recommends for this reason measuring "compliance critical" quality separately, alongside the quality that matters to the customer and the quality that matters to the business, because combining these dimensions in a single score weakens it (May 2024).
What carries weight is defensibility. In insurance in France, the burden of proof lies with the distributor, and the decisions of the Enforcement Committee of the ACPR, the French banking and insurance supervisor, show how an inspector proceeds: they listen to 20 to 30 calls of their own choosing, compare them with the advice forms, and that is enough to establish a breach. To face them, you need an audit trail that states which calls were checked, what was found and on which passage, who corrected what and when, and what was done about the breaches. Our article on proving the duty to advise details the four documents of that file.
In the example below, taken from a fictitious call, a sale scored 83/100 does not state the cooling-off period. The finding quotes the customer's question, points to the two passages of the call, and the transcript shows the answer that took the place of the information owed.
This discipline has another particularity: its data does not necessarily live alongside the scoring data. For recordings that serve at once as proof of a contract, as training material and for evaluation, the CNIL, the French data protection authority, recommends a physical or logical separation of the databases. The regulator itself therefore separates the two disciplines.
Listening to the voice of the customer
The third discipline judges no one. It uses calls as a survey that would need no questionnaire: why customers call, what they express, what makes them call back, what signals a departure. It is often filed under speech analytics, after the technique it uses.
Its unit is the call reason, across a population of customers. Its operation is counting: how many calls for a given reason, up or down, on which product, in which region. Its scorecard is not one in the scoring sense: it is a list of reasons and signals, which must be able to take in those nobody had anticipated.
What carries weight is representativeness: each finding rests on excerpts that can be reread, and on volumes that show whether a case is isolated or widespread. Its limit is well known: customers who leave without calling leave no trace. We described it in the voice of the customer built from calls and in detecting weak signals.
Its typical mistake is to attribute to the advisor what stems from the product or the process. A spike in calls about an unreadable bill is no advisor's fault, and adding it to their scorecard will fix nothing.
Coaching a sales force
The fourth discipline comes from sales. People there talk of sales coaching, conversation intelligence or revenue intelligence. The question is no longer "was the exchange compliant?" but "what makes us win?": discovering the need, handling objections, the next step agreed with the customer.
Its unit depends on how you sell. In telesales, where one call often carries an entire sale, the call remains a good unit, and the discipline comes close to scoring. In B2B sales, where an opportunity plays out over several meetings, often by video call, the unit becomes the opportunity, and a score per exchange no longer says much. Its operation is correlation: which behaviours precede a won sale, which make no difference to the outcome.
What carries weight here is adoption. A coaching tool that salespeople do not open serves no purpose, however accurate it may be. Hence short outputs, focused on the next action, and linked to the CRM.
Why a tool does not carry over from one discipline to another
A tool is judged on what it was designed to do. Transposing it into another discipline means asking it for something it is not trying to do. Four cases come up often.
A scoring tool used for compliance. This is the most frequent case, because the two disciplines look alike. Compliance becomes one criterion among others, weighted, blended into an average score. You get a score, but not the list of sales to remedy.
A sales coaching tool facing a compliance check. It was designed to bring out the moments that win, on the exchanges that are worth it, and to relate behaviours to an outcome. A compliance check demands something else: checking every call in scope, criteria that cannot offset one another, the passage that supports each finding, a record of every human correction, separate retention. If a coaching tool can do all of that, so much the better: have it demonstrated, line by line, on your own calls. It is not the task these tools were designed for.
A contact centre scoring tool facing a B2B sales team. It evaluates each exchange against a scorecard, with a score per call and per advisor. That is what is needed for numerous, comparable calls. For a complex sale that plays out over five meetings, the score of a single meeting says little about the opportunity, and the salesperson expects advice on what to do next rather than a score on the past.
A voice of the customer tool used to score advisors. It counts reasons across thousands of calls. Its figures describe a population, not a person. Attributing to an advisor the volume of dissatisfaction expressed on their calls means blaming them for the flaws of the product, the pricing or the process they inherit.
The vocabulary that feeds the confusion
Part of the misunderstanding comes from the words themselves, which French-speaking contact centres often borrow from English. Some name a discipline, others a technique, others again a family of software.
| Term | What it refers to | Nature |
|---|---|---|
| Quality monitoring | The evaluation of exchanges against a standard, whatever its use | The common label |
| Quality assurance (QA) | Scoring advisors against a scorecard | A discipline: scoring |
| Quality management (QM) | The full cycle: scorecard, evaluation, calibration, feedback to the advisor, corrective action | A process, often a module in a software suite |
| Compliance monitoring | Checking regulatory obligations, call by call | A discipline: checking |
| Voice of the customer | Listening to what customers express, to understand their requests and their pain points | A discipline: counting |
| Sales coaching, revenue intelligence | Analysing sales conversations to improve sales performance | A discipline: correlating |
| Speech analytics | The transcription and automated analysis of call content | A technique, serving all four |
| Conversational analysis | AI analysis of conversations, transcription included | A family of software |
Our comparison of conversational analysis software sorts these tools into four product families. This article sorts them by discipline: the two readings complement each other, since one product can serve several disciplines.
Even in our own product, one word does double duty. In Raisetalk, a call's score is called the "Compliance score": compliance with the scorecard, whether its criteria concern a legal obligation or the quality of the greeting. The usage is common in contact centres, and it is one more reason to state, in a specification, which compliance you mean.
Automation is not a discipline
Rephrased by a buyer, the question often becomes: "which software can automate quality monitoring?". It mixes a technique and a discipline. Automated analysis of every call changes coverage and cost, and that is no small thing: with 30 calls listened to per month, an 80% compliance rate in fact lies between 63% and 91%, as our article on sampling shows. But it does not change the question being asked. You can analyse 100% of calls to score, to check, to count or to correlate, and the tool that does one very well does not necessarily do the others.
That is also why the usual presentation as a maturity ladder, from manual listening to automated evaluation by way of assisted listening, is partly misleading. It describes a progression in technique, not in discipline. A contact centre can have automated the scoring of all its advisors and have no compliance set-up capable of withstanding an inspection. A bank can check every one of its sales and know nothing of the reasons that make its customers call back. There is not one maturity ladder, there are four.
The cycle, on the other hand, is shared: a written standard, an observation, a decision, an action, a check. Each discipline fills it in its own way, and it is the decision in view that says which.
One call, several readings
The four disciplines do not always live in separate teams. In an insurance telesales operation, the same call is at once a sale to improve on, an advisor's work to evaluate and an obligation to prove. The temptation is then to put everything into a single scorecard and a single score.
Three rules avoid mixing what must remain distinct:
- One output per discipline. The advisor's score, the list of compliance breaches and the counts of call reasons are not read in the same place, nor by the same people. A regulatory criterion can appear in the advisor's scorecard, but its failure is read separately, as a number of calls concerned.
- No offsetting for compliance. A missed obligation must trigger an alert and action on the call concerned, not just a few points off an average.
- Access rights and retention periods specific to each use. Internal control does not need the same access as supervisors, and the database used as evidence does not follow the same retention periods as the one used for training.
Writing a specification that does not mix the disciplines
1. Name the discipline, and the decision-maker
Before any list of features, write down which discipline the tool must serve, and who will make the decisions it informs. If the answer is "several", write them all down, with a decision-maker for each. A specification that serves two disciplines must say so, so that each vendor responds on both.
2. For each discipline, write down the question, the unit and the proof
Three lines are enough: the question the tool must answer, the unit it judges (an advisor, a call, a call reason, an opportunity), and what must carry weight with the decision-maker. These three lines are often enough to rule out offers aimed at a discipline other than yours.
3. Have it demonstrated, discipline by discipline
The points to see working in a demo, on your own calls, are not the same for each discipline:
| Discipline | What to have demonstrated |
|---|---|
| Scoring advisors | Calibration between the analysis and your evaluators, criteria that are N/A when the situation does not arise, contestation by the advisor, the debrief |
| Checking compliance | Checking every call in scope, the passage that supports each finding, a record of every human correction (author, date, reason), an alert when a sensitive criterion fails, separate access rights and retention |
| Listening to the voice of the customer | Reasons that emerge without having been listed in advance, excerpts supporting each finding, volumes per period, no individual score derived from these counts |
| Coaching a sales force | Linking exchanges to an opportunity, CRM integration, the link between behaviours and the sales outcome, actual use by salespeople |
Our twelve-criteria checklist for choosing a conversational analysis solution covers the rest of the purchase: data hosting, languages, integrations, cost.
4. Be wary of the tool that does everything
A tool can serve several disciplines, but each row of the table is demonstrated separately. The vendor who ticks every box without a demonstration is answering the label, not your need.
The test: which discipline does your need point to?
The phrases found in specifications often point to a discipline without naming it. A few examples, with the question to ask in return:
| What is written | The discipline targeted | The question to ask in return |
|---|---|---|
| "Reduce our supervisors' listening time" | Scoring advisors | To evaluate what, and to decide what? |
| "Evaluate 100% of our calls" | Scoring, or checking | For fairness between advisors, or for the completeness of a compliance check? |
| "Prove the duty to advise to the ACPR" | Checking compliance | On which scope of calls, and with what trail? |
| "Know why our customers call back" | Listening to the voice of the customer | Who will decide to fix the product or the process? |
| "Detect customers who are about to leave" | Listening to the voice of the customer | Who will call them back, and how soon? |
| "Improve our conversion rate" | Coaching sales, or scoring in telesales | Is the sale won in one call or over several meetings? |
| "Harmonise quality across our outsourcers" | Scoring advisors | With a common scorecard, or one scorecard per site? |
If your specification contains phrases from several rows of this table, it serves several disciplines. That is entirely possible, provided you write it down.
Where Raisetalk fits
Raisetalk is designed for contact centre calls, analysed after the call, across the entire volume. Here is what it covers, discipline by discipline.
- Scoring advisors. Evaluation scorecards with conditional criteria and N/A, a justification for each criterion, contestation by the advisor, the debrief, and re-evaluation with another scorecard to calibrate before deploying.
- Checking compliance. Every call goes through the same scorecard, and every finding refers back to the passage of the call that supports it. A human correction requires a comment; the author, their role and the date are kept, along with the version history. An alert rule fires when a sensitive criterion fails. A criterion that neither adds nor removes points keeps its verdict without affecting the score: it remains filterable and is read separately. Access is set through roles and permissions, and the retention period separately for the audio and for the transcription.
- Listening to the voice of the customer. The Understand page accepts a question in natural language and answers with the excerpts it relies on. The management report shows why customers call, reason by reason, and the signals they express, a threat to cancel or a dispute for example, team by team.
- Coaching, on the phone. The debrief helps advisors improve from their own calls. For telephone sales, the "mechanism" page of the management report compares the rate of an expected outcome, an appointment booked for example, depending on whether an advisor behaviour is present or absent, at equal score.
What Raisetalk does not do: coaching B2B salespeople on video-call meetings linked to an opportunity in a CRM. For that discipline, a revenue intelligence platform is designed for the purpose, as we wrote in our comparison. And Raisetalk is not the reference archive of your recordings: for proof, the unaltered original remains the one held by your recording system or your telephony provider.
Before your next call for tenders, write one line per discipline. The discipline targeted, the decision it informs, the unit judged, what must carry weight. If a line stays empty, that discipline is not part of your need, and the tool does not have to cover it. If every line is filled in, ask for one demonstration per line.
Key terms
- Quality monitoring: the systematic evaluation of exchanges between advisors and customers, against a written standard, in order to reach a decision
- Unit judged: what the decision is about (an advisor, a call, a call reason or a sales opportunity)
- Offsetting: a property of a weighted score, where a passed criterion makes up for a missed one; acceptable for scoring an advisor, not for checking an obligation
- Defensibility: the ability of a finding to stand up before a third party, inspector or judge, because it identifies the call checked, the passage that supports it, and the author and date of each correction
- Representativeness: the ability of a count to describe all customers, and not only those who called
- Speech analytics: a technique for transcribing and automatically analysing calls, serving several disciplines
Frequently asked questions
What is quality monitoring in a customer contact centre?
It is the systematic evaluation of exchanges between advisors and customers, against a written standard, in order to reach a decision. Depending on the decision in view, the term covers four disciplines: scoring advisors to help them improve, checking compliance with regulatory obligations, understanding customers' requests, or improving a sales team's performance. They share the technique, not the scorecards or the standards of proof.
What is the difference between quality monitoring, quality assurance and quality management?
Quality monitoring is the common label. Quality assurance most often refers to scoring advisors against a scorecard. Quality management refers to the full cycle, from the scorecard to corrective action by way of calibration and feedback to the advisor, and often names a module in contact centre software suites.
Is speech analytics quality monitoring?
It is a technique: the transcription and automated analysis of call content. It serves advisor scoring, but also the voice of the customer and compliance. A speech analytics tool can count call reasons without scoring a single advisor.
Which software can automate call quality monitoring?
It depends on the discipline. To evaluate and check every call in a contact centre, whatever the telephony, look at conversational analysis tools designed for the contact centre, Raisetalk among them. For sales meetings held by video call, look at revenue intelligence platforms. Our comparison presents twelve tools. In every case, have what your discipline requires demonstrated on your own calls.
Can a sales coaching tool be used for compliance checks?
It is not the task it was designed for. A compliance check requires checking every call in scope, criteria that cannot offset one another, the passage that supports each finding and a record of human corrections. If a tool allows this, have it demonstrated line by line before entrusting it with your compliance.
Getting started
Before choosing a tool, name your discipline. To then see how Raisetalk evaluates your calls, try it on your own conversations.
