Get started

Taking an assessment rather than buying one? This page is written for employers. Here is the page for candidates.

High-velocity and transactional sales · Entry level

How to assess a Inbound Sales Advisor

This role is screened as a support role, because it sits in a contact centre and the candidate is answering rather than initiating, and that misclassification is the whole problem. Support screening rewards empathy, patience and process adherence: all necessary here, none of which produce revenue. What separates the top quartile is a commercial diagnosis performed inside a service conversation — hearing that the caller asking for the premium tier will not use it, saying so, and closing the smaller sale anyway, which loses value today and wins the renewal and the absence of a refund. No interview can observe that trade, because in an interview every candidate says they would recommend what is right for the customer. The second unobserved thing is the turn: inbound calls arrive with a service question wrapped around a buying intent, and the moment where the advisor stops solving and starts selling — or never makes the turn at all — appears in no current step.

The inbound sales advisor's advantage over every other seat in this family is that the customer called them. Intent is already present, the qualification question is half answered, and the base conversion rate is high enough that a mediocre advisor still looks acceptable on a weekly report. That is precisely what makes the role hard to hire for: the floor is high and the ceiling is invisible. A gym chain, a booking platform, a broadband provider or a subscription business will convert some proportion of inbound calls almost no matter who takes them, and the difference between a median advisor and a strong one shows up not in the conversion number but in average order value, refund rate, and whether the customer is still there in six months.

Hour to hour the work is a queue. A call arrives with a question — what is included, can I change my dates, why is this more expensive than the website said — and the advisor has to answer it accurately and then decide whether there is a sale in the room. Two failure modes bracket the median. One is the order-taker, who answers well, processes exactly what was asked for, and never discovers that the caller has three people to book rather than one. The other is the pusher, who hears intent and immediately reaches for the largest tier, converting a customer who cancels within the cooling-off period and leaves a one-star review on the way out. Both hit their weekly number. Only one of them is still profitable at ninety days.

What separates the top quartile is the willingness to recommend down. It is a genuinely counter-intuitive commercial move — it reduces today's value, it feels like leaving money on the table, and it is the strongest single predictor of retention in a subscription or membership business. The advisor who tells a caller "you'd be paying for four sessions a month and you've just said you'll come once, so take the smaller plan" closes something the customer keeps. The second separator is the transition itself, which is a matter of timing more than of words: solve the question completely first, confirm it is solved, and only then open the commercial thread. Advisors who reverse that order are experienced by the customer as being sold to instead of helped, and their conversion falls.

What a hiring manager is really trying to predict is contribution after cancellations and, in booking and membership businesses, whether the customer comes back. Neither is contained in an interview about handling difficult customers, and neither is contained in the typing test that sits beside it.

What the job actually needs

How people fail in this seat

What most employers do instead

A customer-service style competency interview, a listening or comprehension test, occasionally a shadowing session — very often the same process the contact centre uses for support hiring with one sales question bolted on.

This role is screened as a support role, because it sits in a contact centre and the candidate is answering rather than initiating, and that misclassification is the whole problem. Support screening rewards empathy, patience and process adherence: all necessary here, none of which produce revenue. What separates the top quartile is a commercial diagnosis performed inside a service conversation — hearing that the caller asking for the premium tier will not use it, saying so, and closing the smaller sale anyway, which loses value today and wins the renewal and the absence of a refund. No interview can observe that trade, because in an interview every candidate says they would recommend what is right for the customer. The second unobserved thing is the turn: inbound calls arrive with a service question wrapped around a buying intent, and the moment where the advisor stops solving and starts selling — or never makes the turn at all — appears in no current step.

The assessment

About 18 minutes end to end.

The systems it runs in

A queue. The call arrives from a contact-centre routing surface with a wait time already on it and other calls visibly waiting behind it, and the candidate works it against two panels: the customer or booking record, and the product sheet as a tier catalogue with the out-of-headline fee held as a field on the tier rather than as a line in a document. Nothing about the call surfaces that fee — the candidate has to open the tier to find it, which is what makes the third fork a real omission rather than a reading test. At the end the candidate sets a wrap-up or disposition code, records what was sold against the customer record, and writes t2's confirmation from that record while the next call waits.

Any routing platform that presents a queued call with a wrap-up code, over any customer record that can hold a purchase and a catalogue of tiers. The business layer varies more than the telephony does — a membership platform, a reservations system or a subscription billing account all work — and the fixture is rebuilt on the buyer's own tier structure and fee names where they provide a sandbox, since the fee that is not in the headline price is always specific to the buyer's own product.

Working speed is scored. This is a queue seat and handle time is a real constraint in it, so the design puts the queue on screen and reports handle time and wrap-up time. Both are scored strictly against the disclosure and recommend-down criteria and never on their own, because in this fixture the fastest available call is also the worst one: selling the top tier the caller asked for by name, without resolving their administrative question and without mentioning the fee, is the shortest path to a completed sale and a delighted caller. A buyer who reads the speed column alone will rank that candidate first, so the report states beside the number that handle time is not a positive signal in isolation and shows it next to the disclosure score rather than in a separate column.

What the candidate actually does

TaskWhat happens
The service question with a sale inside it
live_call · 10 min		The candidate is given a one-page product sheet — three tiers, what each includes, one fee that is not in the headline price, and a cancellation window — and then takes a queued call. The AI caller opens with an administrative question, not a buying question: their bill is higher than the website said, or they need to change a booking. Intent is present but wrapped. Three forks are laid in and they fire in sequence. The first is the turn: if the candidate pivots to selling before the caller's question is fully resolved and acknowledged, the caller cools measurably and becomes resistant for the rest of the call; if the question is resolved and confirmed first, the caller volunteers a fact that changes what they should buy — there are three of them now, not one, or they are moving in six weeks. The second is the recommend-down fork: the caller asks for the top tier by name, having already described usage that plainly does not need it, and will buy it instantly and gratefully if sold it. That is the largest sale available, in the shortest time, with a delighted customer at the end of the call. The third is silent: the fee outside the headline price applies to this caller and the caller never asks about it. Nothing in the call surfaces it unless the advisor does.
What the customer receives afterwards
written_artifact · 5 min		The confirmation the caller gets, written without access to the transcript. It must state what was bought, what is included, what is not, the fee, and how the cancellation window works. Scored against the call: anything it promises that the call did not, and anything the call promised that it omits, is the same error in two directions. In a subscription or membership business this message is the first thing a customer re-reads when they are considering cancelling. It is written from the customer record, after the candidate has set the wrap-up code and recorded what was sold, with the next call in the queue already waiting.

The mark scheme

Each criterion is scored 1 to 5 against written anchors, and every score is reported with the excerpt that earned it. A criterion marked floored is reported as a finding rather than averaged into the total. The first is open; open any other to read its anchors in full.

Resolves the service question before opening the commercial threadweight 0.2Answers, confirms out loud that it is resolved, and opens the commercial thread with a question about the caller's situation rather than an offer.
1 Begins selling while the caller's original question is unanswered, or answers it and pivots in the same breath without checking it landed.
3 Answers the question fully, then pivots with a general offer.
5 Answers, confirms out loud that it is resolved, and opens the commercial thread with a question about the caller's situation rather than an offer.
Diagnoses what the caller needs rather than processing what they asked forweight 0.2Asks enough to establish real usage and surfaces a fact the caller had not mentioned that changes what they should buy — and then acts on it.
1 Processes exactly the request; asks nothing about how the caller will actually use the thing.
3 Asks one situational question and uses the answer.
5 Asks enough to establish real usage and surfaces a fact the caller had not mentioned that changes what they should buy — and then acts on it.
Recommends the smaller option when the evidence says so, and still closesweight 0.25States the mismatch using the caller's own numbers, recommends the smaller plan explicitly, and then still asks for the order and confirms the next st…
1 Sells the tier the caller named, or a larger one, against stated usage that does not support it.
3 Mentions the smaller option as an alternative and lets the caller talk themselves back into the larger one without correcting the assumption.
5 States the mismatch using the caller's own numbers, recommends the smaller plan explicitly, and then still asks for the order and confirms the next step rather than leaving the sale unmade.
Discloses the fee or exclusion unpromptedweight 0.2Raised by the advisor before payment, in plain terms, at a point where it could still change the caller's choice, with a check that it was heard.
1 Never mentioned; the caller ends the call believing it is included.
3 Mentioned only because the caller happened to ask, or read out at the end after the sale is agreed.
5 Raised by the advisor before payment, in plain terms, at a point where it could still change the caller's choice, with a check that it was heard.
The written confirmationweight 0.15Included, not included, the fee, the cancellation window and how to use it — all matching the call exactly, in language a caller would not need to pho…
1 Plan name and price only, or contains a commitment the call does not.
3 Accurate summary of what was bought.
5 Included, not included, the fee, the cancellation window and how to use it — all matching the call exactly, in language a caller would not need to phone back about.

How it is scored

Weighted mean of five criteria scored 1-5 against the anchors, each with its excerpt. Two derived signals are reported alongside the score rather than inside it, because they are what the seat is actually managed on. The first is which branch the caller took at the turn, with its timestamp. The second is a simple sold-versus-should-have-sold comparison: which tier the candidate closed against the tier the caller's stated usage supports. A candidate who closes the top tier here scores well on nothing and should not be ranked highly by a total that quietly rewards order value.

Integrity

The log describes what happened. It does not produce a cheating verdict — the follow-up conversation is the control, because a statistical accusation is not something we would ask a reviewer to defend.

What you receive

Who decides

Required before rejection, and specifically required before any hire made on a high score, because this is the design where an automated total is most likely to be wrong in the flattering direction. The reviewer reads two things: the ninety seconds around the recommend-down fork, and the confirmation message against the transcript. Under three minutes. The reviewer is asked one question in writing before seeing the score — did this advisor sell the caller something they will still be paying for in six months — and only then confirms or overrides. Deployers running seasonal cohorts should review the cohort's tier-sold distribution before making offers; if almost everyone sold the top tier, the scenario's usage signals were too subtle and the round should be rerun rather than ranked.

What this does not measure

No criterion scores accent, dialect, first language, speech rate, disfluency, voice pitch, or perceived warmth and empathy, and there is deliberately no rapport or customer-manner criterion — in inbound contact-centre hiring that is the standard route by which speech characteristics get scored under a commercial-sounding name. Every anchor here is decidable from the transcript text; deployers should rescore a sample transcript-only and treat systematic divergence from the audio score as evidence that delivery is leaking in. Three specific risks. First, this seat is frequently hired for multilingual queues and then assessed in English; run the assessment in the language of the queue the person will work, and if a candidate will serve two queues, assess in the one they will spend most time in rather than the one the hiring team speaks. Second, the written task is fifteen percent of the score and should be marked on content only — never on spelling, grammar or register — because the business risk it stands for is an omitted fee, not a misplaced comma. Third, do not use voice, sentiment or emotion analysis to derive any part of this score. Offer a chat-channel variant of t1 on request, which is a genuine version of this job rather than an accommodation fiction, and publish how to request it up front. Monitor pass rates criterion by criterion, since the two criteria carrying the diagnosis work will show impact long before the total does.

The inbound advisor is screened as a support agent almost everywhere, because they sit in a contact centre and answer rather than dial, and the role file identifies that misclassification as the whole problem. Support screening rewards empathy, patience and process adherence. All three are necessary here and none of them produces revenue, which is why the floor in this seat is high and the ceiling is invisible: a mediocre advisor converts inbound intent well enough to look fine on a weekly report. This design is built to find the ceiling, and it does so by putting three things in the same ten-minute call that no interview can hold at once.

The first is the turn. Inbound calls arrive with a service question wrapped around a buying intent, and the moment where the advisor stops solving and starts selling is a matter of timing rather than words. Making the AI caller cool measurably when the pivot comes too early is the single most important mechanic in this design, because it converts a soft coaching observation into an observable branch: the transcript either contains the confirmation that the question was resolved, or it does not, and the caller's subsequent behaviour diverges from that point. Advisors who reverse the order are experienced by real customers as being sold to instead of helped, and their conversion falls; here it falls inside ten minutes and is visible.

The second is the recommend-down fork, which is the reason this design cannot be a find-and-replace of the inside sales representative's next door. That role's fork is about holding price under a discount temptation. This one is the opposite shape: the caller is offering to buy the most expensive thing, is grateful, will not object, and should not have it. The cheap path is not a lie — the advisor is simply giving the customer what they asked for, which is the most defensible thing in the world at the moment it happens and shows up as a cancellation, a refund and a one-star review at ninety days. The strong behaviour, per the role file, is the counter-intuitive one: state the mismatch in the caller's own numbers, sell the smaller plan, and still close it. That final clause matters and is written into the anchor, because an advisor who recommends down and then forgets to ask for the order has swapped one failure mode for another.

The third is the fee nobody asks about. It is silent by design and no branch prompts it, so the only route to a five on that criterion is the advisor deciding, unprompted, that a customer about to pay should know. Scoring it before payment rather than at any point in the call is deliberate: a disclosure made after the sale is agreed is a notification, not information the customer could act on.

Eighteen minutes — ten of call, five of writing, three of brief and setup — is set by the same constraint that governs the rest of this family. These seats are hired in seasonal waves at contact-centre scale, and a screen that only runs on a shortlist stops being run. Two tasks rather than three is a deliberate cut: the judgment scenarios that suit the outbound roles add little here, because the judgment in this seat is not separable from the call. It happens live, inside someone else's question, or it does not happen at all.

What the design does not predict is queue endurance, systems fluency or adherence — the operational half of contact-centre performance, all of which is measured directly and cheaply after hire. What it predicts is contribution after cancellations, which is the number this seat is actually run on and the number that arrives too late to inform the decision.

Sources

Every figure on this page is traceable. Where a claim could not be sourced it is stated qualitatively instead.

  1. US Bureau of Labor Statistics, Occupational Employment and Wage Statistics, National employment and wage data by occupation, May 2025: Sales Representatives of Services, Except Advertising, Insurance, Financial Services, and Travel (41-3091), 1,256,010 jobs, mean annual wage $82,430, median hourly wage $33.65, https://www.bls.gov/news.release/ocwage.t01.htm
  2. US Bureau of Labor Statistics, 2018 Standard Occupational Classification Definitions, 41-3091 Sales Representatives of Services: 'Sell services to individuals or businesses. May describe options or resolve client problems.', https://www.bls.gov/soc/2018/soc_2018_definitions.pdf

See what the employer actually receives. A full report for one role, with every score shown beside the excerpt that earned it, conduct findings reported rather than averaged, and a reviewer sign-off required before any decision. No form.

Read a sample reportOr talk to us about this role