Why Harsh Braking is Also a Routing Problem and Not Just a Driver Problem

The coaching plateau

Most last-mile operations run a version of the same safety programme. Telematics captures harsh braking, harsh acceleration, speeding, and cornering. Drivers are scored. The worst-scoring cohort receives coaching. Scores improve.

Then they stop improving. The programme keeps running, the coaching keeps happening, and the number settles at a level nobody is happy with. The usual reading is that the remaining offenders are harder to change, so the response is more coaching, tighter thresholds, or consequences attached to scores.

There is a better reading of the plateau. Coaching works on incidents caused by habit and disposition, and those are real. It does not work on incidents caused by a plan the driver cannot complete safely, because in those cases the behaviour is not a defect. It is a rational response to conflicting instructions: serve these stops, in these windows, in this sequence, and do it safely, when the four cannot all be true.

That distinction is testable with data most operations already hold, and it determines whether the next intervention should target drivers or routes.

The mechanism is time debt

Every route carries an implicit time budget. When the plan issues more work than the budget supports, the difference has to go somewhere. Call it time debt.

A driver can absorb time debt in exactly three ways.

Drive harder. Higher speeds between stops, later braking, more aggressive gap acceptance. This is the option telematics measures and safety programmes target.

Compress the stop. Park closer than is legal or safe, skip the walk-around, photograph the parcel carelessly or not at all, hand off without the verification step. This option is largely invisible to telematics and shows up later as disputed deliveries, damage claims, and thin evidence.

Miss the window. Accept the service failure and preserve the driving and the process. This is visible immediately, and it is the option most operations implicitly punish hardest.

Every one of those is a reasonable response by a person trying to satisfy incompatible instructions. Which one a given driver chooses depends on what they believe will be counted, and the plan set up the choice.

This is why coaching alone plateaus. Coaching removes the first option. It does not remove the debt. The debt migrates to the second and third options, which is how a safety programme can succeed on its own metric while proof-of-delivery quality quietly degrades and on-time performance stalls.

What the government data shows

The link between operational scheduling and crash risk is not a hypothesis. It has been quantified by the regulator’s own oversight body.

The DOT Office of Inspector General examined commercial driver detention and found that a 15-minute increase in average dwell time, meaning the total time a truck spends at a facility, increases the average expected crash rate by 6.2%. The same work estimated detention was associated with reductions in annual earnings of $1.1 billion to $1.3 billion for for-hire commercial drivers in the truckload sector, and reduced motor carrier net income in that sector by $250.6 million to $302.9 million annually.

Read those two findings together, because they describe one mechanism. Dwell is a scheduling and facility variable. It measurably raises crash risk and measurably lowers driver earnings. The driver experiences both consequences of a variable they do not control.

Dwell is also not rare. ATRI found drivers were detained at 39.3% of all stops in 2023, losing between 117 and 209 hours per year depending on sector. A planning model that treats dwell as a fixed allowance is mispricing something that happens on two stops in five.

The regulator now treats schedule as a crash variable in its own right. FMCSA has moved forward with a study titled Crash Risks by Commercial Motor Vehicle Driver Schedules, examining relative crash and violation risk against factors in the driver’s work schedule. When the safety regulator is studying schedules, treating safety as purely a behavioural matter is a choice rather than a default.

Also Read: Why New Drivers Quit Before They Learn the Route: Earnings Variance Is a Dispatch Decision

Where time debt comes from

Seven planning decisions generate most of it. None is a safety decision, and none is usually reviewed by the people running the safety programme.

The last row is worth separating. Where remaining hours are validated after a plan is built rather than modelled inside it, the constraint bites in execution instead of in planning. A driver approaching an hours limit with stops remaining is in the sharpest version of the time-debt choice, and the plan created that position.

Travel time deserves a note too. INRIX’s 2025 Global Traffic Scorecard recorded 62% of urban areas worldwide seeing increased congestion, with only 26% improving. A model calibrated on last year’s averages is calibrated on a distribution that has moved, and the gap between planned and actual travel time is the rawest input to time debt.

The diagnostic: four cuts

This is the part worth doing before the next coaching cycle. Take incidents per thousand stops and cut them four ways. The pattern tells you where the cause sits.

The third cut does the most work. Holding the driver variable constant is what turns an argument about attitude into a question about route design, and it is usually answerable from data already in the telematics and dispatch records.

The interpretation is straightforward. Clustering on cuts one to three with little clustering on cut four means the operation has a planning problem wearing a safety label. Clustering on cut four means the safety programme is aimed correctly. Most operations have some of both, and the value of the exercise is learning the proportion rather than assuming it.

Also Read: From Static Route Plans to Continuous Re-Optimisation: A European Last-Mile Efficiency Benchmark

The incentive problem nobody designed

Where safety scores carry consequences, and where unsafe behaviour is the only way to complete the assigned plan, the operation has built something it would not have chosen deliberately.

The sequence runs like this. A driver receives routes with structural time debt. They absorb it by driving harder, because that is the option that preserves service. Their telematics score falls. A poor score restricts access to preferred routes, affects standing, or in some operations affects pay. The driver is now penalised for a planning defect, and the routes they can access get worse, which increases the debt.

Learn how a driver management software can enhance rider management, click here.

That is the same compounding pattern that shows up in earnings variance, and it has the same resolution: the metric is fine, the attribution is wrong. A score that measures driver behaviour without controlling for plan feasibility is measuring two things and reporting one.

The National Academies work on long-distance trucking is relevant here as well, noting irregular schedules and extended hours among the primary causes of driver departure, and finding that effective frontline dispatchers correlate with lower turnover. Dispatch quality already shows up in retention data. It shows up in safety data too, if you cut it properly.

What to change in planning

Five changes, in order of how much they reduce time debt relative to effort.

1. Model dwell per site from that site’s own history, not from a network average. The sites with the worst detention are known and stable, and pricing them correctly is the single largest correction available.
2. Allocate buffer where variance is, rather than evenly or not at all. Slack spread thinly protects nothing; slack placed ahead of high-variance stops protects the back half of the route.
3. Put remaining driver hours inside the solver as a hard constraint rather than validating them afterwards. A plan that cannot legally be completed should be infeasible rather than merely flagged.
4. Model access complexity as a stop attribute. Parking difficulty, stair access, gated sites and permit zones change the real duration of a stop, and treating a stop as a stop guarantees the estimate is wrong in exactly the places it matters.
5. Re-decide when the day changes. Time debt accumulates during execution, so a plan that cannot be revised transfers all of it to the driver. Gartner’s finding that 95% of supply chains must react quickly to change while only 7% can execute decisions in real time describes how common that transfer is.

Also Read: How AI Dispatch and Allocation Works for Freight Carriers: A Practical Guide (2026)

How Locus makes feasibility a constraint rather than a hope

Locus, the world’s first Decision-Intelligent, Agentic TMS, addresses time debt where it is created rather than where it surfaces.

Within DiSCO, the Dispatch agent plans and re-sequences against more than 250 real-world constraints per computation, including time windows, vehicle attributes and driver hours, so a plan that cannot be completed inside legal hours is infeasible rather than flagged for someone to notice. The Capacity agent holds remaining hours as live state across the roster, which is what allows the constraint to bind during the day rather than only at planning time. The Hub agent models facility readiness and dwell, which is the mechanism for pricing detention per site instead of per network. And because the cycle is Sense, Decide, Execute, Learn, a route that begins slipping is re-decided rather than left to the driver to absorb.

A clarification worth making plainly: none of this is a safety product, and Locus does not claim a crash-reduction outcome. What it changes is plan feasibility and the ability to re-decide, which the government data indicates are upstream of crash risk. That is the honest version of the claim.

Locus has been recognized by Gartner for seven consecutive years, featured in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies, named a Leader in TMS by QKS Group (SPARK Matrix), and ranked #1 in Route Planning on G2’s 2026 Best Software Awards. In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Locus continues to operate independently.

Two US deployments show the mechanism, with outcomes measured in exposure and execution rather than in incidents.

A global lottery operator running field service across more than 25 states described the infeasibility problem in its own terms: even a well-built plan went stale within the hour as urgency, traffic and weather shifted, while contracts, labour laws and SLA windows differed by state and some carried one-hour response commitments with liquidated damages above $100 per hour. That combination is a time-debt factory. Locus models each state’s contracts, labour laws, SLA windows, zones and skills as live constraints rather than post-plan checks, and re-optimises against live traffic, weather and urgency. The operator reported drive distance and time down 15% and fuel spend down 18%. Less time on the road is less exposure, and lower fuel consumption on the same work is consistent with less aggressive driving.

A Fortune 50 parcel and logistics leader shows the execution side. Its 4,500-strong driver pool spanned captive fleets on zone-based routing and third-party carriers on tendering and on-demand assignment, with no single tool unifying them, so plans could not be re-decided across the whole network. After Locus took over pickup, transit and delivery decisioning against 250-plus operational constraints, weekly execution across 51 service-centre locations moved from 75% to 92%. An execution rate that low means most weeks were absorbing failure somewhere, and time debt is one of the places it lands.

Also Read: The Real Cost of Manual Dispatch in North American 3PLs

Run the four cuts before the next coaching cycle

The next time a safety review lands on the agenda, bring one additional table: incidents per thousand stops, cut by stop position, hour of shift, route, and driver.

If the incidents cluster on routes and on the back half of the day rather than on individuals, the coaching programme is aimed at the wrong layer. It will keep producing modest improvements and keep plateauing, because it is asking drivers to stop responding to a problem nobody has fixed.

That is a more useful conversation than the one about driver attitude, and it has an owner. Safety belongs to operations in the same way retention does, because both are partly produced by the plan. The dispatch layer is where the fix lives.

Book a Locus demo to review how dwell, access complexity and remaining driver hours are currently modelled in your plans, and where time debt is being created before the day starts.