Get started

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

Accounting and finance operations · Entry level

How to assess a Accounts Receivable Specialist

The AR screen tests the wrong half of the job. Employers interview for tenacity, but the expensive failures in a sales ledger are diagnostic rather than motivational. The defining judgment is whether a customer who has not paid is a payer or a disputer, and being wrong in either direction costs real money: chase a legitimate dispute and you damage an account the sales team owns, treat a delinquent as "in query" and the cash never arrives while the debt ages past the point where anyone will authorise action. Nothing in a CV or a conversation observes a candidate looking at a remittance advice that does not tie to the invoices it claims to pay and working out why — which is the actual first hour of the actual job.

This is the ledger side of getting paid, and it is worth stating the boundary first because the title is used for two different jobs. An accounts receivable specialist raises invoices correctly, applies incoming cash to them, resolves the differences, keeps credit notes and disputes documented, and maintains an aged debt report that means what it says. The conversational side of collections — arrears calls, payment plans, hardship, the escalation to legal — is a different discipline with different failure modes, and it is covered in this corpus by the collections and retention family. A business that conflates the two usually ends up with someone good at one and poor at the other, and then concludes that AR is a low-skill function.

The daily rhythm is cash application. A payment arrives for £48,317.40 against a remittance listing eleven invoices totalling £48,600. The specialist has to find the £282.60: it might be a bank charge deducted at source, an early settlement discount taken correctly or taken without entitlement, a credit note the customer applied that the ledger does not have, a short payment for two damaged units, or a deduction the customer has invented and will keep taking every month until somebody notices. Each of those has a different correct action, and only one of them is "chase the customer". The median performer, under time pressure and with a month-end cash target, posts the payment and leaves the difference on account as unallocated. The compound effect is a debtors ledger where nobody can tell which invoices are genuinely outstanding, which is how a business ends up chasing customers who have already paid.

The other half of the role is deciding what an unpaid invoice means. Aged debt reports flatten every reason for non-payment into a number of days, and the single most useful thing an AR specialist does is put the reason back in. An invoice at ninety days because the customer's purchase order number was omitted and their AP system auto-rejected it is not a credit risk; it is a self-inflicted billing error, and the correct action is to reissue and apologise. An invoice at ninety days because the customer disputes the quantity delivered is a commercial matter that needs the account manager. An invoice at ninety days with no reason recorded at all is the dangerous one. The top quartile in this role are distinguished almost entirely by their intolerance of that third category.

There is a written component that hiring managers consistently underrate. AR correspondence goes to somebody else's accounts payable department, which is to say to a person with their own queue and their own tolerance thresholds, and the message either makes it easy for them to pay or it does not. A weak AR email attaches a statement and asks for an update. A strong one names the invoice, the amount, the purchase order it was raised against, what the customer's own system said when it rejected it, and the single thing needed to release it. The difference is not politeness; it is whether the recipient can act without a reply. That artefact is trivially easy to set as an exercise and is never asked for before an offer.

What a credit manager is really trying to predict is whether this person will keep the ledger honest when the incentive is to make it look clean — because a tidy aged debt report and a true one are not the same document, and only one of them tells the business how much cash is actually coming.

What the job actually needs

How people fail in this seat

What most employers do instead

CV screen, a competency interview about persistence and being comfortable chasing money, and occasionally a spreadsheet test.

The AR screen tests the wrong half of the job. Employers interview for tenacity, but the expensive failures in a sales ledger are diagnostic rather than motivational. The defining judgment is whether a customer who has not paid is a payer or a disputer, and being wrong in either direction costs real money: chase a legitimate dispute and you damage an account the sales team owns, treat a delinquent as "in query" and the cash never arrives while the debt ages past the point where anyone will authorise action. Nothing in a CV or a conversation observes a candidate looking at a remittance advice that does not tie to the invoices it claims to pay and working out why — which is the actual first hour of the actual job.

The assessment

About 30 minutes end to end.

The systems it runs in

A sales ledger in an ERP with the customer's open items on it, a cash application screen where a receipt is matched against those items line by line, a set of deduction and dispute reason codes maintained on the account, and an aged debt report drawn from the same sub-ledger. The remittance advice arrives the way it actually arrives — as an attachment on an email, in the customer's own layout rather than the ledger's. Where a receivables automation layer sits over the ERP, that is the surface: HighRadius's published material describes capturing remittance from email attachments, check stubs, EDI 820 and 823 files and customer portals, identifying short payments and mapping them to the reason codes already defined in the ERP, which is exactly the decomposition this task asks a person to do by hand. Naming it changes what the task observes, because the graded artefact stops being a description of four differences and becomes four coded open items somebody now owns, plus a residual that is either zero or stated. A spreadsheet is the working surface for the decomposition: eleven invoices on one sheet, one payment and a remittance on another, and the candidate has to reconcile across them and end with a residual they can name. What separates fast-and-correct from slow-and-correct here is not formula fluency but whether the workbook is arranged so the four causes stay separate; a single netted difference is fast, is arithmetically right, and answers nothing.

Any sub-ledger with customer open items, a match-and-allocate screen, a reason-code list and an ageing report. The fixture is rebuilt inside the buyer's own instance where they provide a sandbox; where they do not, the open items and the remittance ship as plain tables and the allocation is recorded on the face of them. The reason-code list is the one field worth taking from the buyer even without a sandbox, because a candidate's choice between "short payment" and "pricing dispute" is only meaningful against the codes their team actually uses. No criterion is satisfied by knowing a named product's menus.

Working speed is scored. Cash application is a queue worked against a month-end reporting date, the brief states that date, and how much of the receipt was allocated inside the window is scored. It is scored only against the residual criteria and never alone: allocating the whole payment fastest by parking the difference on account is the cheap path the fork exists to catch, and it must not read as a good result. Typing speed and transaction volume per hour are not measured anywhere.

What the candidate actually does

TaskWhat happens
The remittance that does not tie
data_task · 13 min		One customer, one payment, and a remittance advice that does not agree with it. The pressure is stated in the brief in the same words for every candidate: it is the last working day of the month, the cash collection figure is reported to the board tomorrow, unallocated cash does not count toward it, and the candidate's team is short of target. The payment is a single round-ish sum against a remittance listing eleven invoices. The difference decomposes into four distinct causes and the candidate has to separate them: a bank charge deducted at source, a settlement discount taken outside the terms the customer is entitled to, a credit note the customer has applied that the ledger does not hold, and a short payment on two units against one invoice with no explanation given. One of the eleven invoices was raised without the customer's purchase order number and was auto-rejected by their system, which is visible in the pack and is mentioned by nobody. The candidate allocates the cash line by line on the cash application screen, states the residual, and records a reason code against every difference, so that each one leaves the task as an open item on the account with an owner rather than as a sentence in a summary.
The fork. Posting the payment and parking the difference on account clears the item, hits the cash figure, takes ninety seconds and is invisible until somebody tries to age the ledger three months later. It is also what a large share of working AR specialists actually do at month end, which is why it has to be the cheap path here rather than an obviously wrong one.
What the aged debt report is not telling you
judgment_scenario · 9 min		An aged debt extract for the same customer plus five others, with the days column and nothing else. Seven balances sit past ninety days. Behind each is a short note — a delivery query, an omitted purchase order number, a customer who has gone quiet after promising payment twice, an invoice disputed on quantity, a balance where the only record is three chase dates, a balance that is genuinely just slow, and one where the goods were never delivered. The candidate assigns each a disposition: chase, reissue, route to the account manager, route to collections, or write up as unknown. No option is labelled correct and the option set deliberately includes one that is honest rather than tidy.
The fork. Every balance can be given a plausible category, and a report with seven categorised balances looks like work done. The item with three chase dates and no reason is the one that matters: assigning it "slow payer" makes the report complete, and recording it as unknown makes the candidate look like they did not finish.
The email to somebody else's accounts payable
written_artifact · 8 min		Two short messages written back to back, graded against what the candidate actually found in t1 rather than against a model answer. The first goes to the customer's accounts payable department about the invoice their system rejected. The second goes internally to the account manager about the quantity dispute. Neither recipient has the ledger open, and the AP recipient has their own queue and their own thresholds.
The fork. Attaching the statement and asking for an update is the standard AR email, is quick, and is answered at a rate close to zero. Naming the invoice, the amount, the purchase order it should have carried, what the customer's own system said when it bounced, and the single action that releases it takes four times as long and is the only version that gets paid without a second round.

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.

The difference is decomposed rather than parkedweight 0.3Separates all four, states the amount attributable to each, distinguishes the settlement discount taken without entitlement from the deductions that a…
1 Posts the payment and leaves the residual as unallocated cash or on account, with no reason recorded, or forces the allocation by spreading the difference across invoices so the ledger appears clean.
3 Identifies two or three of the four causes, typically the bank charge and the credit note, and leaves the remainder as a single unexplained balance without saying what it might be or what would resolve it.
5 Separates all four, states the amount attributable to each, distinguishes the settlement discount taken without entitlement from the deductions that are legitimate, and leaves nothing in the residual that is not described.
A dispute is diagnosed as a dispute and a delay as a delayweight 0.2Identifies the invoice the customer's system rejected as ours to fix rather than theirs to pay, routes the quantity dispute to the account manager rat…
1 Treats every past-due balance the same way, either chasing all seven or routing all seven onward, so the aged report is processed rather than read.
3 Separates the obvious disputes from the obvious delays but assigns the self-inflicted billing error to chasing rather than to reissue, so the business pursues a customer for an invoice it raised wrongly.
5 Identifies the invoice the customer's system rejected as ours to fix rather than theirs to pay, routes the quantity dispute to the account manager rather than to collections, and separates the balance that is genuinely slow from the ones that are stuck for a reason.
An unknown is recorded as unknownweight 0.15States plainly that the reason for non-payment is not recorded, says what is known and what is not, and names the specific thing that would establish …
1 Assigns the balance with three chase dates and no reason to a category that fits the shape of the report — slow payer, or in query — with nothing behind it.
3 Flags it as needing investigation without stating what is actually known about it or what the next step would be, so it stays exactly as opaque as it was.
5 States plainly that the reason for non-payment is not recorded, says what is known and what is not, and names the specific thing that would establish it. The correct output here is the absence of a category, and the anchor is written so a reviewer can reward it.
The ledger is not tidied by concession flooredweight 0.15Names the concession as a decision that belongs to somebody else, states the amount and the recurrence risk, and continues to treat the balance as due…
1 Credits, writes off or discounts a balance to make a query disappear, or accepts the unentitled settlement discount silently so the account agrees. Reported as a finding on the face of the report and never averaged into the total, because this is the one move in this role that gives away money and leaves no trace of having done it.
3 Declines to concede but does not say who would authorise it or what the customer will be told, so the difference sits open and the customer keeps taking the same deduction next month.
5 Names the concession as a decision that belongs to somebody else, states the amount and the recurrence risk, and continues to treat the balance as due until it is decided.
The recipient can act without replyingweight 0.2Names the invoice, the amount, the purchase order it should have carried, what the customer's own system did with it, the single action that releases …
1 Attaches a statement and asks for an update, or lists a balance with no invoice reference, so the recipient's cheapest response is to ignore it.
3 Names the invoice and the amount and asks for payment, without saying why it was rejected or what specifically would release it.
5 Names the invoice, the amount, the purchase order it should have carried, what the customer's own system did with it, the single action that releases it, and a date. The internal note to the account manager states the same facts and the same amounts as the external one.

How it is scored

Weighted mean of the five criteria, each scored 1 to 5 against the anchors and reported with the ledger line or sentence that earned it. Two raw numbers are reported beside the score and annotated explicitly as not positive signals: the value of cash the candidate allocated, and the value left unallocated. A buyer scanning the report will otherwise read a high allocation figure as competence, when in this fixture a candidate who allocates everything has almost certainly forced something. The report also shows the four causes the candidate named next to the four planted, so a reviewer can see which were found rather than only how many.

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

Recommended. One judgment needs a person and cannot be encoded: how hard this buyer wants a customer pushed depends on whether they sell to a handful of strategic accounts or to a long tail, and the same disposition on the quantity dispute is correct in one business and wrong in the other. The reviewer reads the routing decisions and confirms or overrides the diagnosis criterion with a written reason. The ranking entitles a buyer to conclude that this candidate will keep a sales ledger honest under month-end pressure and can write a message another company's payables department will act on. It does not establish familiarity with this buyer's billing system or credit policy, and it says nothing about the conversational side of collections, which is a different discipline assessed elsewhere in this corpus.

What this does not measure

The design does not measure typing speed or spreadsheet formula fluency, and neither is scored anywhere. Throughput is scored, and narrowly: how much of the receipt was allocated inside the stated window, read only against whether the residual was resolved honestly, because the fastest allocation available is the one that parks the difference on account. Any arithmetic method is permitted including a calculator, and the residual can be reached by any route. No criterion reads register, idiom or fluency in the written messages; the fifth criterion is satisfied by the presence and accuracy of five specific facts, and a blunt message containing all five outscores a polished one that omits the purchase order reference. What the design does not reach is volume: a real AR specialist applies cash across hundreds of receipts a week, and one customer in thirteen minutes samples the diagnosis without sampling the endurance. It also cannot observe the slow social erosion that produces the characteristic failure in this role — the fourth month in which raising the same customer's invented deduction has produced nothing, and parking it becomes easier than escalating it again. Deployers should monitor the data task separately by group; a gap that shows there and not in the written segment indicates numeric-test format familiarity rather than the competency. Extended time and assistive technology should be available on request; the time box exists to control cost.

The accounts payable design in this family looks for a stop-threshold. This one looks for something different and slightly harder to see: whether the candidate will let a ledger stay untidy in order to keep it true. Everything in a sales ledger pulls the other way. Unallocated cash is a visible failure and a clean allocation is a visible success; an aged debt report with a reason against every line looks finished and one with an admitted unknown looks unfinished; and the person doing the work is measured on cash collected in a month that ends on Friday. The design's whole job is to make those pressures real and then see what survives them.

That is why the pressure is stated in the brief rather than left implicit, in the same sentences for every candidate: last working day, cash reported to the board tomorrow, unallocated cash does not count, team short of target. Without it, this is an arithmetic exercise and every candidate decomposes the difference carefully, because in a quiet room with nothing at stake everybody does. The competency the buyer is paying to observe is not whether the candidate can find a bank charge. It is whether they still find it when parking the residual takes ninety seconds and hits the number they are judged on.

The remittance is built with four causes rather than one because the diagnostic skill in this role is separation, not detection. A single planted difference tests arithmetic. Four differences inside one payment — a bank charge, a discount taken without entitlement, a credit note the ledger does not hold, and an unexplained short payment — force the candidate to decide how far to decompose before stopping, and each has a different correct action. Only one of them is "chase the customer", and a candidate who treats the whole residual as a collections problem has produced a chase that will annoy a customer who was mostly right.

The aged debt segment exists because the central judgment of the role appears exactly once in the cash application task, and one observation of a judgment is close to luck. Seven balances is cheap — a disposition and a reason, nine minutes — and it converts a single observation into several, including two directional traps. Chasing the invoice the customer's own system auto-rejected for a missing purchase order number is the self-inflicted error that damages an account for nothing; routing the quantity dispute to collections rather than to the account manager is the same mistake pointed at a commercial relationship.

The item worth designing most carefully is the balance with three chase dates and no reason. The option set includes "write up as unknown", and that option is correct for exactly one of the seven. Candidates who cannot bear an incomplete report will assign it to slow payer, which is a guess wearing a category's clothes, and the ledger then contains a fact nobody checked. This is the case where the correct output is the absence of a classification, and the anchors are written so that a reviewer can reward it rather than reading it as a candidate who ran out of time. It is also the reason human review is recommended rather than optional here: a purely mechanical ranking will always put a confident wrong category above an honest blank.

The concession criterion is floored. Issuing a credit note or accepting an unentitled discount to make a query go away is the one move in this role that gives money away and destroys the evidence that it happened, and a candidate who does it must not appear as a strong hire because the rest of the allocation was neat. A score of 1 there is reported as a finding on the face of the report and never folded into the weighted mean.

The written segment is short, comes last, and is graded against the candidate's own allocation rather than a model answer. An AR email goes to another company's payables department — that is, into somebody else's queue, governed by somebody else's tolerances, and read by a person whose cheapest available action is to do nothing. The difference between an email that gets paid and one that does not is not politeness but whether the recipient can act without replying, which is a property of five specific facts being present. Grading it against what the candidate actually found also does the anti-coaching work: a memorised template cannot state what the customer's system did with an invoice in a fixture the candidate has not seen.

Thirty minutes, and the arithmetic is against the buyer's realistic alternative rather than an ideal one. This is a high-volume seat, filled by a credit manager or a shared-services lead who is hiring several people and whose competitor for attention is not a better assessment but no assessment. Thirteen minutes buys the diagnosis, nine buys the resampling that makes the diagnosis more than one data point, and eight buys the artefact the job actually produces. There is no live call in the design, which is deliberate and is a real limit: the role page draws a firm line between ledger work and collections conversations, and a buyer whose AR specialist will also be making arrears calls should be running the collections design in this corpus instead of, or as well as, this one.

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 Outlook Handbook, Financial Clerks, 2025, billing and posting clerks 418,000 jobs, https://www.bls.gov/ooh/office-and-administrative-support/financial-clerks.htm
  2. US Bureau of Labor Statistics, Occupational Outlook Handbook, Bookkeeping, Accounting, and Auditing Clerks, 2025, https://www.bls.gov/ooh/office-and-administrative-support/bookkeeping-accounting-and-auditing-clerks.htm

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