Supercharged battery swaps: Get the most out of your supply network in a high-demand season

Felix Jonathan Jakobsen
May 25 · 5 min read

Swappable batteries, long a best practice in the moped-sharing industry, are growing increasingly common in scooter and e-bike fleets.

As my colleagues Jay, Nick, and Dan have previously written, Zoba is the leader in micromobility decision automation. Today, we turn to the problem of which batteries to swap first in order to optimize for meeting demand and maximizing fleet utilization. Initially, there is an obvious and straightforward swapping strategy: swap all batteries below a certain ‘unrentable’ threshold, for example: 25%.

In a world with no constraints and steady demand, this is a fine approach and works out relatively well. However, as soon as mobility operators run into times of increased demand, they start facing multiple constraints e.g. staff capacity, charging capacity, the number of batteries available, and the number of operations vehicles.

What does this mean in practice?

Imagine an operator has a fleet of 500 vehicles and usually swaps around 10% of its fleet’s batteries on any given day. Below is a histogram of the battery levels of this simulated fleet. In times of normal demand, the average battery level is 62% and only 35 vehicles have a charge below 25%. With some maintenance vehicles, the team should be able to easily visit all the unrentable vehicles within their typical 50 vehicles.

Typical battery charge distribution in a shared vehicle fleet in times of low demand

Now, imagine demand increases sharply — as we see in most European cities on the first nice Saturday of Spring. They suddenly find themselves in a situation where more than 20% of their fleet is below the unrentable threshold and will require a battery swap. To revisit our example above, in a fleet of 500 vehicles, the mean battery level is down to 46%, and 107 vehicles are under 25%, far more than before!

Typical battery charge distribution in a shared vehicle fleet in times of high demand

At first glance, increased demand looks like a good problem to have. Increased rides should mean increased revenue. However, charged vehicles are necessary to meet this demand and the fleet is losing charge at a faster rate than before. Also, operators need to ensure that the most productive vehicles remain online. Thus, the team now faces hard choices. How should they allocate their scarce resources? Which 50 vehicles are the most important to visit?

The typical response is to serve the lowest battery vehicles first. If the operator swaps the bottom 50 vehicles, they will reach all batteries with a charge under 12%.

However, this strategy does not guarantee that these swaps actually meet the team’s objective of putting the most productive vehicles back online to maximize ridership.

For example, a vehicle that still has 40% charge might be in a position where it could receive significantly more rides within the next 24 hours than a vehicle with a lower charge. If its battery is not swapped, it will run out of battery quickly and miss a majority of those rides. This becomes particularly tricky in moped sharing schemes, where the cost of rebalancing is much higher than in scooter sharing; it is often too cost ineffective to move the vehicle to another location. Further, moped sharing operators often see between 5 and 7 rides per vehicle per day, meaning the impact of the downstream rides is significantly higher if a vehicle runs out of battery early in the day.

In times of increased demand, the logical response is for teams to go for apparently high demand areas first. The problem is that it’s almost impossible to identify high-demand areas using intuition alone. As my colleague Evan previously pointed out in a blog post, demand differs from utilization. And in day-to-day operations, teams tend to mix up demand and utilization. Intuitive choices therefore often lead to operators oversupplying certain hot spots, resulting in lower fleet performance overall.

Locations like train stations often look like areas of high demand, but in fact they’re areas of high utilization. Swapping batteries of vehicles already available in these locations may increase the supply, but it is difficult to know immediately whether the additional battery-swapped vehicles will be ridden. If they are not, the location was oversupplied.

What these examples show is that accurately predicting demand is really hard because there are many variables playing into the equation.

What does Zoba do differently?

At the heart of Zoba Move lies our proprietary demand prediction model. This demand model optimizes swapping operations (as well as deployment or rebalancing, if needed) in order to increase the entire fleet’s utilization — at the network level instead of the vehicle level. This means that Zoba recommends swapping batteries on the vehicles that will receive the greatest number of rides after a swap.

As Zoba optimizes at the network level, we predict not only where their vehicles will get their first ride, but also what happens in terms of downstream rides i.e. which vehicle is driven to a location where it will get another ride, and another, and another.

During times of high demand, operations teams are often forced into an approach of “keeping their heads above the water” aiming to keep as many vehicles up as possible. Coming from a different angle, we aim to equip these teams with the right moves to maximize utilization using their limited resources.

Changing their approach to battery swapping can change the game: from just keeping their head above the water to predictively allocating and executing tasks in order to increase their fleet utilization.

This makes a team’s swapping strategy more proactive and economically efficient as it prioritizes the best vehicles ahead of the less time critical ones.

While to some operators this seems overly complicated as they focus on the basics of reactively keeping as many vehicles as possible afloat, we propose that this is in fact part of getting the basics right. Working with Zoba’s demand-based battery swaps offers tangible benefits:

  1. Increase efficiency, especially when you are strapped for resources and make sure you do the right tasks first
  2. Take work off of your local operations manager’s desk by automating decisions and routes
  3. Streamline your operations staff’s tasks and task distribution
  4. Manage task adherence, especially when working with 3PLs

If you’re interested in improving your fleet or operational efficiency we’d love to hear from you! You can learn more about our work and reach out at our website www.zoba.com.

Zoba Blog

Zoba increases the profitability of mobility operators through decision automation.

Zoba Blog

Zoba uses demand forecasting and optimization to improve the performance of shared mobility services. On this blog, Zoba operations leaders, data scientists, and engineers write about the problems we solve for shared mobility operators and tools we use to solve those problems.

Felix Jonathan Jakobsen

Written by

Business Development Director @ Zoba

Zoba Blog

Zoba uses demand forecasting and optimization to improve the performance of shared mobility services. On this blog, Zoba operations leaders, data scientists, and engineers write about the problems we solve for shared mobility operators and tools we use to solve those problems.