The $40 decision, a hundred thousand times a year

Freight brokerage runs on thin net revenue margins. The public comps sit in the mid-teens across enormous volume. Detention typically bills at $50 to $100 an hour after two hours of free time, so a single event lands somewhere between $50 and $500. Neither number is interesting alone.

Take a brokerage moving 100,000 loads a year, and assume just $30 a load of accessorial charges that nobody had time to check. That is $3 million against net income where a few hundred basis points is the whole year. Unvalidated accessorials stop being an ops nuisance and become one of the largest controllable line items in the business.

The structural trap is that every individual decision to skip validation is correct. The dispute costs more than the charge. Only in aggregate does the pattern become obviously wrong, and nothing in the workflow surfaces the aggregate until the annual review, by which point the money is gone and the timestamps have aged out.

An agent team that makes validation free, so the economics that justified skipping it stop applying.

• Accessorial validation tests each claim against the actual carrier agreement, not industry defaults, reconciling scheduled appointment times against arrival and departure timestamps from every available source, and showing the detention computation rather than asserting a conclusion.
• Exception prediction watches telematics and lane history to flag loads that will miss before they miss, which turns a 2am service failure into a 9pm phone call that prevents one.
• Invoice audit reconciles every carrier invoice against the tender, the contract, and the accessorials exhaustively rather than by sampling, which is what lets systematic overbilling surface.
• Resolution drafting writes the carrier and customer communication, with the evidence attached.
• Orchestration ranks the queue by margin at risk and customer SLA impact.

This is the part that makes the difference between a validation tool and a dispute generator.

A geofence crossing is strong evidence. A gate ticket is strong evidence. A driver-reported arrival time relayed by phone is weak evidence, and a dispute built on contradicting one is a dispute you lose while damaging a carrier relationship. So every timestamp the system relies on carries its source and a reliability assessment, and that weakness propagates into the recommendation. A charge the system thinks is invalid on weak evidence comes back as “approve: evidence insufficient to dispute,” not as a dispute.

The ELD mandate is quietly what makes any of this possible. Arrival and departure times became machine-readable, which turned accessorial validation from a phone-based argument into a data reconciliation problem.

The obvious way to build this is to dispute everything the evidence supports. That is also how you lose capacity in a tight market, and capacity is the brokerage's actual asset.

So carrier relationship value is an explicit, inverted dimension in the scoring model, and dispute recommendations carry a value floor. Some charges the system knows are wrong come back marked approve, because being right about $60 is not worth being the broker that fights about $60. That constraint came from the client, and it was correct.

The agents never refuse a payment. They validate, evidence, and draft; settlement staff decide. And carrier selection never optimizes around safety ratings; CSA scores inform selection in one direction only, because negligent-selection exposure is not a variable to tune.

Validation now runs on every accessorial rather than on the ones someone had time for. In the first four weeks the system reviewed 2,400 charges, flagged 23% as contractually unsupported, and recovered $92,000, with dispute win rate at 79% and no measurable change in carrier acceptance.

That last clause is the one the client cared about most.