Fuzzy Ranking and Filtering
The Fast API recognises the importance of ranking and filtering as the key to reducing often enormous amounts of options to a very short list of exactly the right options for the user.
To do this it uses an innovative technique to consider multiple features of a flight when assessing its value to the user, rather than the black-and-white filtering currently typical within the travel search industry. Travelfusion believe that many users of comparison travel services are not finding exactly the right flight option for them, even though it exists and is available. The best way to illustrate this, is by example:
Using a conventional API
- The user submits a search and receives 10,000 flight options, listed in price order.
- He looks at the first 2 or 3 pages and realises that they are all too early in the morning (hence the low price)
- He uses a conventional filter to set the time to between 11 and 12 - his ideal travel time.
- 2000 options are returned in this time range, but the results are all too expensive.
- He uses the filter to widen the range to between 10 and 13.
- The cheapest flight returned is at 10:05 and is an acceptable price, so he books it. But it was not actually the best option....
- He was not aware that on page 3 of the results, there was a flight at a much better time - 10:55, for only £5 more. There was also a flight that was £100 cheaper, but at 9:55 - outside the time window by 5 minutes.
Using the Fast API
- The user submits a search, specifying his ideal flight time as 11:30
- The Fast API returns 10,000 flight options ranked intelligently considering both the time and the price
- The 'best' flight for the user is at the top of the list - the 9:55, and the 10:55 is in second place. He books the 9:55.
This principle is applied to many different possible features which are all considered at the same time to give a real, human/ fuzzy ranking. Features such as outward and return departure times, distance from airport, duration of flight and price. It is near impossible to find the right flight using conventional filters and interfaces because the user needs to manually scan through all the flights comparing the differences between them all with all these factors in mind.
In addition to this, the service offers a 'weighting' function, so that more importance can be given to one feature over another. For example a weight of 80% could be given to price and 20% to time. This feature can either be offered to the user or it can be hidden and the service provider can decide on appropriate fixed weightings.