The Close Scorecard: Five Numbers That Say Whether the Close Works

Most companies grade their close by how it felt. It felt tight this month, and last month felt like a scramble because the bookkeeper was visibly tired by the end of it. That is not a measurement, and a process nobody measures cannot be improved on purpose. It can only drift, and it usually drifts in the direction of whoever is least willing to stay late.

A close is a process with an output, and the output has properties you can count. Five of them, and none of them requires a system you do not already have. Every one is read off the ledger, the close calendar or the reconciliation index, all of which exist by the end of the month anyway.

The uncomfortable part first, because it changes how the rest of this reads: days to close is the only one of the five that most companies track, and on its own it is the least informative. A close that lands in six days and produces eleven adjustments after issue is worse than one that lands in nine and produces none. Speed is easy to buy by lowering the standard, and a scorecard with one row on it will be bought exactly that way within two quarters.

The five measures

Measure	What it counts	Where it is read from	What a movement in it means
Days to close	Business days from the period end to the moment the period was locked in the accounting system, not to the moment the package was emailed	The period lock date recorded in the ledger, against the calendar. If your system does not record it, the lock log in the close binder	Rising means a dependency has slipped or scope was added. Falling sharply with nothing else moving usually means work was dropped rather than accelerated
Post-close adjustments	Count and absolute value of entries affecting a period after that period was locked, including the year end adjusting entries your external accountant returns, traced back to the month they belong to	The journal entry listing filtered on posting date after the lock date, plus the adjusting entry schedule at year end allocated by period	The single best measure of whether the reported number meant anything. This is the one to watch first
Late dependencies	Count of close tasks that started late, each named with the input that was late and the person who owed it	The close calendar, with a planned start and an actual start column against every task	Tells you whether the close is slow or whether the close is waiting. Those two have completely different fixes and they are routinely confused
Unreconciled balances at lock	Count of balance sheet accounts that were locked without a reconciliation, and the total value carried in them	The reconciliation index in the close binder, compared line by line against the trial balance at lock	This is the leading indicator. Unreconciled balances this month become post-close adjustments in a later month, usually the twelfth
Reopened periods	Count of periods reopened after the package was issued, and whether statements were reissued each time	The lock log, which should record every unlock with a date, a reason and a name	Should be zero most years. A number above zero that nobody can explain is a governance finding rather than a close finding

Read the fourth row and the second row together and you have most of the value. Unreconciled balances are cheap to count, they are countable on the lock date, and they are the thing that turns into adjustments later. A company whose unreconciled count is falling while its adjustment count is still high is fixing the right problem and has not seen the result yet. A company with both rising is not closing, it is publishing.

Absolute value, not just count, on the second row

Counting adjustments alone rewards one large wrong entry and punishes ten small correct ones. Record both numbers on the same row: how many, and the sum of their absolute values. Absolute, not net, because two offsetting errors are two errors and netting them is how a file gets told it is fine.

The same applies to the value carried in unreconciled accounts. Twelve accounts holding immaterial residuals and one account holding a balance nobody can explain are different situations, and a count treats them identically.

There are no targets here, and you should be suspicious of anyone who gives you one

There is no target attached to any of the five above, and the omission is deliberate rather than evasive.

There is no defensible published figure for what days to close should be at a Canadian company of forty people, because the population is not measured and the companies in it are not comparable. A target borrowed from a benchmark built on larger organisations with dedicated finance teams does one of two things. Either it is trivially met and the scorecard becomes a decoration, or it is permanently unreachable and everybody stops looking at it by the fourth month. Both outcomes end with the scorecard being quietly dropped, which is a worse position than never having built one, because the next person to propose measuring the close will be told it was tried.

So the first three months are the baseline. You score them, you do not judge them, and you write down anything unusual about each one. The fourth month is the first month with something to compare against, and from then on the only comparison that matters is against your own trailing scores.

Set a target after the baseline exists, set it on one measure at a time, and set it as a direction with a review date rather than a number with a moral weight attached.

Who scores it, and when

On the lock date, by somebody other than the person who prepared the close. In a company at this size that is usually the reviewer, and where the reviewer prepared parts of the close as well, the scorecard says so on its face.

It is one page. Five rows, this month, the two prior months, and a short note against anything that moved. It is filed in the close binder with everything else, because a scorecard kept in an inbox has a life expectancy of about a quarter.

And it goes to the owner from the first month, not after the numbers look better. A finance function that keeps its own scorecard private until it is flattering has built an instrument for grading itself, which is a different thing with the same name. The first three scores will be uncomfortable in most companies. That is the point of a baseline, and an owner who is shown the starting position is much easier to talk to about the second month than one who is shown a trend that begins after the bad part.

Reading the scorecard when it disagrees with itself

The interesting months are the ones where the rows point in different directions.

Days to close falls, adjustments rise. Work was skipped. Look at the unreconciled count in the same month and it will usually confirm it.

Days to close rises, everything else improves. Somebody is doing the work properly for the first time. This is what the first month of a remediated close looks like and it should be left alone.

Late dependencies high, days to close flat. Finance is absorbing other people’s lateness by compressing its own work, which holds until the month it does not. This is the most common pattern in a company that has never published a close calendar, and the fix is not in finance. When we run the close ourselves this is the row we report on first, because it is the only one the client can act on without us.

Everything flat and good for six months. Check that the reconciliation index has not become a formality. Flat perfection in a growing company is usually a measurement problem rather than an achievement.

What we do not do

We do not set a target for a close we have not yet scored three times, and we say so when asked for one on day one.

We do not score a close we prepared and reviewed without stating on the page that both roles sat with the same party, because a scorecard whose author graded their own work is evidence of an opinion rather than of a process.

We do not net the adjustment values, and we do not report a count without the value beside it.

And we do not add rows for the things that are easy to count. The measures above are on the list because each one changes a decision, and a scorecard where four of nine rows change nothing teaches the reader to skim all nine.

What the scorecard does not measure

It says nothing about whether the reporting pack is any good, whether the variance commentary explains anything, or whether the controls around the close actually work. Those are real questions and none of them is a close metric. The scorecard measures whether the close produces a reliable number on a predictable day. It does not measure whether anybody is helped by the number afterwards.

Keep it at five rows. Every additional measure someone proposes is individually reasonable and collectively the reason the scorecard stops being filled in. If a sixth row genuinely earns its place after a year of scores, add it and drop one.