Note: Before reading this piece, I highly recommend you reading the first part of this series which covers the basics of supply-demand analysis here:
Ordering food, groceries, cigarettes or even booze at the tap of our phones has become a norm these days. And if you are like me, you probably order everything at the oddest hours of the day!
For an on-demand company (aka “Uber for X”), that is primarily a three-sided marketplace, matching demand and supply is at the core of the business and also extremely challenging. Moreover, the real-world introduces far more variation that leads to starker results than in the model.
When supply-demand gaps exists, we are either losing orders or our riders are idle — both of which are contributing to losing money. But, what's causing these gaps to exist? This could happen as a function of demand, supply or both.
Case 1: Demand
The number of orders that you are losing with time because the riders are not available or they are present in the wrong location.
When you are losing orders, how does the order velocity look like? Which means are you getting more orders or lesser orders than usual?
Case 2: Supply
The number of riders that are idle and the amount of time they are spending without being utilized because of the wrong location or lesser demand.
Note: Being idle is a point in time property. Hence, while analyzing historically, we need to analyze the metric “idle session time”.
Total Supply Available
When riders are idle, how does the availability of the total riders look like? Have the number of available riders increased or decreased? While analyzing this historically, we need to analyze the metric “total session time”.
Decoding the reasons for the gaps
Before even trying to bridge the gaps, we need to understand “why” they are happening in the first place. If the orders are getting lost it might be due to the following reasons:
Riders are not available
- Wrong Location: If we are losing orders, it might mean that the riders are not available which means they are either all busy or are idle in the wrong location.
- Excess Supply: There can also be a case where we are catering to all the demand but have excess idle riders. It might be worth analyzing the total available supply then.
The order frequency is falling
If the frequency at which you receive orders on that day and that time is lesser than the usual, that means you need to dive into why that’s happening.
- Type: Is it all types of orders or only a particular kind of order? For example, is it only category A which is falling which contributes to the majority of the order volume.
- Lifecycle: Is there a particular step in the order journey which is seeing a huge number of drop-offs? For example, are there some challenges with the checkout system or the payment gateway due to which users are not able to complete their orders?
Orders are not being allocated
There might be a possibility that orders are coming in but they are not getting allocated to the riders. If this is happening, there might be some problem with the allocation (or matching) algorithm that you are using. In these cases, it is best to raise this concern with the relevant people inside the organization.
In a three-sided marketplace, sometimes there might be an allocation problem with the store or the restaurant. I myself have experienced quite a number of cases when I couldn’t receive the order because the issue was at the restaurant.
To delve deeper on how to carry out analyses on static location such as stores or restaurants, check this out:
Riders are not accepting the orders
Let’s say that the orders are being allocated to the riders but they are not accepting the orders. This might happen because of a number of reasons.
- Distance: Are they not accepting the orders because the distance to pick up is very high?
- Area: Are they not accepting the orders because they don’t wish to travel to the location of the pickup?
- Payment: Are they canceling because the payment mode of the user is not preferable?
Riders are canceling the orders
Often, all of us have experienced riders canceling the orders after they have accepted it — especially once they get to know about the drop location.
- Distance: Are they canceling the orders because the distance to dropoff is very high?
- Area: Are they canceling the orders because they don’t wish to travel to the drop location?
- Origin-Destination: Is it neither about the origin or the destination but about the route? In that case, large incentives also often don’t work because riders just don’t wish to cross those areas.
Cities usually have invisible walls that riders prefer not to cross. Where are these walls? What times of the day are they most prominent in?
As discussed before, supply-demand matching is not just about what’s happening right now. It’s about the flow. Where is demand going to be next and how does it move in different days and different times?
Understanding flow for last-mile delivery companies is super important. Which area is the most likely to get demand next? How far is that from where the drivers are currently?
For example, with one of the last mile delivery companies, studying the flow showed that the delivery partners started moving towards the center of the city and by evening, they started moving out, because most of them lived on the outskirts. But they needed to come inwards because that’s the region in which they receive the maximum amount of orders.
- Points of Interest: Another very interesting analysis that can be done is to analyze how the supply-demand gets influenced by different points of interest such as schools, colleges, malls or tourist destinations.
- External Events: It would also be insightful to analyze how different external events such as rain, a large concert, a sports match, protests affect the demand and the supply for your company.
- Cohorts: A cohort is a group of users, riders or bikes that perform similarly. What kind of users and order do I lose at different times of the day and days of the week? What kind of riders are idle most of the time?
Operational Analytics at Locale
At Locale, we are building an “operational” analytics platform that uses geospatial data of users, vehicles, or riders. Supply and operations teams in last-mile delivery companies can monitor what’s going on on the ground and take more tactical decisions. Product and analytics teams can use it to do historical analysis and take more strategic decisions.
Our mission is to bring the same level of granular analytics as Uber and Grab to every company collecting location data. We want to enable business teams in on-demand companies to get operational insights within a minute, not hours or weeks later!