Heavy Civil Contractors Spot-Check 5% of Their Haul Tickets. Reading All 7,400 in a Season Now Costs About $30.

Take a mid-size grading and paving outfit — call it $28 million a year, three crews, a plant hauling agreement with the nearest drum plant and a pit twenty-two miles the other way. Across one season that company will take in somewhere near 7,400 delivery tickets: scale tickets for hot mix, aggregate tickets for base and rip-rap, ready-mix delivery tickets for headwalls, box culverts and sidewalk. Ask the controller how many of those get matched line by line against the purchase order and the supplier's invoice, and the honest answer is a spot check. Five percent. Ten in a rain week when a project engineer is stuck in the trailer anyway.

Nobody in that office is lazy. Matching one ticket properly takes a person roughly ninety seconds: find the ticket, read the net weight off the gross and tare, confirm the mix design number is the one on the purchase order, confirm the time stamp lines up with a shift the crew actually worked, confirm the truck number belongs to a hauler you agreed to pay, then post it to the right cost code in HeavyJob or B2W Track and check the unit price against the quote. Ninety seconds times 7,400 is 185 hours. That is a full-time person for a month and a half, doing nothing but paper, in the middle of paving season. So the spot check is what the math allowed.

Ticket reconciliation, said plainly

Here is the definition worth quoting: ticket reconciliation is proving, for every single load that appears on a supplier's invoice, that the load was ordered, delivered to a job you had open, placed against a pay item you can actually bill the owner for, and priced at the number in the purchase order. Four checks. Any one of them failing is money.

What the 5% spot check misses is not dramatic fraud. It is drift. A plant runs a price escalation off the liquid asphalt index and applies the March number to April tickets. A hauler bills standby on a day the trucks were released at 10 a.m. A ready-mix supplier charges a short-load fee on nine-yard loads that were never short. Someone at the pit keys mix 610 instead of 57 stone and the ticket comes across at a different unit price. A subcontractor's dump truck runs two loads to a job that closed out in the previous pay period, and the tons land on a cost code that has no remaining pay quantity, which means you bought material you cannot bill.

None of that shows up in a spot check because a spot check finds the loud errors. The quiet ones are one-percent errors on a $7.2 million material spend, and they compound across a season while everybody is busy trying to get the mat down before the temperature drops.

What changed between the 2025 price and the 2026 price

Two years ago the idea of reading every ticket automatically was not new — it was just uneconomic. Running a capable model over a photographed scale ticket, in enough detail to read a smudged mix design number and a handwritten note in the margin, cost real money per document. At 7,400 documents plus the invoice lines to match them against, an owner did the math, saw four figures a season for something that might still need a human to double-check, and passed.

Through 2026 that price collapsed. Frontier models are running roughly ten times cheaper than they were in 2025, capable ones sit around two dollars for a million words of work, and for high-volume repetitive reading, running the model on a machine in your own office is roughly ninety percent cheaper again than sending it to the cloud. The practical effect for a contractor: reading a ticket, matching it, and flagging what does not line up now costs a fraction of a cent. The whole season's worth of tickets costs less than one tank of diesel in a haul truck.

That is the entire development. Not a smarter model — a model cheap enough to point at every transaction instead of a sample.

flowchart TD
  A["Plant scale ticket photographed by the foreman"] --> D["Agent reads ticket: net tons, mix, truck, time"]
  B["Pit aggregate ticket from the loader operator"] --> D
  C["Ready-mix delivery ticket signed on the deck"] --> D
  D --> E{"Matches the PO price and an open cost code?"}
  E -->|Yes| F["Posts to HeavyJob cost code, nobody touches it"]
  E -->|No| G["Exception list for the project engineer"]
  G --> H["Credit memo requested from the supplier"]

A Tuesday in June, ticket by ticket

The paving foreman on the Route 9 mill-and-fill does what he already does: the tickets ride back to the trailer in the truck, or he photographs them on the tailgate at the end of the shift. Nothing about the field changes. That matters, because any system that asks a grade foreman to key data into a phone at 6:40 a.m. will be dead by Thursday.

Overnight the agent reads all 214 of that day's tickets. It pulls ticket number, plant, mix design, gross, tare, net, truck number and time. It compares net tons against the purchase order price, against the quote the estimator loaded into HeavyBid, and against the quantity the crew reported placed on the daily report. It posts the clean ones — 197 of them — straight to cost codes, and it leaves seventeen on an exception list.

By 7:15 the project engineer opens the exception list. Nine are nothing: driver wrote the wrong job number, easy fix. Four are late-night loads billed at a night differential the purchase order never included. Three are tons delivered after the crew shut down for a thunderstorm, which means the mix went somewhere and the question is where. One is a mix design number that does not match what the spec calls for on that layer, which is now a quality question and a phone call to the plant before another day of it goes down. That last one is worth the whole exercise.

The arithmetic, with the assumptions stated

These are illustrative numbers for a company that size — plug in your own and the shape holds.

The thirty dollars is not a typo, and it is the only part of this that is genuinely new. In 2025 the same job priced out near four thousand dollars in machine time, which is exactly why it never got approved.

What it will not do, and who still walks the scale

It does not settle a dispute. When a plant insists the load left at 2:14 p.m. and your foreman says the truck never showed, a machine reading two documents cannot tell you which is true. It can only show you the pair that disagree, faster and on all of them instead of one in twenty. Somebody still has to make the phone call, and that somebody should be a person the plant manager already knows.

It does not know your handshake deals. If you let the pit slide on a fifty-cent difference last October because they held a truck for you during the storm cleanup, that is not written anywhere and the exception list will flag it every week until someone records it as agreed.

It does not replace the field measurement. Tickets tell you what was bought; they do not tell you what stayed on the road. Yield checks — tons per station, thickness cores, the superintendent's own count — are still the thing that catches a hauler dumping short or a mat running thick. Use the ticket check to protect the payables side, and keep your yield discipline exactly where it is.

And on prevailing-wage work, do not let the same system quietly re-code costs between pay items to make a job look better. That is a straight line to a problem with the owner's auditor. Exceptions get resolved by a human with a reason written next to them.

Frequently asked questions

We already get electronic tickets from Command Alkon. Isn't this solved?

Partly, and only for the suppliers who send them. Most contractors run a mix: electronic tickets from the ready-mix producer, paper from the small pit down the road, and a PDF batch emailed monthly by the plant. The value here is that one process handles all three, including the crumpled one out of the truck's door pocket, and reconciles them against the same purchase orders.

How long before it pays for itself?

If it catches one duplicate month of asphalt billing, the first week. More realistically: run it in parallel with your spot check for one month, keep score of what each found, and decide from your own numbers rather than anyone's brochure.

Will my project engineers fight it?

The ones who like ticket entry will — there aren't many. The real risk is the opposite: the exception list gets ignored because nobody owns it. Put one name on it, give them fifteen minutes at 7 a.m., and make the count of open exceptions a line on the Monday meeting agenda.

What about the tickets from our own trucks and our own pit?

Those matter more, not less. Internal tickets drive equipment cost allocation and the rates you charge jobs. Checking them all is how you find out that the 30-ton haul truck has been billing hours against a job that finished in May.

A note on the calls this generates

Every exception becomes a phone call — to the plant, to the pit dispatcher, to a hauler, and often from them to you while your PM is standing on a shoulder with traffic going by. CallSphere builds the AI voice and chat agents that answer those lines when nobody in the trailer can, take down the ticket number and the job, and get the callback on the right person's calendar instead of leaving a message nobody hears until Thursday. It does not read your scale tickets. It makes sure the call about one does not go to voicemail during paving season.