Three kinds of bad day
Files get summarised in a breath: "a few NSFs, runs tight at month end". Inside that sentence are three separate measurements of three separate problems — a payment that was refused, a day that closed below zero, and a day that closed thin. They answer different questions, they fail for different reasons, and merging them into one impression throws away the part that actually predicts how the next advance will perform.
- A refusal, an overdraft and a thin day are not one metric
- The same payment bouncing four days running is four incidents
- A fee refunded is not an incident at all
Three measurements, three meanings
Each is counted from the statement in its own way, and each answers a question the other two cannot.
A returned item is a relationship event
Somebody presented a payment and the bank refused it. There is a counterparty on the other side who expected money and did not get it — a funder, a supplier, a landlord. It is the only one of the three that another party experienced.
A negative day is a hard fact
The account closed below zero. Not thin, not stretched: overdrawn, on a day the bank can name. It is the least arguable number in this group, and the one that most reliably means what it looks like.
A low-balance day is a margin, not a failure
The account closed under a threshold — typically $500, with a wider $1,000 band for context, and the sterling equivalents on a UK file. Nothing failed. The number describes how much room existed at the close of business, which is the room tomorrow’s debits have to land in.
The same total, two different merchants
Take two files with identical revenue and identical financing payments. The first closes below $500 on fourteen days and has no returned items at all. The second is comfortable most of the month and carries five returns.
The first merchant is running lean on purpose. They know what leaves and when, they keep the balance low because the money is working elsewhere, and nothing has ever been refused. That is a management style, not distress — and a stacking decision against them should be about whether another fixed debit fits in the gap, not about the size of the cushion they choose to keep.
The second merchant has a timing problem, and timing problems are how defaults start. Five refusals means five occasions when the order in which money arrived and left did not work out. If those refusals cluster in the weeks after a new position began, the file is telling you exactly what an additional daily debit did to a cash flow that used to cope. That is a different conversation from a thin cushion, and it is invisible if both files are summarised as “some stress”.
Four bounces of the same payment are four incidents
There is a tempting piece of tidiness here that we deliberately refuse. When an account is deeply overdrawn, the same recurring payment can be presented and refused on four consecutive days. It is one payment, so it looks like one event, and collapsing the four rows into one makes a cleaner-looking report.
It also understates the strain, in the direction that approves deals. Each presentation is a separate occasion on which the account could not cover what was asked of it, and often a separate fee. A merchant whose funder retried four times in a week is not in the same position as a merchant who bounced once, and the report should not flatten one into the other.
The opposite error gets corrected too. When a bank refunds an overdraft or NSF fee as a courtesy, the refund is the fee coming back — not a new incident. Counting it would inflate the number with the bank’s own goodwill.
One more piece of discipline sits underneath: the tile, the list you open from it, the month-by-month trend and the risk score all read the same set of rows. That sounds obvious until the day a count and its own drill-down disagree in front of an underwriter, at which point every other number on the page becomes a question too.
Day counts need a ledger that can be trusted
Returned items are read directly off transaction rows, so they survive almost any statement. The day counts do not. Counting days below a threshold requires knowing where the balance closed on every day of the period, and plenty of statements do not print a running balance on each line.
Where that column is missing the ledger has to be rebuilt from the opening figure and the transactions, and a rebuilt ledger is only published when it lands where the statement says it should. If it does not, the low-balance and negative-day counts are withheld rather than estimated, because a chain that has drifted produces a beautifully smooth series sitting at entirely the wrong level — the case for withholding a number is at its strongest exactly here.
What to look at once the three are separate
Where the returns sit. Evenly spread through the period, they are a feature of how the business runs. Concentrated after a particular date, they are an effect, and the cause is usually something that started on that date.
Where the thin days sit. Clustered around the end of the month is often just rent and payroll landing together. Scattered right through it means there is no part of the cycle with real headroom, which is the version that matters when you are adding a daily obligation.
Whether the overdrawn days are isolated. One negative day in a quarter is an accident. A run of them is a state, and no amount of revenue at the top of the file changes what that state means.
None of these is a decision rule, and none of them should be. They are the questions the three counts are good for once they stop being a single word in a summary.
Common questions
What counts as an NSF or returned item on the report?
A row the bank marked as insufficient funds, a returned or rejected item, a chargeback, or an unpaid item, including the credit a bank posts when a debited item bounces back. Fee refunds are excluded, since a refunded fee is not an incident.
Why do repeated bounces of the same payment each count?
Because each presentation is a separate occasion on which the account could not cover what was asked of it, and usually a separate fee. Collapsing them would make a deeply strained account look like it had a single bad day.
What is a low-balance day?
A calendar day that closed below a set threshold — typically $500, with a wider $1,000 band shown for context, and the sterling equivalents on a UK statement. It records how much room the account had at close, not that anything failed.
Is a low-balance day a red flag on its own?
Not by itself. Plenty of well-run businesses keep a deliberately thin operating balance. Low-balance days matter as headroom for what you are about to add, and they matter most when nothing in the month has any.
Why does a report sometimes not show low-balance days?
Because the statement did not print a running balance and the rebuilt daily ledger could not be checked against a figure the bank published. In that case the counts are withheld instead of estimated.
Related
Read the bad days apart
Upload a statement and see returns, overdrawn days and thin days as three separate counts.