We know that you might not be able to check your Data Platform all the time, and now you don’t need to! Our latest tool lets you know when disparity is starting to get ugly, and you don’t need to do a thing.
How it works
Surge Alert sends you “real-time” alerts when your undercut rate looks funky. The way we calculate Surge Alerts is a result of many hours of data manipulation, discussion and trial and error modelling. Here's how it works:
- We calculate your undercut rate every hour, on the hour and keep track of it.
- We record and calculate an average undercut rate for the past 6 hours and for the past 7 days (or 168 hours, phew!), your long-term rate. We keep on doing this every hour!
- If, in the past 6 hours, the average is significantly different to the last 7 days, we send you an alert - this means you know when it needs to be checked out and can take action to beat the surge!
- If you want to know about your current surge multiplier, check out your Disparity Dungeon. We use your 6 hourly average undercut rate and compare it to your threshold to give you an idea about how urgent the surge you’re riding is.
Want more detail?
So how do we work out your threshold? (This is the tech-y bit!)
The threshold is worked out using the upper bound of a 99% confidence interval based on the long-term undercut rate. A confidence interval is used by statisticians to express the level of uncertainty associated with a sample statistic.
Our interval is at 99%. This means that if we picked a random undercut rate and figured out its interval, this would be the same as what your long-term rate looks like 99% of the time.
So, when we send you a surge alert, we're pretty sure your disparities are out of the ordinary.
We use a method called ‘bootstrapping’ to figure out the confidence interval.
Bootstrapping - the lowdown
Bootstrapping is used when data isn’t normally distributed. For our undercut data, it is heavily skewed to the right.
For our confidence interval to work we need the data to be normally distributed and look like this. So, in order to get our confidence interval we use the following steps:
- Collect the hourly undercut rates over the last week for your hotel.
- Take a random sample, with replacement, (that means putting the information back before sampling again) from this collection to create a new one, now work out the mean from this new collection.
- Repeat step (2) 10,000 times to get a sample with 10,000 means. (phew again!)
- Create an artificial normal distribution of these weekly undercut rate means, all 10,000 of them!
- Calculate the upper bound on the confidence interval. This is found by adding 2.56 x the standard deviation to the mean undercut rate.
The threshold is then 3x the difference between the upper bound of the 99% confidence interval and the long-term average undercut rate.
What can you do next?
- Sort it out. You'll want to immediately check out and address the cause of the surge. It is often a mistake that can be quickly and easily fixed.
- Track it. You might find that there is a pattern, do you get an alert weekly, monthly? If so, there might be something you can do to stop the surge!