Here’s a good question to write on a Post-it Note and put on your desk: “What assets do we have right now that are not making money for us?” — Bill Walsh

For a micro-mobility company (aka “Uber for X”), matching demand and supply is at the core of the business. For every unit of demand which is not met, you have lost an order. For every unit of supply which is idle, you are losing money on your vehicles.

Normally, every micro-mobility company has a matching algorithm. But, how does that algorithm perform at different times of the day, days of the week, locations? Do they have a pattern? How do you analyze that?

In this piece, we would delve deeper into supply-demand gaps for micro-mobility companies that have vehicles on the ground. Let’s explore both these gaps in a lot of detail and understand how it relates to both monitoring and historical analysis.

Monitoring helps to know what’s happening on the ground right now and be proactive about it. The real world is very fickle and chaotic and your model always cannot accommodate these sudden changes.

For example, when demand peaks or troughs beyond the average value (it could be rain, traffic, protests, local events etc.) monitoring helps to understand where is the demand abnormally high and what can you do about it.

While monitoring is good for these immediate, abrupt events, analyzing the patterns historically helps you understand these minute nuances of how areas behave (or rather, users or riders behave in certain areas).

For example, you learn from your data that when the trip of the user ends in an area that is on the outskirts in the evening, it is likely not going to be picked up by another user — which means the ops team would have to pick it up!

Locale Console

I. Demand >> Supply (Dropoffs)

Let’s take a scenario: At 8 pm on a Saturday, your users are dropping off at one of the localities with a bunch of pubs as there is a music concert.

For the scope of this article, we can assume a dropped off user to be someone who didn’t book a ride. A canceled user would mean someone who didn’t opt for a ride after booking for the ride.

Note: Dropoffs per unique user and per unique session are both important to look at. So, even if the user searches five times and drop-offs, he would be counted as only one.

Case 1: Monitoring

Some important metrics that you would want to monitor in real-time to be able to match your demand and supply together and for other tactical decisions:

  • Where: Which locations are these users dropping off?
  • Rate: What’s the drop-off rate per minute like? Is it increasing, decreasing or constant?
  • Distance to Nearest Supply: Where is the nearest available supply and how far is it?

Case 2: Analyzing Historically

The aim of analyzing historically would be to find patterns in the mismatch and improve your model over time and take more strategic decisions. For example,

  • Frequency: Do the drop-offs in those locations have a pattern and do they happen at a certain frequency? Example, weekday vs weekend, morning peak vs evening peak
  • Duration: For how long do the users drop off in those locations on an average? 5 mins, 30 mins or an hour?
  • Properties of the Area: Is there something inherent about the area or supply? For example, in some cases, the riders cancel the trips deliberately because the areas don’t seem to suit their preferences (high traffic, outskirts).
  • Next Demand Time: It’s not just about what’s happening at that instance. But, what is the flow like. Where are you most likely to get demand next?
Understanding flow for micro-mobility companies is super important. Where area is the most likely to get demand next? How far is that from where the drivers are currently?
  • Profitability: Would the ride have been profitable? Maybe it was not a bad idea for the user to drop off the ride was not profitable.

II. Supply >> Demand (Idle Riders)

Your riders are in a  residential area where they receive plenty of ride requests usually. However, you are not getting requests and hence, riders are idle whereas users might be dropping off in a commercial area.

Note: Idle Riders and Rider Sessions are both important parameters to look at. This will help us understand whether the problem is with a particular set of riders (high number of riders sessions) or other parameters such as location or time (high number of idle riders)

Case 1: Monitoring

As discussed above, monitoring works well to know what’s going on right now, especially in cases of anomalies.

  • Where: Which areas are riders (or vehicles) idle right now?
  • Rate: What’s the rate per minute like? Is it increasing, decreasing or constant?
  • Distance to Nearest Demand: Where is the nearest available demand and how far is it?

Case 2: Analyzing Historically

Historical analysis is primarily to understand the patterns in the inefficiencies.

  • Frequency: Do the idle riders have a location pattern and do they become idle at a certain frequency?
  • Duration and Active Hours: For how long are the riders idle those locations on an average How does it correlate with the active hours of the riders?
  • Properties of the Supply or Area: Is there something inherent about the area or supply? For example, the drivers cancel when the mode of payment is not acceptable to them — they want some cash early in the day.
  • Earning and Incentives: It would be useful to evaluate the earnings and incentives of the rider in those cases. Can the incentives compensate for their preference to not accept the particular kind of rides?
  • Next Demand Time: As discussed, in the on-demand kind of companies, the match is not about what ‘s happening right now but also where is the flow. If your demand prediction model says that you would get some orders here in 30 mins, it doesn’t make sense to move the riders 5 kms away!

The Online Vs Offline Dichotomy

If you are a company that as a web app, you use web (or app) analytics products (like Google Analytics, Heap, Mixpanel, Amplitude, Clevertap) to get granular details of how users are performing. You create “user personas” based on their tastes, preferences, and actions to personalize your strategies, which increases your engagement and retention.

If you are a company with moving drivers or vehicles on the ground, getting visibility and analyzing how your different dimensions interact with each other on the ground becomes very crucial.

If you could similarly create location personas or micro-markets based on how different areas perform, you could personalize your strategies, and make these micro-markets profitable.

You get the full picture of what’s happenning on the ground once you understand how your supply and demand behave together in conjuction to how certain areas perform as well.

Bridging Supply-Demand Gaps

Once you know the characteristics of these gaps — when do they happen, where do they happen, how often do they happen and so on, here are all the things you can do to bridge them!


In areas of high supply (which is not very mobile), you can focus on offline promotions and awareness activities. For example, setting up a hoarding in a residential area.


You can create user personas depending on how they move and send contextual marketing promotions. For example, in office areas at 5 pm whenever it’s time for snacks!


If you discover the patterns in your demand and flow for different days, time and area, you can make your supply available and increase its utilization. For example, the operations team would redistribute vehicles at different times on different days.

Unit Economics

From the revenue side, you can make your pricing more location-based. From the cost side, you can identify the bottlenecks in the delivery journey and learn which steps in the lifecycle you spend the time and costs on.

Damage Control

You can stay on your toes in case of an anomaly, for example, a sudden dip in revenues or orders placed. For instance, sometimes a local event happening somewhere can drastically affect your KPIs and you might need to react on that!

At Locale, our geospatial analytics product is built for the product and operations team in on-demand companies. It helps you understand which locations you can double down on to improve your most important KPIs such as unit economics, asset utilization, user acquisition & retention.

In our previous piece, we talked about some of the strategies that companies like Uber, Grab, Airbnb use to match their demand and supply. You can check that out here:

Similar Reads:

Making Location-Based Experimentation a part of our DNA
What can we learn about experiments on the ground from web-based experimentation?
How analyzing supply-demand gaps can optimize your unit economics!
After all, we all need our own Google Maps.