Introducing the PredictHQ Features API: Prebuilt intelligence and features to extract value from event data faster
The last 14 months have been incredibly volatile and virtually impossible to predict. Given this chaos, you need reliable external data sources to better understand demand. But you don't want to collect data just to collect data. Data does not always equal intelligence. What you need is smart data, to predict demand. Targeting smart data instead of big data makes the value of a data set compound rapidly.
And that couldn’t be more true for data that will be integrated into demand forecasting models. Over the years — and this year in particular — we’ve come to discover that incorporating smart data, like intelligent event data, into forecasting models can take time as teams learn about the unique data set and sift through what matters to their business. PredictHQ has always focused on building a smart data set and infrastructure, following standards like QSPD, to ensure that you have access to smart data rather than a dumb data feed, but we’ve taken it to a new level.
Today we’re announcing the launch of our newest API called Features API that is pre-built intelligence and features that companies will be able to use directly into forecasting models. Why does this matter? Features API will reduce the time it takes to use intelligent event data in your demand forecasting models from months to days.
Turn months of work into days with the Features API
Let's start by putting things into perspective. Before Features API, customers had to take the following steps before they could integrate our intelligent event data into their models:
Write code to download large volumes of events into a data lake - could take several hours, if not days.
Your code needs to be built to keep the copy of events in your data lake updated.
You would then need to figure out 'features' to create on top of the individual events. This requires a lot of research and development to determine what patterns and aggregations impact your business. This could take weeks, depending on how familiar you are with event data.
Then you need to test these hypotheses that you’ve determined in the research and development step. This testing process is what takes the most time, with many data science teams spending weeks, if not months on this.
Once testing is complete and the features are finalized, data engineers need to write code for the features to then update your demand forecasting models with the updated code.
Final step is to test the code and then deploy into production.
This process can take months and often does take months. This process isn’t unique to event data. It's required for any external data set that you want to pull into your models. We’ve decided as a business it's important for us to solve this time suck, so that you can focus on improving your models to make business decisions faster. With the Features API simply call the API and use the data returned from the API in your models.
Access library of features to use directly in models for faster results
So what does this look like in practice? You can think of Features API as a library of features for use in your models. Maybe you want to know the total amount of sports events happening in a radius around your location – we’ve got you covered. Or maybe you want to know the average event size happening around your store – we have that. The Features API is incredibly fast, allowing you to aggregate large amounts of data in seconds.
For example aggregating attendance across different categories for all events within 5 miles of a location. Lets walk through some real scenarios to paint a better picture:
Transportation: Perhaps you’ve determined that your ride sharing business is heavily impacted by conferences. You can use a pre-built feature to find the top 25 cities in the US with the most conferences and focus your efforts on those locations.
Retail: By understanding the ebbs and flows of non-attended, scheduled events such as religious holidays and school holidays by each location, you’re able to integrate our features to ensure sales and marketing are aligned for the potential footfall for each of your geographic store locations.
Accommodation: Knowing that almost all attended events have an impact on the demand for rooms in your central city location, you’re able to find the average attendance rate over the next 90 days easily, enabling you to integrate the Features API attendance into your revenue management system.
These are just a couple of potential use cases. At the end of the day, no matter your business, the Features API will enable you and your teams to extract value from intelligent event data rapidly. We would encourage you to reach out to test it out today. Want a head start? Check out this video where I walk through the new API.
