If not DSO, what measures AR performance?

Part 4 ended with shift C. Overdue grew across the book, fastest in the oldest bucket, and AR found it was theirs to fix. DSO needed a control chart and a split to get there. The ageing showed it directly.

That leaves Part 1’s other question. Not why DSO moved, but whether the team collecting the cash is doing a good job. If DSO can’t answer that, what can?

The usual answer: CEI

Ask what to use instead of DSO and the usual answer is the Collection Effectiveness Index. The Credit Research Foundation says “CEI can be used as a departmental or individual performance measure.”

Current means not yet past due. In plain words: of what customers should have paid by month end, how much did they pay?

It fixes DSO’s terms problem. Customers who pay on time score 100% on any terms, and a big sales month doesn’t move it.

It brings a problem of its own. This part assumes performance is measured once a month, from the month-end balances. CEI then takes overdue from the last day and never asks how late the collected invoices were.

Suppose a customer on 30 days end of month terms, so each invoice is due on the last day of the following month. They always pay about 25 days late. The March invoice, due 30 April, is paid around 25 May. On 31 May nothing is overdue: March is paid, and April is due that day.

Customer pays	CEI	Paid on time
On the due date	100%	100%
1 day late	100%	0%
25 days late	100%	0%

One customer, one invoice a month, measured at each month end. Made-up numbers.

CEI can’t tell one day late from 25. Neither can the month-end ageing. Collectors see this customer late most of every month in the live ageing. The scorecard never does.

Write off a bad debt and CEI goes up. Failing to collect counts as collecting.

Four measures

No single number answers the performance question. Most lists of AR KPIs agree, then give you fifteen. Four is enough for a manager, and each passes the test DSO fails: change the terms or have a big sales month, and they don’t move.

Overdue and ageing. Overdue in money at month end, by ageing bucket. Overdue is a red flag from the first day, and the buckets show whether it is getting older. As a share of total receivables, it makes a quick pulse check.

Write-offs. Money AR didn’t collect. Low isn’t always good: it can mean old debt kept on the books, and the oldest bucket shows that. Credit notes clear overdue without cash too, so know how much of a fall in overdue came from credits.

Paid on time. Of invoices paid in the month, the share paid by the due date, by value and by count. A customer who scores well by count and badly by value may be paying the small invoices and holding the big ones. Perhaps their own on-time target counts invoices. Worth a conversation.

Days late when paid late. Of invoices paid late in the month, the average days past due, weighted by value. Paid on time says how often customers are late, this says how late. If the near misses start paying on time, this goes up, because only the very late ones are left.

Both payment measures count from the day the cash arrived, not the day it was applied, and both look only at invoices paid in the month. Unpaid ones are the ageing’s job. Because they are built from every payment, not from the last day’s balance, they see what the month-end snapshot misses.

Why it takes four

Each of the four can be fooled in a way another one catches.

Scenario	Overdue and ageing	Write-offs	Paid on time	Days late
A customer always pays 25 days late	Nothing at month end	No change	0%	25 days
Late invoices go unpaid instead	Up	No change	Improves	No change
A large customer pays the day after month end	Spike in 1–30 days	No change	Dips, next month	1 day, next month
Old debt is written off	Improves	Up	No change	No change
A 200-day invoice is finally paid	Oldest bucket falls	No change	Dips	Jumps
Shift C	Up, oldest bucket fastest	Up later	Down	Up

In each row, one measure points the wrong way and another explains it. Late invoices that go unpaid drop out of paid on time, so it improves, and the ageing shows where they went. A 200-day invoice only enters days late once it is paid, so the average jumps, and the oldest bucket shows why. Read alone, any one of the four would send someone after the wrong thing.

Where they sit

DSO isn’t going anywhere. It answers how quickly sales turn into cash, and that is still the question at the top. Part 3 split it into sales and receivables. Part 4 split receivables into not yet due and overdue. Not yet due is mostly terms and customer mix. Overdue is AR’s, and the four sit under it.

Below the four is how the team got there: disputes and how long they take, promises to pay and whether they were kept, whether anyone picked up the phone. A dunning process that sends a third email to a customer who ignored the first two is not much of a process. That level has measures of its own, for another day.

Checklist: is your scorecard measuring AR?

Would a change in terms, or a big sales month, show up as better or worse collection?
Would a customer who always pays 25 days late show up anywhere?
Can overdue fall without cash arriving, and would you see it?
Do the payment measures start from when the cash arrived, not when it was applied?
Is it reviewed every month, whether DSO moved or not?

Back to the monthly review

Part 1 opened with DSO three days above forecast and “what happened?” going to AR. It is a fair question. It is a question about DSO. Parts 2 to 4 answer it.

The performance question gets the four measures, every month, whether DSO moved or not. In shift C, DSO took eight months to fire a run rule. If the ageing started moving in February, the scorecard would have shown it months earlier.

Part 2’s loudest signal was December, a spike gone by January. If that was a large customer paying a day after month end, the scorecard shows a jump in the 1–30 bucket and, in January, one day late. Nothing to fix. The loud signal was the least of AR’s worries. The quiet one was AR’s.

If not DSO, then what? Not one number. Four, read together, with DSO above them doing the job it was built for.