Termco
Termco

The Deduction Code Knows Why

21 November 2026

The envelope from Halvorsen Ready Mix holds a check for $14,262.75, stapled to a remittance stub for invoice 8841. Your office sent that invoice for $14,512.25. The stub explains the gap in one line: ticket 7209, 18.2 tons of washed sand, paid at zero, deduction code SBD. In their system that code means signed backup missing. Your clerk has never seen their code key, so what she sees is a check that's $249.50 short.

Nothing about this is a dispute. Nobody called. Halvorsen's AP software matched your invoice against their own delivery copies, found one line it couldn't verify, zeroed it, and cut a check for the rest. The deduction cost their office nothing. What it costs yours gets decided at the cash application desk, in about ninety seconds, by a clerk with eleven more checks in the stack.

Three doors, one winner

She has three options, and she's taken one of them hundreds of times.

She can chase it. Pull invoice 8841, dig ticket 7209 out of the October box, check the signature line, call Halvorsen's AP, send the copy, follow up when nobody answers. Forty minutes, spread across a week, for $249.50.

She can leave it open. Apply the $14,262.75 and let a $249.50 stub balance ride on the invoice. Now 8841 ages. It surfaces on the thirty-day report looking like Halvorsen is slow, and somebody asks about it at the AR meeting. The explanation takes longer than the balance is worth, every month, until someone writes it off anyway.

Or she can write it off now. Most offices carry a tolerance for exactly this, written down or not. Under some number, $250 in one shop, $500 in another, the shortage posts to an adjustment account with a memo like "small bal w/o" and the invoice closes clean. Note what set the threshold. Nobody priced it. It's the number at which doors one and two stop being affordable, which makes it a workload setting, sitting in your books disguised as a credit policy.

At month end, with a close to finish, the third door is the only one open. The check posts, the aging stays readable, and the clerk did her job well by every measure her job is given.

The money leaves, and the reason goes with it

Every deduction is one of two things, and the write-off treats them identically.

Sometimes it's their error. Ticket 7209 went out signed; their scan was bad, or the copy landed in another job's folder. The deduction was wrong, and writing it off means you paid for their filing mistake.

Sometimes it's yours. The ticket really did leave the window unsigned, a night loadout maybe, where the signature line loses to a dark cab and an idling truck. Then the deduction is a correct audit finding about your scale window, delivered free, with a code on it naming the failure.

One of those deserves a contest and the other deserves a process change, and your ledger records both as "small bal w/o," a memo that describes the amount instead of the cause. A year later you can total what short pays cost you. You can't say which customers' deductions were legitimate, which failure mode repeats, or whether the unsigned-ticket rate is climbing. Halvorsen can. Their system holds your failure rate by code and by month. Your customers keep a better audit file on your tickets than you do, and they built it out of the lines you didn't contest.

flowchart TD
  A["Invoice 8841: 23 ticket lines"] --> B["Customer AP matches lines to delivery copies"]
  B --> C["Ticket 7209: no signature found"]
  C --> D["Check cut $249.50 short, code SBD"]
  D --> E{"Cash application desk"}
  E -- "write off" --> F["Posts as small balance: reason erased"]
  E -- "log and trace" --> G["Ticket pulled, verdict recorded"]
  G --> H["Their error: batch dispute with copies"]
  G --> I["Our paper: fix the night window"]

A tolerance is a price

AP departments notice which vendors contest and which don't. Not as a conspiracy, as a workflow: a deduction that comes back with documentation gets harder to repeat, and a deduction that vanishes into silence gets easier. Leave every short pay unworked for a few seasons and your tolerance stops being a bookkeeping convenience. It functions as a standing discount, one you never quoted, extended first to whichever customers run the most aggressive match.

Size it against your own volume. A yard sending 30 account invoices a month might see two or three short pays among them. Average them at $140 and you're near $4,500 a year, before you count the customers who learned the route. And notice the direction. In all the years your office has applied checks, how many came in over the invoice with a note saying your ticket undercounted? Deductions run one way.

Work the reason, not the dollars

The objection is obvious and half right: chasing $73 costs more than $73. True, if working a deduction means a phone call per check. So don't start with the call. Start with a log.

One line per deduction: date, customer, invoice, ticket serial, their code if the stub gives one, and a verdict after someone pulls the ticket: our paper, their error, or unknown. The verdict column is the whole point. When "their error" clusters on one account, you contest a quarter's worth at once, copies attached, and a documented batch gets read differently than a call about $73. When "our paper" clusters on a failure mode, you've found something worth more than the refund. Unsigned tickets turn out to come off night loadouts, or from one relief operator, and the fix happens at the window, upstream of every future invoice.

The step that kills this in a manual office is the middle one, pulling the ticket. Finding 7209 in a box is twenty minutes, so the verdict never gets rendered and the log never starts. That lookup is the part Termco removes. Tickets are digitized as written, signature box included, so tracing a deduction to its ticket is a search, not an afternoon. Better, the grounds for the deduction get rare. What each account requires rides in reference data. A ticket missing its signature is flagged as non-compliant the day it's read, and invoice totals assemble only from tickets that cleared, so every line you send has backup behind it. If you want to know how many of your tickets would survive your toughest customer's match, send 10-20 samples to hi@termco.ai, night loadouts included, and you'll have findings within 48 hours.

Either way, pull the adjustment account for last quarter and read the memo lines. Count how many tell you what actually happened. Somewhere across town, a customer's AP system has the answer for every one of them, coded and dated. It was on the stub the whole time.