The participation rate matters. On its own, though, it does not yet explain what effect collaborative routing can have.
A small share of vehicles heading for the same bottleneck at the same time can matter for its load. The same share, spread across a large region and many times of day, describes a different situation. So what also decides is which trips are influenced, when they take place and where they meet.
Small changes can have large effects
The paper “Understanding Road Usage Patterns in Urban Areas” by Pu Wang and co-authors including Marta C. González (2012) shows this relationship particularly clearly.
In the model studied, a targeted reduction of one percent of trips cut the congestion-related additional travel time by around 14 percent in the Bay Area and 18 percent in the Boston area. A random reduction was considerably less effective.
What was studied there is the reduction of trips, not participation in NUNAV. The result therefore does not give a direct minimum share for NUNAV. But it shows: small, targeted changes can have a disproportionate effect. The mere number of influenced vehicles is not enough for an assessment.
Why other road users can benefit too
This network perspective is worth taking for the traffic experiment with the ADAC as well.
The direct comparison asks: how do the NUNAV vehicles fare compared with the Google vehicles? For the network as a whole, a second question comes in: how does the entire mixed group fare compared with a group whose trips are all optimized individually?
If NUNAV distributes vehicles and thereby relieves a shared bottleneck, vehicles without NUNAV can benefit as well. Put pointedly: NUNAV can optimize the Googles, too.
A comparison within a mixed group does not fully capture this possible shared benefit. Putting a number on it requires a suitable reference state under comparable conditions. Roughgarden’s work provides the theoretical background for the difference between individually optimized and coordinated route choice. It does not replace this practical proof.
How well do we know today’s traffic?
Before we judge how measurable a new system is, we should also ask how traffic has been recorded so far.
A traffic count shows how many vehicles pass a certain point at a certain time. On its own, it does not yet explain where these vehicles come from, where they want to go and why they are using this particular road.
Historical movement data broaden the picture. How much they tell us, however, depends on coverage, resolution and evaluation method. Origin–destination relationships partly have to be reconstructed. The actual purpose of a trip often remains unknown.
And yet traffic has long been steered: by signage, traffic radio, news, enforcement services and hearsay. Who receives this information, who follows it and what changes it causes is often hard to capture.
Connecting analysis, guidance and impact measurement
NUNAV brings these tasks together in one system. For participating trips, destination and planned route are known. Recommendations can be issued specifically, and the routes actually driven can be evaluated.
This creates a basis for putting traffic guidance into practice and examining its effect. It does not yet create a complete picture of all road users, and it does not replace a reference state. But it links the intended guidance more closely to observable driving behavior.
The limits remain concrete: Which alternative routes exist? What capacity is available? How many recommendations are followed? And which changes would have happened without the intervention as well?
As a citable starting point from practice, the ADAC report (in German), the video of the traffic experiment and the published driving data with analyses are available. When reusing graphics, the respective rights of use apply; Graphmasters’ own figures can be requested for approval.
No universally valid participation threshold can be derived from these sources. The more important question is: which concrete demand can we influence – and how do we measure the effect on everyone affected?
