Andy Garcia · freelance dev
FRENES
Back to the blog
Product & UX

1,099 Last-Minute Lane Changes in 21 Days at One Toll Plaza. The Pay-Online Sign Came After It.

October 19, 2026 7 min read

On September 16, 2026, the National Transportation Safety Board published its findings on a crash at the Swanton Township Toll Plaza, on the Ohio Turnpike (Interstate 80). On Thursday, August 15, 2024, at about 5:29 a.m., an SUV in the left high-speed lane slowed and steered right, toward the low-speed manual toll booths. It crossed in front of a truck-trailer doing 69.6 mph, with cruise control set at the 70 mph limit. The truck struck the SUV, hit the concrete barrier between the two sets of lanes and overturned. Two passengers in the SUV died. The driver, another passenger and the truck driver were seriously injured. Hazardous materials spilled from the trailer; no one was injured by exposure. Two more crashes followed in the queues that formed while the turnpike was closed, each killing one person; their causes were outside the scope of the NTSB investigation.

The press release led with the recommendation, close the slow lanes and go fully electronic, and with one number: 1,099 similar last-minute lane changes over three weeks. I read the full report, HIR-26-08. Two details in it matter well beyond highways, and one of them is a sign.

A word on scope first. The NTSB does not assign fault, and neither will I. Its probable cause is the SUV driver's deceleration and lane change across the truck's path, with reduced situation awareness, likely due to fatigue, contributing. The three rear passengers were not belted and were ejected; the board found a belt would likely have prevented one of the two passengers' deaths. What follows is about the design around that moment, not about the people in it.

Two speeds, one late fork

The Swanton plaza reopened as a hybrid on April 10, 2024, four months before the crash. On the left, two electronic lanes at 70 mph. On the right, low-speed lanes leading to the booths, with a 10 mph limit at the booth channels. Between them, a 950-foot painted gore area, then a concrete barrier.

The SUV's event data recorder tells the rest. It was doing 43.5 mph and slowing 4.75 seconds before impact, and 14.9 mph at impact. Against the truck's steady 69.6 mph, the speed difference grew from 26.1 to 54.7 mph. The NTSB estimates the truck driver could, at the earliest, perceive the closing rate about one second after the SUV entered his lane, when they were about 135 feet apart. He began braking after the impact had already happened. The board found he did not have enough time to avoid or mitigate the crash.

The sign was past the fork

The press release mentions "clearer information about toll payment options" in a single clause. The report is specific. Its probable cause says: "Contributing to the SUV driver's decision to change lanes was the lack of signage stating that missed tolls could be paid online."

The overhead signs above the fast lanes read "E-ZPass Only". The SUV had no transponder. But the fast lanes did not really require one: the 2024 upgrade put license plate readers in every lane, so a driver without a transponder could be billed. The only sign saying a missed toll could be paid online stood on the roadside 1.3 miles after the plaza.

The SUV had met both systems that night. Earlier, before the SUV driver took the wheel again, it went through four Illinois plazas where the low-speed lanes were closed and the only options were E-ZPass or online payment; each passage was logged as a "violation" because the SUV had no Illinois transponder. About 45 miles before Swanton, at Westgate, a hybrid plaza on the same turnpike, she paid an operator at a cash and credit booth. The NTSB concluded that a better presentation of payment options would have made the lane change less likely, "and therefore the crash could have been avoided."

Two more facts make this uncomfortable. The signage conformed to the 2009 edition of the federal Manual on Uniform Traffic Control Devices, whose toll plaza standards, the report says, have not been updated since. And no new equipment is needed to fix it: beyond signage, the report calls the remaining changes "administrative". Compliant, equipped, and the fallback was still announced after the decision.

The swerves showed what the crash count did not

The second detail is how the NTSB showed this was not one driver's problem. The turnpike commission handed over 21 days of footage from one camera, July 25 to August 14, 2024. Investigators counted only obvious, clear crossings of the gore area, left out commission and police vehicles, and skipped any crossing hidden by darkness or by larger vehicles. They still counted 1,099, an average of 52 a day, more than two an hour. The board calls the figure conservative. Crossing a gore area is illegal in Ohio.

Now look at the number a dashboard would have shown. The plaza's vicinity recorded 54 crashes in 2023 as a conventional plaza, and 51 between April 2024 and April 2025 as a hybrid. The NTSB itself calls that "about the same number." The outcome metric said nothing much had changed. The behavior metric said a dangerous maneuver happened twice an hour.

What the evidence does not show

The count comes from one camera, over three weeks, before the crash. There is no comparable count from other hybrid plazas, and none after the commission added two portable message signs on June 11, 2025, and updated its overhead signs. So the report cannot tell us whether better signs alone shrink the number. The video does not say why each driver crossed: no transponder, confusion, habit. The crash totals count all crashes near the plaza, without severity. And the NTSB writes that research is "not definitive" on whether hybrid plazas are safer than conventional ones; it is the fully electronic design that published studies clearly favor.

The commission's side, as 13abc reported it on September 16: sympathies to the victims and families, thanks for the investigation, "We will carefully review its recommendations," and a fatal crash rate of 0.4 per 100 million vehicle miles traveled in 2025, against 1.1 nationally. The NTSB asks it to close the low-speed lanes at four plazas (Westgate, Swanton, Newton Falls and Eastgate) and to say when. Illinois and Pennsylvania, whose toll roads connect to the same east-west corridor, have already gone fully electronic. Why keep the slow lanes at all? The report calls the decision "largely financial", and the toll operators' association, IBTTA, told investigators it is primarily a financial one.

What I take from it

A pricing page is not a highway. Nobody is thrown from a seat by a confusing checkout, and I do not want to put the two on the same scale. What transfers is narrower: a fork between two paths that move at very different speeds, decided late, with the fallback explained after the choice. Products have those forks. Here is what I would check.

Count the swerves, not only the crashes. The NTSB used a camera that was already installed and three weeks of footage the commission already had. Most products already log the equivalent: the user who reaches the last step, backs out and redoes it another way. In BidForge, the AI I built to read French public tender files for small businesses, that is the count I would pull before any incident review: who gets to the end of a flow and turns around. A support ticket is the crash. The turnaround is the gore excursion.

Put the fallback before the fork. If a path accepts people who lack the obvious prerequisite (no card, no account, no transponder), say so where they choose, not in the confirmation email or the FAQ. The Ohio sign said the right thing. It said it 1.3 miles late.

Do not read compliance as clarity. Swanton met the standard of its time. In the iOS apps I ship, passing App Review tells me nothing about whether a user understood which button keeps them on the free plan.

Write down why the slow path survives. When the capability for the simpler design is already installed, keeping the old path is a business decision. That can be legitimate. It should come with its own count of what it costs.

At Swanton, the fast lanes, the plate readers and the camera were already there. What was missing was a sign before the fork, and someone counting the swerves.

Sources

A project like this one?

I design and deploy products like this. Let's talk.

Let's talk