Member-only story
You might all notice the uber price changes all the time. Earlier this month there was a storm when I tried to go to a restaurant, I saw the price almost tripled, so the dinner was cancelled… I wondered how Uber calculated the price?
1. Basic price calculation
Well, I don’t know how the price is set in the first place. Maybe the oil used, the cost of the vehicle spreading out for years and labour. I learned from my last google research[1] and I know this is called cost-based strategy. We can use a competitor-based strategy then. So I googled the taxi rate. The base fee is $2.9 and $2.14 per kilometre, so for 10 km is $24.3. Well, Uber is cheaper, sort of around $20. However, $20 is not peak hour; it could be tripled, as I mentioned above.
The price is dynamic based on the demand-supply in specific areas. Higher price can attract more drivers into this area, or decrease customers who might shift to public transportation. I believe their model has a basic setup as taxis and some cost for running the platform, but how the dynamic part works? An article [2] 8 years ago said those surge pricing only affected 10% of the trips during peak hours, events and bad weather.
2. Dynamic Pricing Algorithm
They must have used a time series model like LSTM behind the scenes, which could predict price in particular events like Xmax. The problem is how do they collect the data in the first place. I mean, now they have tons of data to use accumulated over the years, but at first it might be handcrafted rules? Anyway, now they can use in-house data to make prediction every few minutes, sort of real-time prediction.
The other interesting point or scary one is they could charge you more if AI thinks you will pay more. The normal features for the model might be location, time, weather, traffic, number of drivers, number of customers, and your trip history. Thankfully…
