The Account That Was Lost on Paperwork, Not on Work

Why this matters

A shop finishes a twelve-month term at a municipal customer with a clean service record and is told the work is going out to bid, then is not renewed. Nobody can name a complaint. The technicians were liked, the equipment ran, the two returns in the term were both legitimate parts failures.

That shop asked the wrong question for a year. It measured itself on work quality and the customer measured it on something else entirely, and the whole answer was sitting in the shop's own accounts receivable aging the entire time. This is that reconstruction, done after the fact from records that were already there. The interesting part is not the cause. It is that every input needed to see it was in hand in month five.

What we had to work with

The account was gone, so there was no diagnosis to perform live. Four records existed: the shop's invoice register for the term, its accounts receivable aging by customer, its own work order history, and a short exit conversation with the facilities director who did not renew.

That last one produced one usable sentence and no explanation: "It was never about your technicians." Shops hear that as a kindness. Read literally it is a fact, and it eliminates an entire branch of investigation, which is why it is worth writing down verbatim rather than paraphrasing it into "they said it was fine."

The gaps in the four records turned out to matter as much as the contents, and they are noted where they appear.

Confirming the service file really was clean

Worth doing properly rather than taking the exit sentence at face value, because if the service record had been weak the whole reconstruction would point somewhere else.

Across the 118 work orders, first-visit completion ran about 79%, there were 2 return visits inside 30 days, no safety incidents, no missed emergency response, and no formal complaint logged in the customer's system. Those are respectable numbers on a municipal account, and the shop knew them, because it reviewed them monthly.

That is the trap in one sentence. The shop had a monthly performance review, the review looked at the metrics the shop believed the customer cared about, the metrics were good, and the shop concluded the account was healthy for eleven months. A dashboard that only carries the numbers you chose is a dashboard that will keep telling you the thing you already believe.

Reading the receivable aging before the service file

Start with the aging rather than the work orders, because the exit sentence has already ruled the work out.

The term ran 118 work orders and produced 74 invoices. Twenty-one of the 74, about 28%, were returned at least once by the customer's accounts payable and resubmitted. The first-pass invoices paid in an average of 38 days against net 30. The rejected-and-resubmitted invoices paid in an average of 79 days, roughly double.

The shape of the aging is the first real clue, and it is not the shape shops expect. The balance did not pile up past 90 days. It concentrated in the 60 to 90 day bucket, which is the signature of a rejection loop rather than a slow approver: a returned invoice does not sit still accruing age against the original submission, it restarts as a fresh submission with a fresh clock, so the aging fills the middle bucket and keeps refilling it.

Twenty-eight percent is the number to hold, and it is a share of invoices, not of billed hours.

Sorting the rejections by reason

Twenty-one rejections, sorted by what accounts payable wrote on the return:

• 13 for a missing or incorrect purchase order number
• 5 for a description that did not match the wording of the purchase order line
• 3 for a missing signed field ticket

Thirteen of 21 in one category, about 62%, is not a process running slightly loose. It is a single step failing repeatedly, and it points somewhere specific.

Here the second record gap appears. Pulling the 13 work orders behind those invoices, none of them recorded who authorized the work. Not a wrong name. An empty field. The technician had a verbal go-ahead from someone on site, did the work, and closed the ticket without capturing the name or a reference number, so the office had nothing to put in the purchase order field and guessed, or left it blank, or used the blanket number from a different cost center.

That relocates the fault. Everyone in the shop had been treating this as an invoicing problem and looking at the invoice template. The invoice template was fine. The capture step in the field was the failure, and it failed eight to twelve weeks upstream of where the symptom appeared.

What the customer's fiscal year did with the delay

A municipal fiscal year commonly closes on June 30, and an invoice that arrives after the close is generally paid from the following year's appropriation rather than the one the work was performed under, unless the department deliberately encumbered the funds. Whether it can be paid at all from the prior year depends on that encumbrance, which is precisely the thing a missing purchase order fails to create.

Nine of the 21 rejected invoices crossed the June 30 boundary. Those nine covered 138 of the account's 1,120 billed hours for the term, about 12%.

Now read that from the facilities director's chair. Her cost center closed the year roughly 12% under the work actually performed, which reads as an underspend and invites her line to be cut next cycle. The new year then opened already carrying that same 12% as an unplanned charge against a budget built without it. She had to explain a variance twice, in two directions, for work that was correctly performed and that she had correctly requested.

Nothing about that involves technician quality, and it is exactly the kind of thing a director is measured on. "It was never about your technicians" was not a kindness. It was the finding.

Checking the direction at both ends

Before accepting a mechanism, state what it predicts at the other end of its range, because a mechanism that only explains the case in front of you explains nothing.

The mechanism is that a rejection returns the invoice to the front of your queue and to the back of theirs, restarting the payment clock. So it predicts aging concentrated in the middle bucket with a high resubmission count, and it predicts that the delay scales with the rejection rate rather than with the size or complexity of the job.

At the other end: an account with a very low rejection rate and a balance concentrated past 90 days is a different fault entirely. That is an approval chain stalling, usually a department head who has to sign before accounts payable can release, and the fix is chasing the approver by name, not fixing the field ticket. Same symptom family, opposite mechanism, opposite remedy.

Testing it here: the rejected group averaged 79 days and the first-pass group 38 days on the same account, same approvers, same job types. The delay tracked rejection, not job size. The mechanism holds.

One honest note on that comparison. The 38-day first-pass average is itself above the 30-day term, so this customer ran slow on everything by about a week. The rejection penalty is the difference between the two groups, roughly 41 additional days, not the whole 79.

The month it became visible

Reconstructing month by month from the invoice register:

Months 1 through 3 produced 18 invoices with no rejections. The remaining 21 rejections all fall in months 4 through 12, across 56 invoices, a rejection rate of about 38% over that stretch.

Month 4 was the first month the account's first-pass acceptance fell below 90%. Month 5 was the second consecutive month below it. That is seven months before the non-renewal conversation, with the term's worst damage, the June crossings, still ahead.

Nobody was watching first-pass acceptance per account, because the shop's collections review looked at total receivables across all customers, where one municipal account's rejection rate disappeared into a book that was otherwise healthy. The number existed. It was never computed at the level where it meant anything.

The rule worth setting, and how big the response is

Trigger: any single account whose first-pass invoice acceptance falls below 90% for two consecutive months. Unit of analysis is one account in one calendar month, on a month containing at least 5 invoices, so a two-invoice month with one rejection does not fire it. The Boolean is two consecutive months, not one, because a single bad month is usually one person on vacation.

The response, sized: audit the field capture step for that account, not the invoice template. Pull every ticket behind the rejected invoices and check two fields, who authorized and what reference number they gave. If more than half the rejections trace to those fields being empty, the correction is a required field on the ticket that will not let a technician close a job on that account without a name and a number.

That is one change to one account's ticket, not a shop-wide process rewrite. Applied in month 5 here, it addresses the 13 purchase order rejections directly and leaves the other 8 untouched, which is worth stating plainly: it would have cut the rejection count by about 62%, not eliminated it. The description-mismatch group needs the purchase order wording copied onto the invoice line, and the missing-signature group needs a signature before the technician leaves site. Three separate fixes for three separate causes.

What it costs to get wrong in the other direction: setting the trigger at one month, or at any rejection at all, produces an audit every month on every account and the audits stop being done. The threshold exists to keep the response rare enough to actually happen.

The portable claim out of all of this: on institutional work your invoicing accuracy is a service level, visible to your customer as budget behaviour, and it is scored by someone who never sees your technicians. A shop can be excellent at the work and still be, from the only chair that decides renewals, unpredictable.

References

• See related: Why Work Done Without a Purchase Order Often Goes Unpaid, What a Purchase Order Actually Obliges Both Sides To, What a Vendor Portal Wants and Why Invoices Come Back
• See related: Net Terms and What They Do to a Small Shop's Cash Position, The Facilities Director and What They Are Actually Judged On
• Trade-standard practice for three-way matching in institutional accounts payable