Turning Roadside Counts Into Usable Reports
In the early 1970s, traffic agencies could collect vehicle counts with pneumatic tubes stretched across a road. Each passing vehicle sent a pressure pulse through the tube to a counter, which recorded traffic over time. But collecting the counts was only the first step: turning the recorded data into tables and reports could mean tedious manual work.
Traf-O-Data aimed to automate that step. The venture, founded by Bill Gates and Paul Allen with Paul Gilbert, built a system to read the counters’ recorded data and process it into reports. Its central challenge was practical as much as computational: the machine had to reliably accept the data in the format the counters produced, then turn it into results traffic officials could use.
A Demonstration Is Part of the Product
That made reliability crucial. A system that worked in theory—or could process a prepared example—was not enough if it faltered when potential customers watched. Traf-O-Data’s machine barely worked during demonstrations, undermining confidence in the product and helping sink the business.
The failure illustrates a basic rule of applied computing: the useful invention is not just the program or the processor, but the whole chain from input to dependable output. If a machine cannot consistently read its data, its calculations do not matter to the person waiting for a report.
A Small Venture, A Lasting Lesson
Traf-O-Data did not become a major business, but it gave the young founders a concrete problem to solve: connect a real-world measurement system to software that could save people work. Building for that kind of task forced them to confront hardware, data formats, and customers’ expectations—not merely write code.
The telling detail is that the product’s weakness appeared in public, at the moment it needed to earn trust. In computing, a demonstration can reveal what a neat design on paper cannot: whether the entire system works when someone else is counting on it.
