What Share of Collections Arrives by Card
Why this matters
Card share of collections sits on the same page as revenue, invoices issued and average ticket, and it is not the same kind of number as any of them. Those are sales figures anchored on when work was billed. This one is a cash-timing figure anchored on when money arrived, and the money that arrives in a month was billed across several. Read it as a sales number and you will conclude that a quiet month was a strong one, or that a payment-offer change worked when what actually happened was one customer clearing an old balance.
Two clocks, and this number runs on the second one
Every figure in a shop's monthly pack runs on one of three clocks, and which clock it runs on decides what question it can answer.
- Issue date. What you billed. Invoices issued, booked sales, the collections-figure denominator.
- Payment date. What arrived. Total collected, card share, days to pay, card fees.
- Snapshot. What stands right now. Open receivable, aging buckets, part-paid balances.
Card share is computed by taking every payment received inside the window, whatever invoice it was against and whatever date that invoice carries, and splitting the total by how it arrived: card, check, bank transfer, cash. This card owns the arrival-date explanation for this group; the siblings below cite it rather than restating it.
The practical consequence is that the numerator and denominator of this figure can both be dominated by work performed and billed months ago. There is no version of it that is about this month's work, and trying to make one by filtering to same-month invoices produces a different metric with a different bias, covered at the end.
What is deliberately left out
Two things are excluded on purpose, and both exclusions change the number enough to be worth stating whenever you quote it.
Credit-memo applications are not money. Applying a credit to an open invoice settles a balance without any cash arriving. Leave those out of both the numerator and the denominator. Include them and a month of housekeeping looks like a month of collections, and the method split gets a category that is not a payment method at all.
Refunds are read separately, not netted in. A refund is cash going the other way and it belongs in its own line with its own count. Net refunds into the arrival total and a single large reversal quietly depresses the month's collected figure and distorts the method split, because a refund is nearly always returned by the method it came in on, so it lands entirely on one side of the split.
Deposits and prepayments, on the other hand, are money and belong in. They arrived.
Card is three channels, and they do not behave alike
"Card" as a single bucket hides the only thing this figure is useful for. Split it three ways, which costs nothing to set up and is the difference between a number you look at and a number you act on.
- At the door, card present. Dipped or tapped on a terminal in front of the customer. Fastest possible arrival, lowest fee tier, and the only one of the three a technician controls directly.
- Card not present: a payment link, a phone call, a keyed number. Prices above card present because the card was never physically read, and its arrival date is whenever the customer gets round to it, which can be any time from an hour to a quarter.
- Card on file, charged automatically. Arrival timing is set by you rather than by the customer, which makes it the only card channel with a predictable date - and the only one that needs a stored-credential authorization captured before you use it, because charging a saved card without the customer's express prior consent to that specific arrangement is a chargeback you will lose under the card networks' own rules (which are contract, not law), and where the saved credential is a debit card a recurring charge is a preauthorized electronic fund transfer requiring the customer's written authorization under Regulation E, 12 CFR 1005.10(b). Capture the authorization and the amount-change notice terms at enrollment, not at the first charge.
In the month worked through below, 0.23 of the 0.665 card total came through a payment link, so at least 35 percent of that month's card money was card not present. A shop at 61 percent card where most of it is at the door and a shop at 61 percent card where most of it is keyed or linked have different processing costs, different days to pay, and different levers. The blended figure cannot tell them apart.
A quiet month where collections rose and billing fell
All figures below are in collection-months, where one collection-month is this shop's trailing three-month average of money actually received.
| Trailing three-month average | Review month | |
|---|---|---|
| Money received | 1.00 | 1.09 |
| Received by card | 0.44 | 0.665 |
| Received by every other method | 0.56 | 0.425 |
| Card share of money received | 44.0% | 61.0% |
| Invoices issued | 1.02 | 0.78 |
| Jobs completed (count, vs the trailing average) | - | down 22% |
Card arrivals are 0.665 over 1.09, which is 61.0 percent. Collections rose 9 percent against the trailing average while invoices issued fell 23.5 percent and completed job count fell 22 percent. Money received exceeded money billed in the same month by a factor of 1.09 over 0.78, which is 1.40.
Every one of those statements is true simultaneously and there is nothing wrong with the shop's records. The month was quiet and the month collected well, because collections in any month are mostly the previous months' billing arriving.
Split the arriving money by when it was billed and the picture finishes itself:
| Billed when | Share of the 1.09 received |
|---|---|
| In the review month | 0.47, which is 43% of what arrived |
| In the month before | 0.39, which is 36% |
| Two or more months back | 0.23, which is 21% |
So 0.62 of the 1.09, which is 57 percent of what arrived, was billed before the month even started.
Splitting the 17-point move
Card share went from 44.0 to 61.0 percent, a rise of 17.0 points. Two separate things did that, and they call for different responses, so separating them is the work.
The 0.23 in the oldest billing bucket was one property-management account clearing two aged invoices through the payment link, on a company card. Strip that single event out: card falls to 0.435 and total collections to 0.86, giving a card share of 0.435 over 0.86, which is 50.6 percent.
- 10.4 points of the move (61.0 down to 50.6) was that one aged catch-up.
- 6.6 points (50.6 down to 44.0) was mix: a quiet month is a residential-heavy month, residential work pays at the door by card, and the commercial work that pays by check on terms was the part that went quiet.
Those add to the 17.0 points. Reported as a single number, "card share up 17 points" reads as a behaviour change across your customer base. It was one payment and a mix shift, and neither is a behaviour change.
The inclusion map
Before you put two of these figures in a sentence together, check they are on the same clock. Most of the bad readings in a monthly pack are one clock compared against another.
| Figure | Anchored on | Credit applications | Refunds |
|---|---|---|---|
| Card share of collections | Payment date | Excluded, no cash moved | Separate line, not netted |
| Total collected | Payment date | Excluded | Separate line, state which convention you used |
| Invoices issued, booked sales | Issue date | Reduce sales in the credit's own period | Not applicable |
| Collections-figure denominator | Issue date | Netted out of window sales | Not applicable |
| Aging buckets | Snapshot, aged from due date | Reduce the open balance | Reduce the open balance |
| Days to pay by method | Payment date, measured back to issue date | Excluded | Excluded |
| Effective card rate | Payment date | Excluded | Fees stay taken; read the fee line, not the net |
The decisions this number can carry, and the one it cannot
It cannot carry a read on demand or on the health of the month. It is not a sales figure and it has no denominator in common with one.
It can tell you where your money is actually coming in, which is the input to two real decisions: what your effective processing cost is going to be next month, and how much of your collections depends on a channel you do not control. A shop at 61 percent card has a materially different cost structure and a materially different dependency than one at 20 percent, and that is true regardless of why the figure moved.
It can tell you where to steer, but only alongside days to pay by method. Card share on its own says nothing about whether card is worth what it costs, because it does not contain the fee or the speed. The sibling card on days to pay by method puts those on one scale.
It can be used as a mix alarm. A steady figure that jumps or drops by more than about 10 points in a month almost always has a single identifiable event behind it, and finding that event takes about five minutes with the payments list sorted by size. Do that before anyone writes an explanation into the board pack.
Testing whether a payment-offer change actually moved anything
If you add a payment link to invoices, or start offering to keep a card on file, the arrival-anchored figure cannot tell you whether it worked, because the month after the change is full of money from invoices issued before it. Re-anchor the test deliberately:
- Take invoices issued after the change date that are now fully settled, and compute card share of the money on those invoices.
- Compute the same thing on invoices issued in an equal-length window before the change.
- Compare those two.
Wait before reading it. The minimum is one full terms cycle plus your median days to pay: on net 30 with a median of 22 days, that is about 52 days after the change. Reading it earlier restricts the sample to invoices that have already settled, and the fast settlers are disproportionately card payers, so an early read will show the change working whether it did or not. That is the same survivor bias the time-to-invoice card owns, arriving in a different costume.
What changes the answer: if the change also altered which customers get the offer, you are no longer running the same test. A payment link added only to residential invoices moves card share by changing the population, and the honest comparison is within residential invoices only, before and after.
References
- See related: universal-the-effective-card-rate-is-not-the-rate-you-were-quoted, for what this share actually costs you.
- See related: universal-days-to-pay-by-method-and-what-the-gap-buys-you, which puts the fee and the speed on one scale.
- See related: universal-days-sales-outstanding-and-the-mismatch-inside-it, for the issue-date figures this number sits beside.
- See related: universal-time-to-invoice-only-counts-the-invoices-you-sent, which owns the survivor-bias explanation referenced above.