When Sylvestre Duroudier was recruited, the Blablacar project has already begun: the data collection protocol had already been implemented. This one had been mainly written by Sophie Kuegler, then data scientist at the Grenoble Alpes Data Institute, and stored on the KIMA server created by Jean-Marc Francony.
The PACTE team of researchers (Estelle Ployon, Patrick Juen and Annette Casagrande) collected travel from a query of 120 French cities on 5 times. The dataset represents 8,5 millions of travel offers.

A huge dataset

Sylvestre Duroudier mostly worked on cleaning the huge dataset. There were indeed a lot of mistakes and lack of data due to the data registration itself. For example, addresses have to be written in the same way to be usefull for next analysis. So he built a script on R environment. In other words, he has set up a reproducible procedure for all collections to have the cleanest and the most exhaustive dataset.
 

Blablacar: a platform make accessible places that are not or poorly served by other means of transportation?

The second part of his work, when the database was stabilized, was to analyze it. He sought to see what happens in spatial patterns. Its guideline was as follows: does the Blablacar platform make accessible places that are not or poorly served by other means of transportation?

... a market offer strongly linked to supply and demand

His statistical and cartographic analysis carried out with R software tend to confirm that Blablacar is a market offer, and that, as such, it is strongly linked to supply and demand.
The offers are much more numerous in the large cities. However, the scheme "the larger the city, the more offers" does not always work. There are corridors linked to the road network: the frequented roads (A6, vallée du Rhône, etc.) are those on which there are the most offers.

... a good coverage of rural areas

The research results also show that Blablacar allows a good coverage of rural areas since there are relatively few places where there is no offer.
 

A dataset open to other analysis

Sylvestre Duroudier and Magali Talandier are writing an article about these first results. The dataset of Blablacar carpooling offers can be use for further research. It can be exploited for instance to study the distance in travel time or price instead of kilometers. It can also be interesting to compare their first results with other transportation networks (TER, Flixbus, Blablabus, etc.).