Original benchmark · synthetic
Receivables Operating-State Benchmark: measuring what an ageing view cannot see
This benchmark runs on a synthetic ledger of 60 open invoices, constructed for the purpose. It is not client data — At Par has no customers, and nothing here is sampled, anonymised or derived from any real business. Each invoice carries both an ageing bucket and a full operating state: owner, last action, next action, promise and whether it was kept, dispute, evidence, and reconciliation status.
Four questions were then put to each invoice from each view. The result, computed and reproducible below: every one of the four ageing buckets contains all five required next actions, and 60 of 60 invoices sit in a bucket alongside at least one invoice in a materially different state. Sorted oldest-first, 2 of the top ten deserve the follow-up that sorting implies. The honest conclusion is narrow: a richer operating state improves what you can see and plan. It is not a claim that anyone pays faster.
Two views, four questions. What each one is able to contain.
An ageing view and an operating-state view are both descriptions of the same open invoice. They differ in which fields they hold, and therefore in which questions they can answer at all. Both are defined here before anything is counted, because a benchmark that defines its instruments after seeing the data is not measuring anything.
| Ageing view | Operating-state view | |
|---|---|---|
| Fields held | Invoice number · customer · balance · date issued or due · days elapsed · bucket | All of the ageing fields, plus: owner of the next action · last action and when · next action and its date · promise, with the person who gave it · whether that promise was kept · dispute state · evidence the customer still needs · reconciliation state |
| Q1 Is a date credibly expected? | Cannot answer | Committed · known cycle · nothing |
| Q2 Is anything blocking payment? | Cannot answer | No · evidence · dispute · never delivered |
| Q3 Does this need an owner decision? | Cannot answer | Yes · no |
| Q4 Is the balance collectable as stated? | Cannot answer | Yes · no |
Across the 60 invoices the ageing view answers none of the four questions and the operating-state view answers all four. That is not a finding. It is a restatement of what each view contains, and it would hold for any ledger, real or constructed. It is stated here so it is not mistaken for a result later.
Sixty invoices, built on purpose.
The ledger is synthetic. 60 open invoices across 18 customer accounts, in one notional currency shown as USD, with balances from 890 to 48,000 and 4 to 214 days past due. Stated total: USD 582,280. No real business, no anonymised extract, no client involvement of any kind — At Par has no customers.
| Ageing bucket | Invoices | Stated value (USD) | Distinct operating states | Distinct required actions |
|---|---|---|---|---|
| 0–30 | 20 | 182,740 | 8 | 5 |
| 31–60 | 18 | 206,300 | 9 | 5 |
| 61–90 | 12 | 126,850 | 8 | 5 |
| 90+ | 10 | 66,390 | 7 | 5 |
Synthetic ledger — constructed data, not client data.
The states were placed deliberately rather than drawn from a distribution, so the ledger exercises every combination the rubric can express. What it contains:
- 10 invoices with a live commitment from a named source, and 7 inside a customer’s known payment cycle.
- 4 with one broken commitment; 3 where a second date has also passed and a concession has been asked for.
- 5 with an open dispute, and 9 blocked on something the customer still needs — a purchase-order reference, a portal submission, an approved timesheet, a corrected legal entity.
- 3 the customer never received at all.
- 11 where the stated balance is not what will arrive: cash already received and unallocated, a credit note never applied, a duplicate, or a standing deduction never modelled.
- 5 never actioned since issue, and 8 with no owner for the next action at all.
This mix was chosen, and that matters for how the numbers should be read. A ledger built with fewer disputes and fewer reconciliation problems would produce smaller magnitudes. The construction is stated openly so a reader can judge it, and the full ledger is published at the end of this page so the counts can be recomputed against a different mix. What does not depend on the mix is the direction: an ageing bucket cannot separate states it does not hold.
How much an ageing bucket throws away.
Each invoice was given an operating-state class: the four answers, in order, as a single label — for example committed / no / no / yes for an invoice with a committed date, nothing blocking it, no decision needed and a balance that is collectable as stated. The rubric permits 48 such labels, some of which are logically impossible; the ledger exhibits 9 distinct classes, which is a modest spread and is already enough to make the point.
- Distinct operating states inside a single ageing bucket: 0–30 holds 8 · 31–60 holds 9 · 61–90 holds 8 · 90+ holds 7. Every bucket also contains all five required next actions.
- 60 of 60 invoices (100.0%) share an ageing bucket with at least one invoice in a materially different operating state. Counted as pairs: 404 pairs of invoices are indistinguishable by bucket and different in state.
- The tighter test. A reader who sorts by age and size still cannot separate them: 28 of 60 invoices (46.7%) have a near-twin — same bucket, balance within 15% — in a different operating state. That is 31 such near-twin pairs.
Working: bucket = 0–30, 31–60, 61–90, 90+ by days past due. Class = Q1/Q2/Q3/Q4 as defined above. “Materially different” means the class labels differ, so at least one of the four answers differs. Near-twin means the same bucket and balances within 15% of the larger. Synthetic ledger — constructed data, not client data.
Five invoices, one bucket, five different correct actions.
Selection rule, applied mechanically to the ledger: every invoice in the 31–60 bucket with a balance between USD 9,000 and USD 10,000. That returns 5 invoices, spread across just 6.5% of balance and 8 days of age. On any ageing report they are one row repeated 5 times. They require 5 different next actions, and four of the five would be actively wasted effort if the fifth were applied to all of them.
| Invoice | Days past due | Balance (USD) | Operating state | Required action |
|---|---|---|---|---|
| INV-1020 | 33 | 9,450 | Finance director committed in writing to this Friday. Next action scheduled. | Wait |
| INV-1021 | 35 | 9,200 | Committed to a date six days ago by an accounts clerk. The date passed in silence. | Chase |
| INV-1022 | 37 | 9,800 | Customer has queried the scope in writing. No follow-up will settle it. | Decide |
| INV-1023 | 39 | 9,600 | Paid three weeks ago in a lump transfer. The cash is in the bank and unallocated. | Fix |
| INV-1024 | 41 | 9,350 | Never accepted by the customer’s supplier portal. They cannot pay what their system will not take. | Ask |
Synthetic ledger — constructed data, not client data.
The consequences of treating them alike are not symmetrical. A reminder sent on the invoice already committed for Friday costs a little goodwill. A reminder on the disputed invoice is worse than nothing, because it answers a commercial question with an administrative act. A reminder on the invoice already paid three weeks ago is the one that ends a customer relationship, and it is entirely self-inflicted: the money is in the bank and nobody allocated it.
What an age-sorted worklist actually puts in front of you.
Each invoice was assigned exactly one required next action by a published precedence: Decide if an owner decision is needed; otherwise Fix if the balance is not collectable as stated; otherwise Ask if something specific is blocking the customer; otherwise Wait if a date is credibly expected; otherwise Chase. The precedence is a convention, stated so it can be argued with, and it is applied identically to all 60 invoices.
| Required action | What it means | Invoices | Stated value (USD) |
|---|---|---|---|
| Wait | A date is credibly expected and nothing is blocking it | 14 | 246,300 |
| Ask | One specific thing is missing on the customer’s side | 12 | 60,190 |
| Chase | Nothing is blocking it and nothing has been committed | 15 | 92,240 |
| Fix | An internal correction — no customer contact required | 11 | 84,300 |
| Decide | A commercial decision belonging to the business | 8 | 99,250 |
Synthetic ledger — constructed data, not client data.
Three results follow from that table and the ageing sort:
- 19 of 60 invoices (31.7%, USD 183,550) require no customer contact at all — they need an internal correction or a decision inside the business. On an age-sorted chase list, most of them are near the top, because unresolved internal problems age exactly like unpaid invoices.
- The top ten by age — the list an ageing report hands you first — contains 1 Wait, 2 Ask, 2 Chase, 2 Fix, 3 Decide. Only 2 of the 10 deserve the follow-up that sorting by age implies.
- Decisions are scattered, not stacked. The 8 invoices needing an owner decision sit at age ranks 4, 7, 10, 21, 22, 33, 38, 54 of 60. 5 of them fall outside the top ten, at ranks 21, 22, 33, 38, 54 — far enough down that a weekly pass over the oldest balances will not reach them.
Measure 5: what the stated total is worth. An ageing report totals USD 582,280. Reading down the reconciliation state changes that figure twice over, and in different ways.
| Component | Stated value (USD) | Share of total |
|---|---|---|
| Collectable as stated | 467,280 | 80.3% |
| Not collectable as stated — already paid and unallocated, credited, duplicated, or subject to a standing deduction | 37,350 | 6.4% |
| Undetermined — disputed, so the collectable amount is not yet knowable either way | 77,650 | 13.3% |
| Stated total | 582,280 | 100.0% |
Composition of the 37,350 not collectable as stated: Cash already received, never allocated — 4 invoices, 22,000 · Credit note agreed, never applied — 2 invoices, 1,960 · Duplicate raised for the same work — 2 invoices, 8,500 · Standing deduction never modelled — 3 invoices, 4,890. Disputed balances are held separately and deliberately: a dispute makes an amount undetermined, not uncollectable. Synthetic ledger — constructed data, not client data.
The disputed share is the honest one to sit with. In this constructed ledger 13.3% of the stated total is neither collectable nor written off — it is unresolved, and only the business can move it. An ageing view reports that money as simply late. It is the same boundary the month-end exception map codes as a contested balance — a category no volume of follow-up can clear.
Method, and what this does not show.
Corpus. A synthetic ledger of 60 open invoices across 18 accounts, one notional currency, stated total USD 582,280, published in full below. Synthetic and constructed by the author. Not sampled, not anonymised, not derived from a real business, not a market study. At Par has no customers and no client data exists to draw on.
Rubric. Four questions (a credibly expected date, a blocker, an owner decision, collectability as stated), each with a fixed set of permitted answers, applied identically to every invoice from each view. One required next action per invoice, assigned by a published precedence: Decide, then Fix, then Ask, then Wait, then Chase. Ageing buckets are 0–30, 31–60, 61–90 and 90+ days past due.
What was measured. Informational difference only: how many distinct operating states each ageing bucket contains; how many invoices are indistinguishable under ageing and materially different in state, at bucket level and at bucket-plus-size level; the distribution of required next actions; the composition of the ten oldest balances; where owner decisions fall in an age-ranked list; and how much of the stated total is not collectable as stated or undetermined. Version 1, 29 July 2026.
- What was not measured — and is not claimed. Nothing here measures whether anyone gets paid, or gets paid sooner. No days-sales-outstanding effect, no payment-speed effect, no cash-flow effect, and no collection-rate effect was measured, modelled or estimated.
- No statistical or industry-wide validity. The ledger is one constructed example, not a sample of any population. Nothing on this page supports a statement about receivables in general, about any sector, or about any market.
- The magnitudes are a property of the construction. Change the mix of disputes, promises and reconciliation problems and every number here moves. What does not move is the direction, and the reason is structural rather than empirical: a view cannot answer from fields it does not contain.
- Cost was not measured. Capturing and maintaining an operating state takes work. This benchmark says nothing about how much, or about whether it is worth it for a ledger of any particular size.
- The rubric is arguable. The four questions, the permitted answers and the action precedence are authored choices. A different set would produce different counts from the same ledger. Both are published so a reader can substitute their own.
Reproducibility. The complete ledger follows, with each invoice’s bucket, balance, operating-state class and required action as computed. Every number on this page is derived from these 60 rows by the rules above — nothing is typed in by hand.
| Invoice | Bucket | Days past due | Balance (USD) | Operating state (Q1 / Q2 / Q3 / Q4) | Action |
|---|---|---|---|---|---|
| INV-1042 | 0–30 | 4 | 12,400 | committed / no / no / yes | Wait |
| INV-1043 | 0–30 | 6 | 3,150 | none / evidence / no / yes | Ask |
| INV-1044 | 0–30 | 8 | 8,800 | cycle / no / no / yes | Wait |
| INV-1045 | 0–30 | 9 | 1,950 | none / no / no / no | Fix |
| INV-1046 | 0–30 | 11 | 24,000 | committed / no / no / yes | Wait |
| INV-1047 | 0–30 | 12 | 4,600 | none / no / no / yes | Chase |
| INV-1048 | 0–30 | 14 | 15,750 | none / dispute / yes / yes | Decide |
| INV-1049 | 0–30 | 15 | 2,300 | none / evidence / no / yes | Ask |
| INV-1050 | 0–30 | 17 | 6,700 | committed / no / no / yes | Wait |
| INV-1051 | 0–30 | 18 | 9,900 | cycle / no / no / no | Fix |
| INV-1052 | 0–30 | 20 | 1,200 | none / no / no / yes | Chase |
| INV-1053 | 0–30 | 22 | 33,500 | committed / no / no / yes | Wait |
| INV-1054 | 0–30 | 24 | 5,400 | none / no / no / yes | Chase |
| INV-1055 | 0–30 | 25 | 7,250 | none / evidence / no / yes | Ask |
| INV-1056 | 0–30 | 26 | 18,200 | cycle / no / no / yes | Wait |
| INV-1057 | 0–30 | 27 | 2,750 | none / no / no / no | Fix |
| INV-1058 | 0–30 | 28 | 4,100 | none / no / no / yes | Chase |
| INV-1059 | 0–30 | 29 | 13,600 | committed / no / no / yes | Wait |
| INV-1060 | 0–30 | 30 | 890 | none / delivery / no / yes | Ask |
| INV-1061 | 0–30 | 30 | 6,300 | none / no / no / yes | Chase |
| INV-1020 | 31–60 | 33 | 9,450 | committed / no / no / yes | Wait |
| INV-1021 | 31–60 | 35 | 9,200 | none / no / no / yes | Chase |
| INV-1022 | 31–60 | 37 | 9,800 | none / dispute / yes / yes | Decide |
| INV-1023 | 31–60 | 39 | 9,600 | none / no / no / no | Fix |
| INV-1024 | 31–60 | 41 | 9,350 | none / evidence / no / yes | Ask |
| INV-1025 | 31–60 | 43 | 21,000 | none / no / no / yes | Chase |
| INV-1026 | 31–60 | 45 | 14,300 | cycle / no / no / yes | Wait |
| INV-1027 | 31–60 | 46 | 5,100 | none / no / yes / yes | Decide |
| INV-1028 | 31–60 | 48 | 7,800 | none / evidence / no / yes | Ask |
| INV-1029 | 31–60 | 50 | 16,400 | cycle / no / no / no | Fix |
| INV-1030 | 31–60 | 52 | 48,000 | committed / no / no / yes | Wait |
| INV-1031 | 31–60 | 54 | 2,600 | none / no / no / yes | Chase |
| INV-1032 | 31–60 | 55 | 11,200 | none / no / no / yes | Chase |
| INV-1033 | 31–60 | 57 | 3,900 | none / evidence / no / yes | Ask |
| INV-1034 | 31–60 | 58 | 6,050 | none / no / no / no | Fix |
| INV-1035 | 31–60 | 59 | 8,350 | none / no / no / yes | Chase |
| INV-1036 | 31–60 | 60 | 1,450 | none / delivery / no / yes | Ask |
| INV-1037 | 31–60 | 60 | 12,750 | committed / no / no / yes | Wait |
| INV-0981 | 61–90 | 63 | 27,500 | none / dispute / yes / yes | Decide |
| INV-0982 | 61–90 | 66 | 4,700 | none / no / yes / yes | Decide |
| INV-0983 | 61–90 | 68 | 3,300 | none / no / no / no | Fix |
| INV-0984 | 61–90 | 71 | 12,800 | none / evidence / no / yes | Ask |
| INV-0985 | 61–90 | 74 | 5,600 | none / no / no / yes | Chase |
| INV-0986 | 61–90 | 77 | 19,400 | cycle / no / no / yes | Wait |
| INV-0987 | 61–90 | 80 | 2,150 | none / no / no / no | Fix |
| INV-0988 | 61–90 | 83 | 15,900 | committed / no / no / yes | Wait |
| INV-0989 | 61–90 | 85 | 1,750 | none / no / no / yes | Chase |
| INV-0990 | 61–90 | 87 | 22,600 | cycle / no / no / no | Fix |
| INV-0991 | 61–90 | 89 | 4,250 | none / evidence / no / yes | Ask |
| INV-0992 | 61–90 | 90 | 6,900 | none / no / no / yes | Chase |
| INV-0903 | 90+ | 96 | 8,400 | none / dispute / yes / yes | Decide |
| INV-0904 | 90+ | 104 | 3,050 | none / no / no / yes | Chase |
| INV-0905 | 90+ | 112 | 7,150 | none / no / no / no | Fix |
| INV-0906 | 90+ | 121 | 11,800 | none / no / yes / yes | Decide |
| INV-0907 | 90+ | 133 | 1,300 | none / delivery / no / yes | Ask |
| INV-0908 | 90+ | 148 | 5,750 | none / evidence / no / yes | Ask |
| INV-0909 | 90+ | 161 | 16,200 | none / dispute / yes / yes | Decide |
| INV-0910 | 90+ | 178 | 2,450 | none / no / no / no | Fix |
| INV-0911 | 90+ | 195 | 990 | none / no / no / yes | Chase |
| INV-0912 | 90+ | 214 | 9,300 | committed / no / no / yes | Wait |
Synthetic ledger — constructed data, not client data. Cite as: At Par, Receivables Operating-State Benchmark v1 (29 July 2026), synthetic corpus of 60 invoices.
Where At Par fits — and the limits.
At Par maintains the operating state described here as part of ordinary bookkeeping, not as a separate collections product: every open invoice carries an owner, a last and next action, any commitment with the person who gave it, whether that commitment held, dispute status kept apart from collection status, and a reconciliation state so a balance already settled in cash stops being chased. A qualified accountant (ACCA) is accountable for the work.
That sits inside bookkeeping and month-end close and feeds management reporting, so the receivables position in the report is the one being worked. The decision about which parts of a receivables process can be handed over at all is a separate question, covered in the guide to accounts receivable outsourcing. Why the ageing instrument itself cannot answer the arrival question is set out in why an ageing report does not tell you when cash will arrive, and the invoice follow-up health check runs the same diagnostic on your own process in the browser.
Asked about the benchmark, and about what it does not prove.
Is this receivables benchmark based on real client data? +
No. It runs on a constructed ledger of 60 open invoices built specifically for the exercise. Nothing is sampled from, anonymised from or derived from a real business — At Par has no customers, so no client ledger exists to draw on. The full corpus and the classification rules are published on the page so the counts can be checked or re-run against a different mix. It is a demonstration of an informational difference, not evidence about any population of businesses.
What does a receivables operating state contain that an ageing report does not? +
Eight things the ageing calculation has no field for: who owns the next action, what the last action was and when, what the next action is and on what date, any promise to pay and the person who gave it, whether that promise was kept, whether a dispute is open, whether the customer is missing something they need in order to pay, and whether the stated balance reconciles to what will actually arrive. Ageing holds the invoice, the balance and elapsed days.
How many invoices in the benchmark are indistinguishable under ageing but different in operating state? +
All 60 of them share an ageing bucket with at least one invoice in a materially different state — 404 such pairs in total. Under a stricter test that also matches on size, 28 of the 60 have a near-twin in the same bucket with a balance within 15% and a different operating state, across 31 pairs. Every one of the four ageing buckets in the ledger contains all five distinct required next actions. These counts are from synthetic data and describe this constructed ledger only.
Does a better receivables view make customers pay faster? +
That was not measured and is not claimed. This benchmark compares what two views can tell you about the same invoices; it observes nothing about payment behaviour, days sales outstanding, collection rates or cash flow, because a constructed ledger cannot observe those things. The defensible conclusion is narrower and still useful: with operating state recorded, the invoices needing a decision, an internal correction, a specific request or simply patience can be separated. Whether a customer then pays remains the customer’s decision.
How much of the benchmark ledger is not collectable as the ageing report states it? +
In this constructed ledger, 6.4% of the stated total — USD 37,350 of USD 582,280 — is not collectable as stated: cash already received and never allocated, a credit note agreed and never applied, duplicates, and standing deductions never modelled. A further 13.3% is disputed, which makes it undetermined rather than uncollectable. The figures are a property of how this ledger was built and are not an estimate of any real business.
What did the benchmark find about sorting receivables oldest-first? +
The ten oldest balances in the synthetic ledger break down as 1 Wait, 2 Ask, 2 Chase, 2 Fix, 3 Decide — so only 2 of the 10 warrant the follow-up an age sort implies. Two need an internal correction with no customer contact at all, and three need a decision from the business. Meanwhile 5 of the 8 invoices requiring an owner decision sit outside the top ten entirely. Age is a weak proxy for what an invoice needs.
Can this benchmark be reproduced? +
Yes, and it is designed to be. The complete 60-invoice ledger is published, along with the four questions, their permitted answers, the bucket definitions and the precedence rule that assigns one required action to each invoice. Every figure on the page is computed from those rows by those rules rather than typed in. A reader who disagrees with the rubric can substitute their own questions or a different mix of states and recompute — the construction is stated openly precisely so it can be argued with.
Does this prove anything about receivables in general? +
No, and the page says so directly. One constructed ledger is not a sample of any population, so nothing here supports a statement about receivables across an industry, a market or a size of business. The magnitudes depend entirely on how the ledger was built. What does not depend on the construction is the direction — a view cannot answer questions from fields it does not hold — and that is a structural property rather than an empirical discovery.
Bring your open invoices. We’ll return the state of each one.
Export the receivables list however it comes out. We will mark each balance with an owner, a last and next action, any commitment and whether it held, dispute status, what the customer is still missing, and whether the balance reconciles to what will actually arrive — then tell you which ones nobody should be chasing.
Prefer to reach us directly? Tell us a little and we'll come back with a time.
Test our receivablesWe use your details only to prepare for and hold this call. No spam, ever.
Reviewed by an ACCA on the At Par team · Last updated 29 July 2026 · The benchmark on this page runs on synthetic, constructed data; it measures information, not payment outcomes. See what we actually do.