Scanning Paper Supply-House Invoices From the Truck

If you want to scan supply-house invoices from the truck, the honest ranking is: get the distributor to email you a PDF or turn on portal access, and photograph paper only for the tickets that never arrive any other way. A native digital feed beats the best phone capture you will ever do, because nothing has to be read back off a picture. Everything below assumes you cannot get that for every branch and every counter run, which is true for most shops.

Second thing up front: photographing a crumpled counter ticket works fine. The failure comes later, when the ticket goes in the door pocket and nobody looks at it again, so the unit price on it never becomes part of your history, which is the exact condition material price creep needs to survive.

So this post is the boring mechanics: your options, what optical character recognition genuinely gets wrong on a counter ticket, how to hold the phone, what routine survives a busy week, and how long to keep the paper.

Why the paper is still paper

Contractors get told to “just go digital” by people who have never stood at a will-call window at 6:40 in the morning behind four other trucks. Here is why the ticket is paper and stays paper.

The counter is a point-of-sale terminal and its native output is a printed ticket. Counter sales run as their own module inside distributor ERPs, separate from order management and delivery (Epicor ships Counter Sales as a discrete module). The person who signs is your tech, not your bookkeeper, and the branch hands the signer a copy. That is the whole mechanism.

Will-call and counter runs are also the highest-frequency, lowest-ceremony purchases you make. Nobody cuts a PO for two boxes of straps. No quote, no packing slip, no email trail. The ticket is the only document that will ever exist for that transaction, which makes it more important than the paperwork on your big orders, not less.

And the truck is a filing cabinet with a bad retrieval system. The ticket lives in the door pocket, migrates under the seat, and surfaces in a wad in mid-March. Any capture plan that costs more than the fifteen seconds between signing and getting back in the truck does not happen.

The capture options and where each one breaks

Five real options. Most shops run two, one for the counter and one for everything else.

The top two rows are a different category. They produce data. The bottom three produce pictures of data, and something has to read the picture.

Ask the distributor first, because a native feed beats any OCR

Before optimizing your phone technique, spend one call trying to make the problem go away.

Ask your branch or your rep for two things. First, email delivery of every invoice to a single address you control. Most distributor ERPs support invoice emailing per account, and it is a settings change on their side, not a project. Second, portal access with invoice history and, if it exists, a line-level export. The systems that run contract pricing and counter sales also keep your purchase history, and asking for a copy is a normal request reps field regularly.

Two practical notes. Ask at the account level, not per branch, or you get one branch emailing and three not. And ask whether counter and will-call tickets are included, because they frequently are not. That is the gap you will still be photographing.

If you only do one thing from this post, do this one. It is free, it takes one call, and it removes the recognition problem for the majority of your spend. It also gives you clean numbers for a real price review with your supply house.

What OCR actually gets wrong on a counter ticket

Recognition on a clean laser-printed invoice is a solved problem. A counter ticket is not. Here is where it genuinely struggles, and none of this is a knock on any particular app.

Thermal paper fade. Most counter printers are thermal. Thermal print fades with heat, sunlight and time, and a truck dashboard supplies all three. A ticket that sat on the dash through July can lose enough contrast that the characters are gone, not just hard to read. No software recovers ink that no longer exists. That is the strongest argument for capturing at the counter rather than at month end.

Carbon and NCR copies. The customer copy of a multi-part form is the faintest in the stack. Light blue or pink stock lowers contrast further, and the impression smears where the pen pressed hard.

Handwritten job numbers. The most valuable field on the ticket is usually the one a human wrote: the job name the tech gave the counter. Handwriting recognition on a scrawled “Ridgeline 2” in a cramped box is the least reliable part of the whole job, and it decides whether the cost lands on the right job.

Ambiguous characters in part numbers. Zero versus capital O, one versus capital I versus lowercase l, five versus S, eight versus B. Distributor catalogs are full of manufacturer catalog numbers mixing letters and digits with no pattern a reader can use to disambiguate (Trade Service, now Trimble, distributes those catalog numbers and UPCs as the underlying product data). One wrong character turns a repeat purchase into a brand new item, and a brand new item has no price history, which quietly defeats the entire point.

Multi-column layouts. Tickets put quantity, description, unit price and extended price in columns, sometimes with a second block for tax, freight and core charges. Recognition reads the characters fine and then associates the wrong price with the wrong line, especially when a description wraps to two lines and the numbers do not.

Wrinkles, folds and oil. A fold through the price column bends the characters. Hydraulic oil and cutting fluid make paper translucent so the back of the ticket bleeds through. Both are normal truck conditions.

Bad photos. Angle causes keystoning, so the far end of the ticket is smaller and blurrier. Overhead fluorescent light on glossy thermal stock gives you a bright wash across the middle. Shadow from your own phone lands right where you are trying to read. And a photo with three tickets fanned out in it is the worst of all, because nothing knows where one document ends and the next starts.

The pattern in that list: recognition rarely fails loudly. It fails by producing a plausible wrong number. That is why the checks in three-way match for contractors still matter after capture, and why misaligned line-item descriptions show up as a named failure mode in ordinary AP matching practice too (Tipalti, Stampli). A captured price still has to be compared against something.

The technique that fixes most of it in ten seconds

Six habits, and they matter more than which app you use.

1. Flat. Put the ticket on the counter, the hood, the toolbox lid, anything rigid. Smooth the fold with your palm. A flat ticket is worth more than a better camera.
2. Straight on. Hold the phone parallel to the paper, not tilted. The ticket should look like a rectangle in the frame, not a trapezoid.
3. Even light. Daylight or shade beats a single overhead fixture. Move so your shadow and the phone’s shadow are off the ticket. If there is a glare stripe, turn the ticket ninety degrees.
4. Whole ticket in frame. Include all four edges with a little margin. Cropping the header cuts off the invoice number and date, which are the two fields every later check depends on.
5. One ticket per image. Always. If you have four, take four pictures. A stack in one frame is unusable data no matter how sharp it is.
6. Look at it before the phone goes away. Two seconds. Can you read the unit price column? If you cannot, nothing else will.

A scanner app helps mainly because it does edge detection, deskew and contrast for you, and because it nudges you into one document per capture. Real benefit, not a huge one. Good technique with the plain camera beats sloppy technique with any app.

The routine that survives a busy week

Capture discipline is a routine problem, not a technology problem. Three rules.

Capture at the counter, before the ticket goes in the door pocket. This is the whole ballgame. Signing and photographing is one motion at the counter, and a fifteen-minute treasure hunt three weeks later.

One destination, and only one. Pick a single address or folder and make it the only place tickets go. If some techs text photos to the office, some email them, and some drop paper in the tray, you get things entered twice and things that arrive nowhere. That split is a documented way the same supply-house invoice gets paid twice: one person photographs the ticket, another drops the paper copy in the tray, and both get keyed.

One weekly sweep. Friday, someone empties every truck door pocket and captures what escaped. Ten minutes. The sweep is the net under the system, not the system. If it produces more than a handful of tickets, capture at the counter is not happening.

Capture the packing slip too, on deliveries. It is the only document that proves what actually showed up, and quantity billed against quantity delivered catches money the same week.

How long to keep the paper after you capture it

Two different clocks run here and they get confused.

The tax and audit clock is long, it is set by the IRS and your state, and it applies to the record, not to the specific piece of paper. A legible digital copy is generally an acceptable record. Ask your CPA where your state lands before you shred anything.

The dispute clock is short, and it is the one that costs you money. Credits get honored on fresh invoices and argued about on old ones. Published AP practice runs on similar timelines in the other direction: one institutional credit memo policy escalates at 45 days to vendor contact and 90 days to collections (Miami University). Returns have their own clock and conditions, since published distributor return policies commonly require prior approval, original packaging and a 25% minimum restocking fee, and exclude wire cuts entirely. More in supply house restocking fees.

A workable rule: keep the physical ticket in a dated box for 90 days, then keep only the capture. Ninety days covers essentially every credit request you will actually win, and it means a credit request email can attach the original if a rep asks. Do not shred anything you have not confirmed is legible in the capture.

The reason any of this matters: price history

A ticket you never capture is not just a missing document. It is a unit price you can never compare against anything. You cannot tell whether the coupling you bought in June cost more than the one in March, because March does not exist in any form you can query. That gap is exactly the condition price creep needs. Nobody has to hide anything. The comparison simply cannot be made.

Counter runs are the worst offenders: the least documented purchases you make and the most frequent. They are also where you are most likely to be paying stock pricing instead of your negotiated number, which is the mechanism behind contract price vs counter price. Miss the tickets and you never find out which column you are actually billed from, covered in column pricing at distributors.

The market context sharpens it. Copper and brass mill shapes ran 645.990 in January 2025 and 803.275 in June 2026 in the BLS producer price index (series WPU102502, recent months preliminary), so some of what you see is genuinely the market. But AGC reported that in May 2026 input prices for new nonresidential construction rose 8.4% year over year while contractors’ bid prices rose only 3.5% (AGC, June 2026). You absorb the difference. The only way to tell which part of your increase is the market and which part is drift on your account is your own numbers, per SKU, over time. That gets built out of tickets, and it is what makes a quote vs invoice mismatch provable rather than a feeling.

The short version

• Call your distributor: emailed invoices and portal access, at the account level. Ask whether counter and will-call tickets are included.
• Photograph whatever that does not cover, at the counter, before the ticket goes in the truck.
• Flat, straight on, even light, whole ticket, one ticket per image, glance before you pocket the phone.
• One destination for every capture. No side channels.
• Friday door-pocket sweep as a net, not as the plan.
• Keep paper 90 days, then keep the capture, after your CPA confirms your state’s retention rule.
• Capture packing slips on deliveries, not just invoices.

Sources

• Epicor electrical distribution software, on Counter Sales as a separate module in distributor ERP.
• Trimble product pricing information (Trade Service), on manufacturer catalog numbers and UPCs as distributed product data.
• Stampli, three-way invoice matching, on matching invoice, PO and delivery receipt.
• Tipalti, three-way match, on tolerance practice and named failure modes.
• Miami University credit memo policy, a published AP credit escalation timeline.
• Colonial Electric Supply return policy, 25% minimum restocking fee, no returns on wire cuts.
• BLS producer price index series WPU102502, copper and brass mill shapes. Recent months preliminary.
• AGC of America, June 11 2026 release, input prices +8.4% year over year against bid prices +3.5%.