As an innovative logistics startup, Shadowfax works in multiple delivery domains — Food, Pharmacy, Grocery, and E-commerce.
We work with large grocery clients for online grocery deliveries from their dark stores to end consumers. Dark stores become a critical point in the delivery process for grocery clients because the shelf life of grocery items such as vegetables, meat, and seafood is rather short. Add to this, the importance of timeliness because grocery deliveries are a part of the hyperlocal slot based delivery system. Customers who order groceries online, choose a specific slot based on their convenience and availability. If this convenience is disrupted due to frequent cancellations or delays it can create a negative customer experience, impacting the business of all the stakeholders involved in the entire delivery chain. This where data visualisation and interpretation can help. Data visualisation over time can not only help you identify the causes for various setbacks in your business, but also help you make informed decisions sooner than the competition.
So, why is data visualisation important?
Now, even before we delve into leveraging data visualization as a tool, for fraud detection for hyperlocal delivery, let us look at the delivery process in brief.
The process flow is roughly* like this:
Shadowfax Receives an order from the client → Our allocation engine assigns it to the most relevant partner → Partner completes the delivery to the customer.
But the last part of it is slightly more complicated. There are multiple endings:
- Order delivered to the customer. Phew!
- Order returned partially because of issues with some of the items.
- Order returned/cancelled because of the customer not being available or any other reason.
*This is only basic. There might be multiple steps in-between that we are not going into the detail of.
A case of suspicious user behavior
On a certain day towards the end of 2018, there was a sudden surge in order cancellations — we had to deep dive into this on high priority because the end customer experience was getting hampered with order re-allocations.
While we did identify the issue even before our client flagged it to us, we had to solve the issue identified as fast as possible because customer satisfaction is at the heart of our business.
How data came to our rescue
To get to the root cause identification, we quickly collated all the data that was relevant. This included geolocation of delivery partners at the time of cancellation, frequency of cancellation, feedback from end-customer about the cancellation and a lot of other pointers.
At that point, we knew that identifying trends would require spatial analytics and hence, we quickly started exploring data on @kepler.gl — a brilliant visualization tool built by Uber’s engineering team made open source!
The first visualization:
- Blue Dot — Location of the dark store
- Height of tile — number of shipments cancelled at that point
Clearly, a large percentage of shipments were getting cancelled at the store! This was completely counterintuitive because typically, these are confirmed orders and delivery partners are expected to go close by to the customer location before initiating an order cancellation.
To dive deeper into this, we added another layer of visualization on top of this!
Color of tile — the average distance of delivery partners from customers’ at the time of cancellation. Darker the tile, lesser the distance, i.e. delivery partner is near customer location.
Brighter the tile, more the distance, i.e. delivery partner is further away from customer location.
Comparing the trend between different cities reinforced the issue — cities with an increase in cancellation (Hyderabad, NCR in this case) also had a higher number of bright tiles in comparison to other cities. Here’s a zoom into some of the areas
The tower at the center shows a large number of orders being canceled at that point. The yellow color is an indicator that the average distance between customer and delivery partner at the time of cancellation was very high! It was now clear that there were miscreants in the system trying to play with the system to earn more through canceled orders.
So what did we do to?
Within days after this,
- Locations with high fraudulent cancellations by delivery partners were flagged
- Bad-apples were identified and were barred from the system
- We setup auto-triggers to detect such fraudulent cancellations on-the-go
…and the result of our smart move?
Our customer experience remained intact!
P.S.: This is almost a year old now. Since then, many more suspicious behaviors were flagged & rectified to improve customer experience and ensuring customer delight in 99% of cases.
So, if you’re struggling with your business, you can use data visualization to learn inefficiencies in your operations, build new processes to improve customer experience.
