Partly Paid Invoices and the Coverage Number

Why this matters

Coverage on part-paid invoices is one of the few collections figures that can improve because something bad happened. It is not a rate about your invoices. It is a statistic about a population whose membership changes every week, and an invoice can leave that population by being paid off, by being written off, or by being sent to an agency. Two of those three are good news and one is not, and the coverage number reads identically for all three. A shop that reports coverage rising without reporting what left the population has reported nothing.

What the population is, and who is not in it

An invoice is in this population only if it carries at least one payment and still has a balance owing. That excludes two much larger groups:

• Invoices with no payment at all. The genuinely unpaid ones. Not in the population, and they are usually the bigger problem.
• Invoices paid in full. They left the moment the last payment landed, taking their 100 percent coverage with them.

So three numbers describe this population and none of them can be read without the other two:

1. The count of part-paid invoices, stated as a share of open invoices.
2. The total still owed across them, stated as a share of the whole open receivable.
3. Blended coverage, meaning total paid divided by total billed on those invoices only, a weighted rate rather than an average of the invoices' own percentages.

In the shop used throughout this card: 118 open invoices, of which 92 carry no payment and 26 are part-paid. That is 26 of 118, so 22 percent of the open invoice count. The balance still owed across those 26 is 18 percent of the whole open receivable. Coverage is 41 percent.

Note that 22 percent is a share of the invoice count and 18 percent is a share of the money. They are close here by coincidence, and in a shop taking deposits on large jobs only they will not be close at all.

Coverage is a weighted rate over a membership that moves

Compute coverage as total paid over total billed across the population, which is a weighted rate. Do not compute it as the average of each invoice's own coverage percentage, which is an unweighted mean of ratios and a different number whenever invoice size and coverage are related. In this population the unweighted mean of the 26 per-invoice ratios is 46.3 percent against the weighted 41.0 percent, a gap of 5.3 points, because the low-coverage invoices here happen to be the large ones. The sibling card on averages of averages owns why that gap exists and what has to be true for it to open; the point here is only to pick the weighted version and say which one you picked.

The harder property is the membership. Coverage can move without a single customer changing behaviour, in both directions:

• An invoice paid off leaves the population, taking a high coverage figure out of it, which pushes coverage down.
• An invoice written off or handed to an agency leaves the population, and those are the low-coverage ones, which pushes coverage up.
• A new deposit taken on a large job enters the population at whatever the deposit percentage is, which pulls coverage toward that percentage.

Any one of those can dominate a quarter.

The quarter coverage rose nearly 18 points and nobody paid anything

The 26 part-paid invoices, split by how they were actually behaving. All shares below are of the part-paid population's own billed total, not of the whole receivable.

Coverage on A is 0.045 over 0.38, which is 11.8 percent. On B it is 0.365 over 0.62, which is 58.9 percent. Across all 26 it is 0.41 over 1.00, which is 41.0 percent.

In the following quarter the office cleared the book: the 7 stalled invoices in Group A were written down and handed to an agency. Nothing else happened. Nobody in Group B made an extra payment, and no new part-paid invoices entered.

Coverage now reads 0.365 over 0.62, which is 58.9 percent. It rose 17.9 points from 41.0. Not one additional unit of money arrived.

Watch what the second number did at the same time. Money still owed inside the population went from 0.59 of the original billed total to 0.255, a fall of 0.335 - and every bit of that fall was an exit, not a payment. Reported on its own, "coverage up 18 points and outstanding balances down by more than half" is a sentence about a good quarter. What it describes is the shop giving up on 0.335 of that billed total, which is 57 percent of the money the population still owed at the start.

The rule that prevents this: never report coverage without reporting the count and what left the population. Three lines, every time. Count at the start, count at the end, and how each departure left.

The two-by-two

Coverage against count, and the four corners have four different causes. Three of them have an owner; the fourth is the one where the answer is to leave it alone. Boundaries below are starting points; set your own once you have four quarters of your own history.

The example shop above is in no corner at either review, and that is the first thing the grid tells you: 22 percent part-paid against a 25 percent line, coverage 41 percent against a 35 percent line, so both coordinates sit inside the boundaries and you read the rows rather than the corner. By the second review the part-paid count is 19, and with the open invoice count otherwise unchanged that is 19 of 111, so 17 percent, at 58.9 percent coverage. Both coordinates have moved toward the top-left, the corner that needs nothing, and they moved there by write-off rather than by payment. That is exactly why where you land has to be checked against what moved you.

What each corner actually asks for

Top-left needs nothing. Do not add a report.

Top-right is a follow-up cadence on final balances, not a deposit change. The customers here already paid you once, so they can pay and they are willing. The final invoice is sitting in somebody's inbox. This is the cheapest money in the whole card to collect and it is almost always neglected because the account does not look delinquent.

Bottom-left is one phone call per account by the owner, since under 10 percent of open invoices is at most eleven of them on a book the size of the example, and a decision on each within the quarter: plan, write down, or hand off. These accounts do not improve by waiting, because a customer who paid a token amount and stopped has already told you what they are going to do.

Bottom-right is the only corner that needs a policy change, and the change is enforcement rather than design. Most shops in this corner have a perfectly reasonable deposit policy written down somewhere and no step in the scheduling process that checks it. The fix is a gate, not a memo.

Two shops where the number says nothing

The metric collapses at both ends of the deposit spectrum, and knowing that saves you from putting it on a report where it will only ever generate arguments.

A shop that takes no deposits and bills on completion. Its part-paid population is two or three invoices out of a hundred open ones, and those are almost always customers who part-paid after a dispute. Coverage over three invoices swings 20 points or more on one payment landing, so it carries no trend at all. Read the invoices themselves, and read the aging ladder. Coverage belongs nowhere near that shop's monthly pack.

A shop that takes a third up front on everything. Nearly every open invoice above a token size is part-paid by construction, so the population is most of the book and coverage sits close to the deposit percentage plus whatever progress payments have landed. It barely moves, and when it does move it is telling you about job mix rather than about collection. The useful reading there is coverage split by how old the invoice is: a balance part-paid 60 days ago at the same coverage as one part-paid last week is the row that matters, and the blended figure hides it.

Between those two ends there is a floor worth stating: do not read coverage as a trend on fewer than about 15 invoices in the population, because at 15 a single average-sized invoice leaving at zero coverage takes a fifteenth of the billed total out from under an unchanged paid total, which lifts the rate about 3 points on its own. That floor is stated on invoices in the part-paid population and is not the 30-record floor the ranking cards use. Below it, read the rows. The worked example above clears the floor at both reviews, with 26 invoices at the first and 19 at the second, which is why it was legitimate to read the change at all rather than just the rows.

The settings this number is really about

Coverage is downstream of three decisions. If the number is wrong, one of these is wrong.

• When a deposit is required. A common starting point is a deposit of about a third on any job priced above roughly three times your average ticket, taken before the job is scheduled rather than before the crew arrives. Taking it at scheduling is the part that matters: at that moment the customer wants the date, and on the morning of the job they want the work to start and the leverage has moved.
• How progress billing is milestoned. On jobs running longer than about two weeks, bill at defined milestones with payment due before the next milestone begins, and name the person who checks that before the crew rolls. A milestone with no gate is a schedule, not a billing term.
• When a part-paid account goes on hold. Hold new scheduling at 30 days past due on a part-paid balance. That is deliberately earlier than the general aging ladder's day-45 credit hold, because a customer who paid a deposit and then stopped has given you information that a customer who never paid anything has not, and because you usually have a scheduled job to hold, which is leverage the general case lacks.

What changes the answer: a shop doing mostly warranty or insurance-funded work has a part-paid population whose coverage is set by someone else's payment schedule entirely, and none of the three settings above applies. Track those separately or they will sit in the bottom-right corner permanently and make a working process look broken.

References

• See related: universal-gross-margin-percent-is-an-average-of-averages, which owns why an unweighted mean of ratios differs from the weighted rate and what has to be true for the gap to open.
• See related: universal-the-aging-buckets-and-what-each-one-actually-costs, for the past-due ladder these balances also sit inside.
• See related: universal-days-sales-outstanding-and-the-mismatch-inside-it, for how deposits compress the whole-book collections figure without anybody paying faster.
• See related: universal-the-close-rate-improved-because-they-stopped-quoting, which owns the general rule that a ratio cannot be read without the counts underneath it.