One of our top priorities is making sure you (our customers) can easily use our products and that we help you when we’ve fallen short on that goal. Please reach out to us or navigate to any of the following areas if you have a question, feedback or are having difficulty understanding or using our products – we would love to hear from you!
API Documentation
Our API documentation is available here. If you have further questions, please reach out to us.
Frequently Asked Questions
Commonly asked questions are outlined below. We’d be happy to chat about anything else!
General Support
If you still can’t find what you’re looking for, please click on the Intercom icon at the bottom right of your screen.
For more information, see status.thinklumo.com
The inputs to Lumo’s prediction models vary by how far from scheduled departure time the prediction is being made.
Further out, the main factors are historical performance for a given airline, seasonality, day of week, and time of day. As it gets closer to flight time, the models begin to consider things like flight paths and air traffic, congestion at various airports and how it impacts other airports (think “butterfly effect”), and weather forecasts. The model simulates air traffic throughout each day, based on all of the inputs, to return the predicted delay probabilities. Read more.
We are 93.5% accurate! But what does that mean? Not so much actually.
Lengthy flight delays aren’t that common; only about 8% of flights worldwide have a delay greater than 60 minutes. So if we predicted that every flight will be on time or delayed less than 60 minutes, we would be correct 92% of the time! Sounds great, but not a very meaningful statistic.
Due to the dynamic nature of flight delays, our predictions for a given flight are intended to indicate a level of risk in the form “a delay of M minutes is p% likely,” as opposed to a binary prediction of “delayed” or “not delayed.”
Therefore we think about accuracy as the answer to the question, “Out of all the times we say a delay of M minutes will occur p% of the time, what percentage of the time does it actually occur?” If the answer is close to p then we consider ourselves to be accurate. Read more.
The Lumo delay indexes are a score from 1 to 10 given to each flight indicating how “risky” a flight is with respect to being delayed. The score is intended to capture both delay frequency (probability of a delay occurring) and delay severity (how long will the delay be if it does happen). Read more.
