While we can't give away all the details of our secret sauce, here's an outline of the basic...
If you're here, you know that we built our reputation on using machine learning to predict flight delays, but we've also long recognized that flight delays are only one component of the delays in a traveler's journey. With Lumo's new TSA and CBP wait time feed, we're adding one more component to the delays a traveler might face, helping passengers, travel agents, and admin assistants better plan for and manage delays.
The source of the raw data
One challenge to estimating wait times has been getting granular data to help train forecasting models. We use data published by the DHS from the TSA and CBP, but the data usually contains only historical averages.
We don't provide real-time data!
The wait time data we provide isn't intended to answer the question "what is the wait time right now?". Instead, it's meant as a planning tool to answer the question "What is my wait time likely to be tomorrow/next week at 2pm? How bad is it going to be?" Given this data, travelers can plan on when they arrive at an airport, or account for wait times on connections that involve immigration.
We don't estimate times for every security checkpoint
Unfortunately, the raw TSA wait time is crowdsourced at the airport level, so our estimate just provides one estimate across the entire airport; this is intended to be a worst case wait time at the entire airport, and is a reasonable estimate of the worst case across all the checkpoints. We do break down the wait times for standard vs precheck travelers.
The CBP data is more granular in that it provides the estimates by terminal (when an airport might have multiple international arrival terminals).
So what's the point? What is Lumo's value add?
Rather than simply provide users with historical averages, Lumo applies filtering algorithms to clean up the raw data and then runs the raw data every day through a forecasting algorithm that estimates future wait times by hour. The estimates account for the number of departures/arrivals, time of day, day of week, etc. that captures recent trends. This provides more robust estimates than just using historical averages, and calculates confidence intervals on the estimate.
We then provide two estimates of wait times by hour – the average or expected wait time, and the 90th percentile wait time – for different types of travelers. The traveler types we currently support are standard vs precheck for TSA data, and US citizens/residents vs visitors for the CBP data.
A single API call for everything you want to know about your flight
The wait time data is bundled in with our flights API; if you are already a Lumo customer, the new data is already being returned with the API calls you currently make. If you're a new customer, a single call to our API returns flight delay information, predictions, CO2 emissions, travel advisories, wait times, and more. If you'd like to learn more, get in touch us to see how the Lumo API can supercharge your travel application.
Or, if you're a developer, you can always test drive the API by signing up for a free trial API key.
