Value of More Objective Data

Value of More Objective Data
April 8, 2019 Michael Wooley
RoadBotics data is objective. That’s because our pavement assessments are done by our deep learning model, not humans. Don’t get me wrong, I like humans. In fact, some of my best friends are humans. However, some of the things that make humans so great – our sociability, for example – also makes us less-than-ideal pavement assessors.
Road managers need to communicate why they’re doing what they’re doing. Here’s an exchange that might be familiar to them:
Constituent or Elected Official: (with a hint of contempt) Why are you repairing that road and not my road?
Road Manager: You see, we’ve put together a plan based on the condition of the roads that will keep them in the best condition while minimizing costs.
Constituent or Elected Official: (contempt rising) Yeah, but who determined the condition of the roads? You?
Even the most selfless civil servants cannot avoid suspicion from certain quarters that their road assessment was skewed in some politically-advantageous manner. The solution, of course, is to have a neutral third-party like RoadBotics do the assessment.



There are a lot of companies out there that will do a neutral assessment. With the exception of RoadBotics and companies that use LIDAR (which have their own cost and reliability drawbacks), all of these companies are going to be doing manual windshield assessments. A windshield assessment is an exercise in concentration and judgement. Who among us is as sharp at the end of a workday as at the beginning? Now imagine trying to maintain the same level of vigilance throughout an entire day of road rating. You’d have to be a machine.
Then there is the issue of what the assessor is seeing. They’d want to sit up front to get a good view. But are they looking down every few seconds to input data or keep their coffee from spilling? And what if they’re rolling by too fast to catch all of the distresses on a road? Do they see more distresses at stop lights because they have more time to take it all in? A video record of the assessment would certainly help with this.
“Ride smoothness” ratings are often an integral part of manual assessments. But then how do you compare an assessment that was done in a tank-like SUV (say) to the experience of a driver in an old micro-compact? And does the driver dodge or swerve towards potholes? RoadBotics has seriously considered these questions. Our conclusion was that the only way that anyone could provide objective data on this point would be to drive the same stretch of road tens or hundreds of times along slightly different paths. This wouldn’t be cost-effective for our customers. One of the greatest responsibilities of a data analyst is to be transparent about the questions that their analysis cannot answer. A RoadBotics assessment does not try to incorporate any ride smoothness rating because we cannot provide direct, objective data on that point.
RoadBotics road assessments are done by a model running on our tireless servers. Once the model has done its thing, we post the ratings to the RoadWay platform and never touch it again. In short, a RoadBotics assessment is not just neutral, it is objective.