What a slow close is made of.
Time a close properly and the shape is consistent across organisations. A small amount of genuine judgement (accruals, provisions, valuation) sits at the end. Everything before it is investigation: reconciling two lists that ought to agree, establishing why they do not, and posting corrections.
In one recent engagement we classified every correcting journal raised over a quarter. Four causes accounted for 92% of them: a missing or wrong cost centre, a posting landing in the wrong period because a goods receipt was entered late, a foreign exchange rate applied at the invoice date instead of the receipt date, and item codes from two outsourced processing partners that had never been mapped.
Not one of those four needs an accountant. All four are rules. They were being caught in the second week of the close because that was the first moment anybody looked.
Accept-and-fix is an expensive default.
Most finance systems are built to accept a posting and let you correct it afterwards. That is a reasonable default for a general ledger and a poor one for an integration.
The cost of a defect rises steeply with how long it survives. Refusing a malformed vendor bill at the point of receipt costs a message to the supplier. Finding the same bill during the close costs a reconciliation, a correcting journal, a line in the variance commentary and (if it crossed a period) a restatement of a figure somebody has already seen.
The useful question is not how to close faster. It is what proportion of postings arrive correct, and what is stopping the rest.
Validate at the boundary.
The change is to move validation to the point of arrival, and to make rejection the ordinary outcome rather than an exception somebody has to authorise.
- A vendor bill that cannot be matched to a purchase order and a goods receipt does not post. It goes back with a reason rather than being accrued and chased.
- A posting with no cost centre is refused at the interface, not defaulted into a suspense account for somebody to clear later.
- Rates are applied from the event date carried on the transaction, not from the date the document happened to be keyed.
- Every posting carries the identifier of the thing that caused it, so a figure in the trial balance can be walked back to a movement without a spreadsheet.
None of this is sophisticated. It is unpopular, because it moves work from the finance team, who are paid to be careful, onto operational teams and suppliers, who are not. That is a genuine cost and it should be acknowledged rather than argued away: the first month after the boundary tightens is worse for everybody except the accountants.
Reconcile continuously, or not at all.
A reconciliation performed once a month is not a control. It is an audit of the previous month's failures. Performed daily, the same comparison becomes an exception list of a handful of items, each fresh enough that somebody still remembers the context.
Intercompany is the clearest case.
Two entities posting from two documents will always need reconciling. Two entities posting from one document have nothing to reconcile: a process change rather than a systems change, and usually the cheapest improvement on the table.
Measure arrival, not the close date.
The close date is a lagging indicator, and managing it directly produces the wrong behaviour: heroics in the first week, and a team that gets faster at correcting rather than an organisation that gets better at posting. Four measures are more useful, because they move first.
- The proportion of postings requiring no correction after arrival.
- The count of adjusting journals raised after day one, and their causes.
- The age of the oldest unreconciled item, checked daily rather than monthly.
- The proportion of vendor bills matched automatically to a purchase order and a receipt.
Move those and the close date falls out of it. In the engagement above it went from a range of eleven to fifteen working days down to two, and nobody worked on the close date itself.
What does not automate.
It would be dishonest to end there. Accruals for work performed but not yet invoiced require somebody to form a view on completeness. Provisions require an opinion about the future. Valuing anything illiquid requires an argument, and occasionally a negotiation with an auditor.
That is the two days. The goal was never to eliminate it, it is the part of the close that is actually accounting. The goal is to stop burying it under twelve days of finding out what happened.