Features
Part of Every expense answers three questions: when, where, and whose
Reading expense recording metrics without fooling yourself
Expense recording metrics track document coverage, accrual accuracy, recoding, holding-account age and entry lag; read them together, not as isolated scores.
Most businesses measure what they spend. Very few measure how well that spend is written down, and the second number decides whether the first one can be trusted. Expense recording metrics make that second number measurable.
Three properties are worth counting on the cost side: how much of the spend is evidenced, how accurate the judgements are, and how long everything takes. Coverage, accuracy, latency. All three come out of a transaction export and a folder of documents, and none needs a report your system does not already have.
Every figure below is invented to show how a calculation runs. Substitute your own.
What to take away
- Measure by value, never by transaction count. Ninety percent of items evidenced can still be half the money.
- Accrual accuracy is the one measure that tells you whether your estimates are worth making.
- A latency figure with no tail is not a measurement. The slow items are the whole finding.
- An old item in the holding account overrides every other reading, however good the rest look.
- Five numbers, taken monthly after the close, beat any single quarter-end review.
Coverage: how much of the money is evidenced?
Documented spend. Value of expense transactions with the supplier's own document attached, as described in IRS Publication 583, Starting a Business and Keeping Records, divided by total expense value.
Count vs Value Coverage
By count
- Documented
- 1,080
- Undocumented
- 120
- Total
- 1,200
- Coverage
- 90%
By value
- Documented
- 164,000
- Undocumented
- 46,000
- Total
- 210,000
- Coverage
- 78%
The by-value part matters. Suppose 1,200 transactions in a quarter, of which 1,080 have a document: 90%, which sounds respectable. Now value them. The 120 without documents total 46,000 against a quarterly expense total of 210,000, so by value the coverage is 78%. Those two numbers lead to different conversations, and only one of them is about the money.
Purpose recorded. Share of value where the document alone would not tell a stranger why the business bought the thing, and a note has been added. This applies to a small subset of spend and is worth measuring separately, because it is exactly the subset anyone reviewing will ask about.
Supply date recorded. Share of expense entries carrying a supply or service date distinct from the payment date. A low figure does not prove the periods are wrong. It proves nobody can check, which at a period end is the same problem.
Accuracy: are the judgements any good?
Accrual variance. For every accrual you reversed, compare the estimate with the invoice that replaced it. Report both the total difference and the largest individual ones.
| Month accrued | Estimate | Actual invoice | Difference |
|---|---|---|---|
| January | 4,000 | 4,180 | 180 |
| February | 2,500 | 2,460 | 40 |
| March | 6,000 | 9,200 | 3,200 |
| Quarter | 12,500 | 15,840 | 3,340 |
The total is not the story. Two of the three estimates were close and one was out by more than half. That one is worth an hour: it usually means the person estimating had no access to what was actually delivered, and the fix is a conversation with whoever did, not a better guess.
Recoding. Value of transactions moved between accounts after their original posting, as a share of expense value. Some of this is healthy correction. A high figure means the coding is being done twice, once by whoever entered it and once by whoever cleans up, and the cure is usually one paragraph added to the account definitions.
Duplicate payments found. Count and value. This is not really a measure of the books; it is a measure of the accounts payable process, and it belongs here because the ledger is where duplicates become visible. Search a year's supplier payments for the same supplier, same amount, close dates.
Holding account clearance. A holding account is usually a temporary balance-sheet clearing or suspense account for items not yet assigned to their final accounts. Measure the value cleared within thirty days and the age of the oldest item still there. To clear an item, match it to its supporting document and purpose, then post it to the appropriate account; do not leave it there or resolve it by guesswork.
Latency: how long does everything take?
Supply date to entry. Compute per transaction, then look at the count over a fortnight rather than the mean. Late entry is what pushes costs across period edges, and it clusters: usually one supplier, one card, or one person's claims.
Latency Along the Expense Path
- Supply dateservice delivered
- Entrytransaction posted
- Claimemployee submits
- Period endlast day of month
- Closenothing more posted
Claim submission lag. Days between an employee spending their own money and submitting the claim. A long lag moves costs into the wrong period and gets worse quietly, because nobody chases a claim except the person owed.
Period end to close. Days from the last day of the month to the point where nothing more will be posted into it. This is the number that makes every other measure possible, because an open period cannot be measured, only estimated.
Locking is what makes it real. IRS Publication 538, Accounting Periods and Methods, describes federal tax accounting periods and methods. For an application to adopt, change, or retain a tax year, see IRS Form 1128, Application To Adopt, Change, or Retain a Tax Year. The accounting periods and methods that apply are set out in the federal guide to accounting periods and methods.
Measures that mislead here
- Transactions processed per hour.It rewards speed at the exact point where a document should be read. A business can double this figure and halve the usefulness of its books.
- Cost per invoice processed, on its own.Useful only beside an accuracy measure. Otherwise the cheapest process is the one that never checks anything.
- Your expense ratios against another business's.Different mix, different boundary between direct and operating costs, different definitions. The comparison that means something is with your own last twelve months.
- The count of accruals.More is not better and fewer is not better. What matters is whether the ones you make are accurate.
Reading the three together
Read coverage, accuracy, and latency together; a single score can hide a problem. These patterns are diagnostic, not pass/fail benchmarks: this article sets no universal percentage or day target for coverage, recoding, duplicate payments, or latency. Compare consistent definitions month to month, then investigate the missing documents, coding changes, duplicate-payment candidates, or late items behind a movement.
| What you see | The usual cause |
|---|
Reading the Three Together
High coverage, high recoding?
documents collected, definitions unclear
Low coverage, low recoding?
- Strong coverage by transaction count but weak coverage by value:missing documents are concentrated among higher-value expenses. Start with the largest unsupported transactions.
- Documents are attached, but purpose is often not recorded:the paperwork alone does not explain why the business bought the item. Add a note for those transactions.
- Accrual variance is concentrated in one estimate:check whether the person estimating had information about what was actually delivered, then discuss the estimate with the person who did.
- Recoding is high and documented spend is weak:entries may be posted before the supporting paperwork is available. Clarify account definitions and get the needed context before posting.
- Supply-date-to-entry or claim-submission lag is long:costs may cross period edges. Look for a recurring supplier, card, or claimant.
- An old holding-account item remains unresolved:investigate its document, purpose, date, and correct account. This concern overrides otherwise favorable readings.
An old holding-account item means an unresolved question is accumulating. At year end, it may be settled by whoever is cheapest rather than whoever knows, so investigate and clear it instead of relying on a guess.
Turning the readings into a monthly routine
Use this five-number routine once a month, right after the close. Record the same measures and definitions each time:
- Documented spend by value:documented expense value divided by total expense value.
- Accrual variance:the total difference between estimates and replacement invoices, plus the largest individual differences.
- Recoding:value moved between accounts after original posting, as a share of expense value.
- Oldest holding-account item:the age of the oldest unresolved item.
- Supply date to entry:the count over a fortnight, rather than the mean.
The other measures in this article are useful follow-ups, but these are the five in the monthly routine.
Monthly Five-Number Routine
- Coverage by value, largest undocumented named
- Reversed accruals, estimate against actual
- Value recoded after posting, and accounts
- Oldest holding account item, in days
- Days from period end to lock
Five numbers, twenty minutes, and a trend that is worth more after six months than any single reading.
The routing decisions these measures help check are set out in three places:
- [Expense recording frameworkdeciding what an outgoing payment is](/expense-recording-framework)
- [Worked expense-recording examplesentries traced from the paperwork](/expense-recording-examples)
Common questions
Where do I get documented spend from?
Most systems can report on whether a transaction has an attachment. Where that is not available, sample rather than count everything: take the fifty largest expense transactions of the quarter and check those, which by value is most of the answer anyway.
What is a good coverage figure?
This article sets no universal target. Compare the figure with your own prior readings using the same definition, and look at the size and purpose of what is missing. Fifty thousand of undocumented spend is a problem whatever percentage it represents.
Should the person doing the entry see these numbers?
Yes, and they should own two of them: coverage and latency. Both are within their control. Accrual accuracy usually is not, because it depends on information from elsewhere.
How do these relate to the cost figures we already review?
They tell you whether those cost figures can be relied on. Running them once a month is what lets you read the expense report without silently discounting it.
