Beyond On-Time Percentage: What a Real Carrier Scorecard Measures

Last updated: July 2, 2026.

Ask a room full of brokers what makes a good carrier scorecard for freight brokers and most will point to one number: on-time percentage. It is the number that fits on a slide, the number the customer asks about, the number the TMS reports without any extra work. It is also the number that hides the most. On-time percentage tells you what happened after a load was already delivered. It says nothing about whether the carrier will accept the next tender, how it behaves at your worst-dwelling facility, how often it files a claim, or whether the dependable partner you scored last spring still runs the same trucks and drivers today. A real scorecard measures the signals that predict reliability before the load moves, not the one lagging signal that confirms it afterward.

On-time percentage is a lagging signal, not a reliability signal#

Start with what on-time percentage can and cannot see. It is measured after the wheels stop, from delivery timestamps the carrier already earned. That makes it a report card on loads that are done, not a forecast of loads you still have to cover. The single most useful thing to know about a carrier on a Tuesday morning is whether it will accept the load you are about to tender. On-time percentage is silent on exactly that question, because it only counts the loads the carrier accepted and delivered. Every tender the carrier quietly rejected is invisible to it.

That is how a carrier posts a beautiful number while being unreliable in practice. A carrier that accepts two of every ten tenders you offer, and delivers those two on time, shows up at nearly 100 percent on-time. The eight loads it walked away from, the ones you scrambled to re-cover at spot, never touch the score. Price does not rescue you here either. FreightWaves has documented that even sizable truckload rate increases did little to improve carrier reliability: when contract rates climbed sharply, tender rejections stayed stubbornly elevated, because a carrier rejects a load for capacity reasons as often as for money. Paying up is not the same as buying dependability, and a scorecard that leans on price and past on-time percentage will keep recommending carriers that fail you at the moment capacity is tight.

On-time percentage has a second flaw: it is an average, and averages lie by blending. A carrier can run at 98 percent on your easy dedicated van lane and 71 percent on your hard reefer lane into the Southeast, and the headline number splits the difference into something that looks fine and describes neither lane. The load you actually care about is the hard one, and the average is designed to hide it.

The signals a real scorecard actually weighs#

A real scorecard is not one number. It is a small set of signals, each of which predicts a different kind of failure, read together. Here is what belongs on it, what each signal predicts, and the uncomfortable part, where the data for it lives today.

Read that right column again. No two of those signals live in the same place. That is the whole problem, and we come back to it below. First, walk the signals themselves.

Tender acceptance rate is the earliest and most honest predictor of reliability, because the load you could not cover on time was usually a load that was never accepted. A carrier's acceptance rate on your specific lanes tells you far more about next week than its on-time percentage tells you about last week.

Communication responsiveness is the leading edge of an exception. The carrier that stops answering check calls at 2 p.m. is the carrier whose driver is already sitting at a closed dock, and the silence shows up hours before the late-delivery flag does. Responsiveness is a reliability signal precisely because it is early, and an exception caught early is an exception that does not cascade into the next three loads on the same truck.

Dwell and detention behavior is where the scorecard gets specific and expensive. How a carrier behaves at a facility is not one number, it is a facility-by-facility pattern, and the pattern is shockingly common. ATRI found that drivers reported being detained at 39.3 percent of all stops in 2023, a share that climbed to 56.2 percent on refrigerated loads and 42.5 percent among fleets that run the spot market. Detention is not just a carrier problem. It is a facility problem that a good scorecard attributes to the right shipper, so you know which lane looks cheap on the rate sheet and quietly burns hours at the dock. The detention and demurrage you can actually recover starts with measuring it at the facility, not the carrier.

Claims ratio measures freight integrity: how often a carrier damages, shorts, or loses freight per hundred loads, and what the rework costs. A carrier can be perfectly on time and still generate a claim on one load in forty, and that claim is real margin gone plus a customer relationship strained.

POD timeliness looks like paperwork and behaves like cash. A carrier that delivers on time but returns the proof of delivery four days late has stalled your billing, extended your cash-to-cash cycle, and created a collections problem that never shows up in on-time percentage.

Safety and authority standing is the vetting signal, and it is public. FMCSA's SAFER system and safety measurement data expose a carrier's operating authority status, insurance, out-of-service rates, and crash history. A carrier that was fine to book last quarter can have its authority revoked or its insurance lapse this quarter, and a scorecard that never re-checks is scoring a company that may no longer be legal to haul your freight.

Performance drift is the signal that ties the rest together, and it gets its own section, because it is the one almost every scorecard ignores.

Why one headline number lies, and a per-lane, per-facility picture does not#

The instinct to compress a carrier into a single grade is understandable and wrong. Reliability is not a property of a carrier. It is a property of a carrier on a lane, into a facility, on a given equipment type, right now. The same carrier can be your most dependable van partner in the Midwest and a coin flip on flatbed into a congested port. A headline score averages those two truths into a third number that is false about both.

This is where the per-lane view earns its place. The question worth answering is not "is this carrier good," it is "how does this carrier compare, on this lane, into this facility, against every other carrier I run there." Answering it requires reading your own records rather than a vendor's opinion of them. When you can rank every reefer carrier you use into the Southeast by acceptance rate, dwell, and claims on that specific corridor, the booking decision stops being a gut call and becomes a comparison you can defend. For an asset carrier scoring its own board, the same per-lane view is what separates a lane that pays from one that just runs up deadhead miles between loads.

Carriers drift, so a scorecard is a moving picture#

Here is the assumption buried in most scorecards: that a carrier is a stable thing. It is not. The carrier you booked last year may not be the carrier you booked last week, because the people and equipment behind the name turn over constantly. The ATA has reported that driver turnover at large truckload fleets has run as high as 91 percent, and it is careful to note that this figure measures churn between carriers, not people leaving the industry. Drivers move from fleet to fleet chasing pay and home time. The consequence for your scorecard is direct: the roster that earned last year's on-time number has largely been replaced by the roster hauling your freight today.

The carrier pool itself churns just as hard. FreightWaves counted a net loss of more than 4,500 carriers across a single recent stretch as capacity exited the market, part of a broader truckload capacity exodus that rising rejection rates revealed even while demand was soft. Small carriers lose authority. Owner-operators park equipment. The mix of who is available to haul your lane is different this month than last, so a scorecard frozen at a point in time is describing a market that has already moved.

This is why drift belongs on the scorecard as a signal in its own right, and why WATCH matters more than a snapshot. What you want to catch is the carrier whose acceptance rate on your core lane has slipped four weeks running, or whose dwell at your top facility has crept up since the new dispatcher took over, while the headline on-time number still looks fine because the trend has not finished playing out. Reliability is a trajectory, not a photograph, and the failures worth preventing announce themselves in the slope long before they show up in the average.

Why nobody assembles the real scorecard#

If these signals are so predictive, why does almost every scorecard still come down to on-time percentage? Because on-time percentage is the only signal that already lives, whole, in one system. Every other signal is scattered. Tender acceptance sits in the TMS. Dwell lives in tracking timestamps. Detention lives in accounting and in invoices the carrier may never get paid on. Claims live in a claims system. PODs live in document management. Responsiveness lives in email threads and dispatcher memory. Safety and authority live at FMCSA. No single screen holds more than a slice, so assembling the real picture means a human stitching six systems together by hand for every carrier, which nobody has time to do, so nobody does it.

The detention numbers make the orphaned-data problem concrete. ATRI found that 94.5 percent of fleets bill for detention, yet get paid on fewer than half of those invoices. The cost of a carrier that dwells is real, it is measured, and it is sitting in an accounting system that never talks to the on-time number in the TMS. The signal exists. It is just stranded one system away from the decision it should inform.

This is exactly the gap a connected dataset closes, and it is the whole thesis of the Mithrilis platform: intelligence from connected data, not automation of a single workflow. When your TMS, tracking, accounting, claims, and document systems are unified against one shipment and one carrier record, each signal on the scorecard becomes visible, and just as importantly, sourced. We do not hand you a black-box grade you cannot argue with, and we do not surface a single opaque score in place of the evidence. Every signal traces back to the row it came from, so before you book, you can see the acceptance rate, the dwell trend, and the claim that drove the number, and verify it. That principle, that you should be able to check every result, is written into our manifesto. With the data connected, Atlas answers the plain-English question directly: how did this carrier actually perform on this lane into this facility over the last quarter, and is it getting better or worse. The scorecard stops being a spreadsheet someone updates once a quarter and becomes a live, sourced view the person booking can trust.

How a 3PL scores carriers across the whole book#

The scorecard matters to both sides of the market, and it means something a little different for each.

For a freight broker, the scorecard is cover-risk management. You are grading partner carriers on whether they will accept and perform on your customers' lanes, and the edge is knowing, before you tender, which carrier is trending up on this exact lane and which one has quietly slipped. A per-lane scorecard built from your own records turns carrier selection from a relationship habit into a defensible decision, and it protects the margin that a bad cover or a surprise claim would erase.

For a 3PL, the scorecard points two directions. You score the carriers you use across every customer account, on the same acceptance, dwell, and claims signals a broker would, but rolled up across your whole book so one underperforming carrier shows up everywhere it runs, not one account at a time. You also live on the other end of your clients' scorecards, because the dwell and POD numbers you post per account are the ones deciding whether you keep the business. Seeing carrier performance across the book, and your own service the way each customer sees it, per facility and over time, is how you route on your real leverage and defend the accounts worth keeping.

See your real carrier scorecard, sourced and per lane#

On-time percentage is not wrong. It is just late, and alone. The carriers that fail you at the worst moment tend to look fine on the one number most scorecards are built from, because the signals that would have warned you, a slipping acceptance rate, a creeping dwell, a lapsing authority, a roster that turned over, were sitting in systems the scorecard never read.

The fastest way to see the difference is to look at your own carriers through connected data. Mithrilis unifies the systems these signals hide in, keeps every number traceable to its source, and updates the picture continuously instead of once a year, so drift shows up as a trend and not a surprise. You still decide who to book. The scorecard just finally shows you the whole carrier instead of the last thing it delivered. Request a demo and we will build the real scorecard on your own lanes and facilities.

Related Mithrilis capabilities

• The Mithrilis platform

  How connected data turns scattered carrier signals into one sourced, verifiable view.

• For freight brokers

  Score partner carriers on acceptance, dwell, and claims per lane, not just on-time.

• For 3PLs

  Roll up carrier performance across every customer account, not one at a time.

• Stop exceptions before they cascade

  Why communication and dwell signals catch failures early, in WATCH mode.

Frequently asked questions

What makes a good carrier scorecard for freight brokers?

A good scorecard weighs the signals that predict reliability before the load moves, not just the on-time percentage that confirms it afterward. That means tender acceptance rate, communication responsiveness, dwell and detention behavior at specific facilities, claims ratio, POD timeliness, and safety and authority standing, read together and compared per lane and per facility rather than compressed into a single grade. Most importantly it tracks drift over time, because a carrier that scored well last year may not be the same operation this quarter.

Why isn't on-time percentage enough to measure a carrier?

On-time percentage is a lagging average. It is measured after delivery, so it only counts loads the carrier accepted and delivered, and it is blind to every tender the carrier rejected. A carrier that accepts two of ten tenders and delivers those two on time can post nearly 100 percent while leaving you to re-cover the other eight. It also blends easy and hard lanes into one number that describes neither, which is why a real scorecard reads it alongside acceptance, dwell, and claims, per lane.

What is tender acceptance rate and why does it matter more than price?

Tender acceptance rate is the share of loads a carrier actually accepts out of those you tender at the agreed rate. It matters because the load you cannot cover on time is usually one that was never accepted, so acceptance predicts next week better than last quarter's on-time number does. And price does not fix it: FreightWaves has documented that even large truckload rate increases did little to improve carrier reliability, because carriers reject loads for capacity reasons as much as for money.

How does carrier performance drift over time?

Carriers are not stable. The ATA reports large-fleet driver turnover has run as high as 91 percent, churn between carriers rather than people leaving the industry, so the roster that earned last year's score is largely replaced by the one hauling your freight now. The carrier pool churns too: FreightWaves counted a net loss of more than 4,500 carriers across a single recent stretch. A scorecard frozen at a point in time describes a market that has already moved, which is why drift belongs on it as its own signal.

Where does carrier scorecard data actually live?

In pieces, across systems that rarely talk to each other. Tender acceptance sits in the TMS, dwell in tracking timestamps, detention in accounting and invoices, claims in a claims system, PODs in document management, responsiveness in email, and safety and authority at FMCSA. That fragmentation is why most scorecards default to on-time percentage: it is the only signal already assembled in one place. ATRI's finding that 94.5 percent of fleets bill detention but get paid on fewer than half of those invoices shows how a real, measured cost can sit stranded one system away from the decision it should inform.

Do 3PLs need carrier scorecards too?

Yes, in two directions. A 3PL scores the carriers it uses across every customer account, on the same acceptance, dwell, and claims signals a broker would use, but rolled up across the whole book so an underperforming carrier is visible everywhere it runs rather than one account at a time. A 3PL also lives on the receiving end of its clients' scorecards, so the dwell and POD numbers it posts per account decide whether it keeps the business. Seeing carrier performance book-wide, and its own service the way each customer sees it, is how a 3PL routes on real leverage and defends the accounts worth keeping.