Sales call accuracy monitoring: correct pitch or unapproved promises?

The short answer

Sales call accuracy monitoring means comparing every claim the agent makes — price, feature, deadline, warranty, campaign condition — with the approved source in force on the date of the call. There are three kinds of result: correct (matches the source), unapproved (not in the source; it may even be true, but the agent had no authority to say it) and wrong (contradicts the source). The transcript shows that a claim was made; the source shows whether it was true.

So checking accuracy on a sales call has two parts: finding the claims — call analysis does that — and comparing them with the source — a person or a separate data connection does that.

Why this criterion is tracked separately

A promise made in a sale is the basis of the customer's decision. If "we'll install it tomorrow" was said but the service time is three days, the customer is unhappy, complains and sometimes cancels. If "the warranty is two years" was said but the terms say one, that is already legal risk. A wrong promise costs more than an ordinary quality criterion, because its consequences show up after the call — in delivery, complaints, returns.

Agents rarely say something wrong on purpose. The cause is outdated information, sales pressure, or simply not knowing the answer and guessing. The criterion's purpose is to find those causes, not to punish.

The source of truth: what you need

• A product catalogue — features and models, with version dates
• A price list — dated; the "current" list is no good for checking last month's call
• Campaign terms — start and end dates, who they apply to, what they cannot be combined with
• Service times — delivery, installation, response time; regional differences
• The warranty and returns policy
• Permitted and forbidden phrases — for example, never using "guaranteed results"

Every source needs an owner: marketing updates the price list, operations updates service times. A source without an owner goes stale, and agents stop using it.

Six kinds of promise

• Price and discount — "we'll give you a special discount"
• Timing — "it'll be delivered tomorrow", "we'll install it in an hour"
• Features — a function the product does not have, or a wrong technical figure
• Promised outcomes — "you'll definitely lose weight", "you'll pass the exam"
• Warranty and returns — a condition that does not exist, or a broader one
• Social proof — unverified claims such as "everyone chooses this", "today is the last day"

The checking process

1. 1Find the claimsPrices, deadlines, features and promises stated on the call are recorded with their transcript lines.
2. 2Classify themEach claim is assigned to one of the six kinds above — which decides which source to look at.
3. 3Compare with the sourceThe claim is compared with the source in force on the date of the call.
4. 4StatusCorrect, unapproved, wrong or unable to assess (the figure cannot be heard clearly).
5. 5Human confirmationWrong and unapproved claims are confirmed by a person with the evidence line.
6. 6Root causeFor confirmed cases the cause is recorded: stale information, training, pressure, no source.

The definition card

• Question: Did everything the agent said about the product, price, timing and conditions match the approved information in force on the date of the call?
• Yes: every claim matches the source
• Partial: the claim is right, but an important condition was left out (the "up to" in "100 Mbit", for example)
• No: at least one claim contradicts the source or is an unapproved promise
• Not applicable: no claims about product, price or timing were made on the call
• Evidence: the line of the claim, and the name and date of the source it was compared with

An illustrative example

Not a real call. A customer wants home internet. The source: the plan is "up to 100 Mbit", installation takes three working days, the contract is one year.

• Line 6, agent: "The speed is 100 Mbit" — partial; the "up to" condition was not stated
• Line 9, agent: "We'll come and install it tomorrow" — no; the source says three working days
• Line 11, agent: "You can cancel the contract whenever you like" — no; the contract is for one year
• Line 13, agent: "The router is free" — correct, if that campaign is in force on that date; needs checking

After this call the customer will wait three days, and when they try to cancel they will face a penalty. Both problems were created on the call, but they will only surface long after it.

"Unapproved" and "wrong" are different

An agent may talk about a product launching next month, and what they say may be true. But it is not yet approved sales information. Counting it as "wrong" is unfair; counting it as "correct" is risky. The "unapproved" status shows that case separately, and it usually points to a gap in the process: nobody wrote down what agents may and may not say.

Fixing by root cause

• Stale information — the source's owner does not get changes to the call center and the standard in time
• Not knowing and guessing — agents are taught that "let me check and call you back" is an acceptable answer
• Sales pressure — check whether targets and rewards encourage wrong promises
• No source — if a question comes up often, the answer is written down and added to the source

Limits

A transcript cannot check whether a fact is true — it only shows what was said. Figures can be misrecognised on a noisy recording. An agent's claim depends on context: by "tomorrow" they may have meant the customer's district. So a "wrong" status should only reach an agent's record after human confirmation, and that result on its own should never be the basis for a disciplinary decision.

Accuracy checks in Vexvon Audio Analyzer

Vexvon Audio Analyzer handles the first part of the process: if your standard includes the step "the agent gives only approved information", every call brings back that step with a status and the transcript lines where the claims were made. The analyzer does not connect to your catalogue, price list or campaign data, though — a person checks from the evidence line whether the claim was true on that date. Because poorly recognised lines are marked, you can also see where a figure may be unreliable.

The separate criteria for a price call are in price inquiry call evaluation, the effect of a confirmed breach on the score in critical errors, and the owners of sources in quality management roles.

First step

Check which of the six sources you have with dates and owners. Then take ten sales calls from last week, pull out every timing and warranty promise and compare them with the source. If most mismatches point to one source, the problem is not the agents. More in the sales call quality section; get in touch.

Further reading on this topic: hospitality call quality assurance.