Document Reconciliation and Error-Checking AssessmentThe one you let through costs more than the one you queried.
Eight decisions under a cost table that is printed on your report, plus twenty-four exercises on finding the difference that is real. About twenty-five minutes.
Checking tests measure how fast you spot a difference. That is not the part that costs money.
Attention to detail is the most duplicated single construct in the hiring market, and almost every version of it is the same drill: two columns of numbers, find the odd one, against a clock. It measures how quickly you can see a difference. It says nothing about the decision that follows, and the decision is where the money is. Passing a wrong payee costs the payment. Querying a line that was fine costs somebody's afternoon and puts a real query behind forty routine ones.
So this is built the other way round. Eight cases put a real difference in front of you and offer the same four actions every time: pass it, record it and pass it, query the line, or hold it and escalate. Each action carries a published cost that depends on what the difference turns out to be, and the eight cases are balanced two apiece across four kinds of difference. That balance is the design. Passing everything costs forty-two units, recording everything twenty-four, querying everything twenty-two, holding everything twenty-four, against a zero that is genuinely attainable. Every policy lands in the same place and all of them land far behind reading the case.
The cost table is printed on your report in full. Deciding that a changed payee on a large invoice is worse than an annoyed supplier is a value judgement somebody made, not a measurement of anything, and burying it in code would make the score unarguable in the wrong way. It is there so you can disagree with it and see what your score would have been.
Twenty-four further exercises cover the rest of the work. Comparing a document against its source and marking only the lines that genuinely disagree — a list of findings that is mostly not findings is how a checking process stops being read. Telling one thousand from one thousand point zero zero, and two cartons from twenty-four units, and knowing which of those is a difference and which is a notation. Matching the shape of a difference to its likeliest cause, because a factor of ten is a decimal point and an exact repeat is a feed that ran twice. And writing the finding so the next person can confirm it without opening the documents again.
The report is a set of receipts. Every sentence is footnoted to the case it came from, so a conclusion you disagree with can be checked rather than accepted, and the two directions of your error are named in the words of the work: what you let through, and what you stopped that did not need stopping.
What you walk away with
The share of the avoidable cost you avoided, on a scale where zero is a typical reader and one hundred is the cheapest action on every case.
All sixteen cells, with your own action mix beside them, because the table is a judgement rather than a measurement and you are entitled to argue with it.
What always passing, always recording, always querying and always holding would each have cost on your own eight cases. The demonstration rather than the claim.
Let through, and stopped for nothing. Opposite errors with opposite fixes, which a single accuracy figure hides completely.
What the difference was, what you chose, what it cost, and what the cheapest action would have been. Footnotes, not a hover.
Picked from the direction you actually lean, because that is the only place a single habit can move a score built from a cost table.
Inside your report
Illustrative sample — your report is generated from your own responses.
| If you | No real difference | Size not yet known | Real, correctable | Real and expensive | You did |
|---|---|---|---|---|---|
| Pass it | 0 ✓ | 5 | 4 | 12 | 2× |
| Record and pass | 1 | 0 ✓ | 1 | 10 | 2× |
| Query the line | 4 | 2 | 0 ✓ | 5 | 3× |
| Hold and escalate | 5 | 4 | 3 | 0 ✓ | 1× |
Deciding that a wrong payee costs more than an annoyed supplier is a value judgement somebody made, so it is on the report where you can disagree with it.
The eight cases are balanced two apiece across the four kinds of difference, which is why the policies land so close together and all of them behind reading the case.
Every sentence in the report is footnoted to a case, so a conclusion you disagree with can be checked rather than accepted.
Built for
- Finance, accounts payable, billing and back-office teams who reconcile documents daily
- Operations and shared-services teams where a missed difference reaches a customer
- Anybody moving into a role where checking is part of the job and wants to know how they read
- Managers who want a reading on judgement rather than on speed
Eight decisions, one published cost table, twenty-five minutes
32 exercises across four formats · a cost-avoided figure with both spans, the table it was computed from, every policy priced, and receipts for every case.
Take free · full report ₹249 (incl. GST)
Frequently asked questions
No exercise is timed and no part of the score uses how long you took. That is deliberate. A checking test against a clock measures reading speed alongside judgement and then reports the mixture as one number, and the two are not the same thing.
Because it is a value judgement rather than a measurement. Somebody decided that passing a wrong payee costs more than annoying a supplier, and that decision drives every figure on the page. Printing it means you can disagree with it, recompute your own score against your own numbers, and argue about the thing that is actually arguable.
No, and the report shows you why on your own cases. Querying everything costs twenty-two units against an attainable zero, which is about the same as holding everything and recording everything. The eight cases are balanced across four kinds of difference precisely so that no policy wins.
It has no pass mark and it is not built as a screen. It reports a figure with its band, the two directions of your error, and a receipt for every case. The report says plainly that it should not be the only evidence in a hiring decision, and that it says nothing about how somebody works on the four hundredth document of the week.
About twenty-five minutes for thirty-two exercises. ₹249 in India, inclusive of GST, or US$2.99 elsewhere, one-time, for the sitting and the full report.
Each one takes a single capability, puts you inside the situations where it is actually tested, and scores your choices against published evidence — with a report designed for that capability alone, not a template. They span hiring, compliance, education, operations and personal skill.
Browse the catalogue →Methodology: Thirty-two original exercises across four formats: twelve keyed single-choice items, of which eight are decisions taken under a published cost table, eight select-every-that-applies comparisons against a source document, six true or false claims about reconciliation practice, and six match-the-following exercises pairing a difference with its likeliest cause. Construct statement: it measures whether somebody can compare a document against its source, tell a real difference from a difference in presentation, choose an action proportionate to what the difference could cost, and write the finding down so that somebody else can act on it. It does not measure accounting knowledge, arithmetic speed, eyesight, sustained vigilance, honesty, or suitability for any particular role. Declared response instruction: knowledge and judgment throughout. Each exercise has a defensible answer that does not depend on the respondent's preferences. Scoring: the eight decisions are scored against a decision-cost matrix rather than as right or wrong. Four actions are offered on every one of them - pass, record and pass, query the line, hold and escalate - and each of the four carries a different cost depending on what the difference turns out to be. The table is a value judgement rather than a measurement, so it is printed in full on the report, and the score is expressed as the share of the avoidable cost that was avoided, between a respondent answering at the declared answer priors and a respondent choosing the cheapest action every time. The eight cases are balanced two apiece across the four states, which is what makes every fixed habit cost about the same: passing everything costs 42 units, recording everything 24, querying everything 22 and holding everything 24, against an attainable zero. No policy beats reading the case. Construct grounding, drawn across sources rather than from one framework: statistical decision theory and the expected-cost formulation of a classification decision, in which the loss matrix and not the accuracy rate decides the right threshold (Wald, 1950; Berger, 1985; Elkan, 2001 on cost-sensitive decisions); signal detection theory and the separation of discrimination from where a threshold is set (Green and Swets, 1966; Macmillan and Creelman, 2005); measured human error rates in clerical and spreadsheet checking, and the finding that self-review detects far less of it than practitioners believe (Panko, 1998; 2008); the slip-versus-mistake distinction that separates a keying fault from a wrong model of the document (Reason, Human Error, 1990); the psychology of visual and proofreading comparison, and why a difference in a familiar field is harder to see than one in an unfamiliar field (Levin and Simons on change blindness, 1997); base rates and the effect of a low rate of real errors on the value of a positive finding (Meehl and Rosen, 1955); the alarm-fatigue literature on what happens to a checking process whose false-positive rate is high (Cvach, 2012); the audit concepts of materiality and of tolerable misstatement as published value thresholds rather than measurements (International Standard on Auditing 320); professional guidance on payment-diversion fraud and the rule that changed payee details are verified on a channel already held; McDonald's omega as the reliability estimate reported in place of alpha (McDonald, 1999); and the standard error of measurement as the reason a band rather than a point is reported (AERA, APA and NCME Standards, 2014). All items are original works. No item, scale name or report section is taken from any commercial instrument, and no affiliation with or endorsement by any vendor is claimed or implied.