How predictive software is keeping Rouen’s bike share in balance

Featured Content Partner

Share this article

Running a bike-share network is often a race to prevent stations from becoming completely empty or entirely full.

In Rouen, a city in northwestern France, the operator has to manage a more precise set of constraints, with specific stations required to maintain minimum bike availability during defined periods of the day.

French software company Qucit has been working with Inurba Mobility on the Rouen system since January 2026, using predictive tools to translate the operator’s complex service requirements directly into priorities for field teams. The system adapts its recommendations to the specific service level agreements (SLAs) of the network and has increasingly become part of the operational workflow, replacing much of the manual monitoring and judgement that previously guided daily decisions.

Rouen’s bike-sharing challenge requires 25 designated stations to maintain at least 50% of their allocated capacity during at least one of three daily time windows. Because requirements roll over if missed, stations face a continuous, shifting operational target throughout the day.

“What makes Rouen demanding is not the size of the network but the shape of the requirement,” Marie Quinquis, Head of Customer Success at Qucit, tells Zag Daily. “Most bike share contracts set a fleet wide availability target. Rouen sets it station by station and window by window.”

Qucit’s system recalculates the status of those 25 stations three times a day, once for each service window. It identifies which requirements have been met, which remain open and which stations are at risk, then incorporates that information into the recommendations given to field teams.

That means a station that still has an opportunity to meet its target can rise up the work queue, while one that has already secured its requirement for the day can fall down it. It remains part of the network, while aiming to avoid consuming  field resources unnecessarily.

“The status is not simply displayed,” Quinquis says. “It weights the rebalancing recommendations sent to the field teams and it is refined hour by hour as the day goes on.”

The Rouen SLA is unusually specific, but Qucit aims to translate different contractual requirements into operational rules and automatically incorporate them into the work assigned to field teams. In Rouen, this is implemented through 41 automated rules developed with the operator, divided into seven families and adapted to various factors like what side of the river the station falls on and how busy it  is.

Rebalancing a city divided by the Seine

Rouen presents a particular geographical challenge. Bike flows are heavily directional between the Left Bank and the city centre, meaning the network does not naturally return to a balanced distribution over the course of the day.

“In most cities, flows partly offset each other over the course of the day,” Quinquis says. “In Rouen there is no equilibrium to return to. The flow runs in the same direction every single day, so the distribution has to be rebuilt rather than corrected.”

The Seine adds another constraint. A field worker moving between stations on the same bank can often complete the next task within a few dozen metres. Crossing the river can require a journey of close to two kilometres before the next intervention begins.

Today, about one in seven movements crosses the Seine, according to Qucit. The software also changes when operators intervene. Instead of waiting for a station to approach a shortage, recommendations can begin with stations that are filling up, allowing teams to move bikes before peak demand arrives.

“On a one way flow, acting on the source two hours before the peak avoids the corrective trip altogether,” Quinquis says.

For Inurba, the operator of Rouen’s LOVÉLO self-service bike-share network, this predictive approach has changed how teams allocate their time between monitoring the network and rebalancing stations.

“Before using Qucit, most operational decisions relied on the teams’ experience and on real-time observations made in the field,” Frank Porta, Operations Services Director at Inurba Mobility, tells Zag Daily. 

“With Qucit, we now have predictive recommendations that allow us to anticipate network imbalances before they happen, while taking into account service priorities and contractual obligations.”

Porta says the operator now moves fewer bikes while visiting more stations for a nearly equivalent level of service. The main change has been the allocation of staff time. Instead of continuously monitoring the network and manually deciding where teams should go next, managers can rely on Qucit to automate much of that analysis. Teams spend less time working out what needs attention and more time carrying out the interventions themselves.

Automating the decisions behind the daily workflow

Six of Qucit’s seven rule families in Rouen deal with the wider state and location of bikes, including preventive and corrective maintenance, bikes left outside stations and journeys that have not been properly closed. Maintenance technicians work through the same application and close their damage reports there.

“Previously, a significant share of our time went into the surveillance of the network,” Porta says. “Today, Qucit automates a large part of that analysis.”

The measurable result cited by Qucit is an average compliance rate of more than 95% against Rouen’s station-by-station, window-by-window service level agreement between February and April.

The software is also being used to fold tasks that previously sat outside conventional rebalancing into the same workflow. Inurba used to have one staff member dedicated to travelling around the network to collect unavailable bikes. Most of those jobs, including collecting bikes left on trips, missing bikes and bikes outside stations, are now incorporated into standard rebalancing tasks.

That allows the operator to use the same field activity for several operational needs. Qucit says 51% of the tasks generated by its automated rules in Rouen relate to maintenance, while all rebalancing tasks are generated by the algorithm.

The broader issue is how operators allocate finite resources when contracts place particularly high demands on parts of a network. Quinquis argues that a service guarantee inevitably concentrates resources somewhere, potentially leaving other stations with less attention.

“Every guarantee concentrates resources and those resources are paid for by the part of the network nobody named,” she says. “A bike in one place is a bike that is not somewhere else.”

The purpose of the software, she says, is to manage that trade-off across the whole network. Once a priority station has met its daily requirement, resources can be redirected to other stations where availability is at risk.

For Inurba, that has meant automating more of the analysis behind day-to-day operations. The field teams still make the physical interventions, but more of the decision-making that determines where they go next is now automated.

As bike-share contracts become more detailed and operators are asked to maintain availability across increasingly complex networks, that division of labour is becoming an important part of managing fleets without simply adding more vehicles or field staff.

Share this article

Photography by