With web analytics tools today, we are able to measure and optimize every user interaction based on their persona and their behavior on our web products and app. But analyzing how our people and assets move today on the ground is a lot harder because the real world is far more fickle and chaotic than the online world.
What we have learned while working with some of the on-demand companies is that today:
- Only 5% of the location data is getting used in models
- Delivery people are idle 30%-50% of the time
- ~30% unrealized revenue: Inefficient surge pricing models
At Locale.ai we are trying to solve some of these business problems spatially. The talk will cover the process and good practices of building spatial models such as what are the nuances of spatial data and how is it different from statistical data? How do we go about data preparation which involves cleaning it for GPS errors and outliers and enriching it to add more context?
We will talk about why it’s important to aggregate the data and why we prefer a system of hexagonal grids over geohashes. Finally, we will conclude the talk by talking about the importance of experimentation in creating models just like we do it with our UI and UX.
Key Takeaways for the Audience
- What are the caveats of location data?
- Why do we need spatial modeling?
- What are the biggest challenges to run spatial models at scale?
