Geospatial data visualisation using Kepler.gl
Kepler is an open-source data visualisation framework from Uber’s Data Viz team, and is an easy way to visualise large sets of geospatial data. Built with deck.gl, Kepler helps you visualise data as beautiful, interactive maps.
Kepler analyses your data, identifying datatypes and suggesting visualisations. You can add multiple layers of data, and filter them. You can also make the map interactive with tooltips, selecting the information you want displayed.
All this is done entirely on your machine, within your web browser, using WebGL to render large sets of data efficiently.
Kepler is extremely easy to work with, and you can get started right away. Check out a few examples to understand how data is processed by Kepler.
Alternatively, you can run Kepler on your local machine. Detailed instructions are available here.
If you’re not a developer, you’ll first need to set up a coding environment on your machine. Make sure your node version is above v6.0
You’ll also need a Mapbox token to display your map. Sign up for a free account to get one, which gives you up to 50000 map views a month, plenty and more for your experiments.
Clone the Kepler repository to your local machine, and you’re all set.
One of the first things I did was to see where customers were opening the app, and what they were doing after. Seeing this on a map made it much simpler to understand than a set of numbers in a table, or even a funnel visualisation.
Shown below is one day for DriveU in Bangalore, and it was surprisingly easy to do this using Kepler!
Visualising data makes it easier to understand patterns, and this is especially the case with geospatial data. For example, looking at the map above:
- There are hot zones that appear during the morning, evening and night times. This helps our team move drivers to these zones at the time they start to peak so that our customers get matched with drivers when they need one.
- A lot of customers open their DriveU app after 10PM. These are probably users who don’t want to drive back after a long day, possibly even after a night out depending on which area they’re booking from.
- Most people book for the same day or the next, and very few book for the day(s) afterwards. This indicates a shift in customer thinking from planning in advance towards a more on-demand approach.
- The lion’s share of users do not make a booking after opening the app — this is a funnel optimisation problem. What are these users doing? Why are they not creating a booking? It’s something that I ponder every day. :)
Data is at the heart of everything we do at DriveU. We’ve taken up the challenge of organising the driver aggregation industry and are building a platform that makes car ownership a pleasure for our customers and empowers our driver partners.
We are constantly looking for motivated, talented people to join us in our mission. If you’re up for the challenge, get in touch. Lets have a chat! :)
