Cartesian Products
This is René Descartes, a Frenchman, who had quite a lot to say about God, mind-body duality, and travel website design.
Anyone who has worked with Cool Travel on a travel website will have heard mention of René on occasion. But, in case you haven’t had that honour, you might need me to provide a little further detail.
We have already explained, with the help of some nice people at Google and MIT, how difficult it is to find the cheapest fare between point A and point B on date C (and back on date D).
Let’s assume that you’ve decided that the low fare search problem was too hard to deal with yourselves (if you haven’t, please give us a shout – we’d love to help or even have a confidential free chat on the subject) and that you’ve decided to get someone else to deal with that problem. This is excellent if you’ve just decided to sell flights but let’s say you decide to sell ‘holidays’ (for the purpose of this conversation, I’m going to assume that a ‘holiday’ is a flight or two and some ground product).
Holidays
If you do want to sell holidays, you’ve got a new problem, which my friend Ernie calls the ‘Flight and Hotel Problem’, but which might easily be the ‘Flight and Cruise Problem’ or even the ‘Flight and Climb A Mountain Problem’.
This is where René comes in. René spent a lot of time philosophising but in those days you could be a world-leading philosopher and a world-leading mathematician, which is hard nowadays because people tend to specialise more. Anyway, whilst he was being a mathematician, he mentioned to various people that if you take a pretty big number and multiply it by another pretty big number then you get a really big number. The people he mentioned this to thought that this was a very clever observation and so they named the process of multiplying two pretty big numbers together after him, calling the result a Cartesian product.
Let’s say a customer living in London wants to go to New York and we want to offer them a holiday. This seems like a pretty simple problem to solve but, according to booking.com, there are 403 hotels in New York (a pretty big number) and we would get around 250 flights (another pretty big number) back from most low fare search engines, giving us 403×250 = 100,750 combinations (a really quite big number).
If we were a tour operator and decided to package each of these combinations and put a price on them, it would take us a while.
More dimensions
There are not only those two dimensions of flight and hotel. We also have to consider departure date, length of stay, airport to depart from, number of travellers and their distribution between rooms, etc…
To make things a little more complicated, this is the number of combinations that one of our clients decided that it might need to package and price for one small island in the Caribbean:
- 365 departure days
- x (let’s say) 28 durations
- x 16 hotels
- x 15 departure points
- = 2.5 million combinations
- If we have (let’s say) 20 occupancy combinations, then we have:
- = 50 million combinations
- Each with potentially hundreds of flight class and flight number combinations
- = More than 1 billion combinations
- Each with several special price rules for particular clients
- = More than 5 billion prices
All of this is fine when we put a person in a call centre between the customer and the product because they can talk to the customer, reduce the size of the problem using their product knowledge and training, and get something decent and reasonably priced in front of the customer quite quickly.
Handling this problem on a website is a different problem.
Travel website design
The Amazon Approach
Quite often digital agencies making their first forays into the travel industry begin with a standard eCommerce approach. They ask for a big list of everything that a travel company sells and a price for each item on sale.
The difficulty in producing this big list often consumes a few weeks of discussions and is often followed by a suggestion to build a cache.
If things are going well and everyone is cooperating, we can usually calculate the expected cache hit ratios and prevent any wasted development. If things aren’t going so well, someone actually tries to build this cache.
The second (and possibly the biggest) problem with the ‘big list’ approach comes whenever flight costs change regularly, as they do when using revenue-managed scheduled flights. This leads to the customer being presented with prices in the search results which then change regularly before the customer can buy.
With lower product volumes, for example at a cruise company, a big list of priced product can sometimes be created.
The Apple Approach
The other approach to the problem is to have a limited number of products which appear in the search results, each of which can be ‘tailored’ later.
Taking this approach means that the customer is tied in to a small group of products early in the book flow and sometimes has to try and try again to find the best product for the price that they are looking for.
So we need a compromise of some sort.
In the next article, we will examine a group of UK sites to look at the different approaches that they have taken to solving these problems.
No comments yet.