09 August 2010

Mirror Mirror On The Wall Who Is The Fairest Of Them All?

If you look at the meta search sites and try and determine which one is best for you... there are a number of subjective criteria. But which tool is best? Who actually serves up the best fares for the consumer?

According to the Sunday Times of London it's Momondo. Congrats to Thorvald and the team in Copenhagen for this honour. Runners up were Kayak and Skyscanner.

However - I will use this opportunity to pick on the subject matter and its inadequate value that results in a poor user experience.

In my alter ego's life, my team have been working on result optimization. As those deeply involved in Search and IBEs know this is a subject matter that is incredibly complex. There are just so many variables.

In looking at the problem there have been two major variable elements which contribute to the quality of the results.

The search system (which covers a lot of areas of the technology)
The source data

In recent years the quality of the technology in search has improved significantly. However the quality of the data and the manner of which it is stored, served and refreshed has not become any better.

So you can have a great engine and lousy data, a great data store and a lousy engine resulting in untrustworthy results.

Then of course there is the result that just is different from the one that you buy. For a consumer anything that is less than 100% spot on represents a lack of trust. I spoke to Jim Young (formerly Marketing Head at Frontier Airlines and now the Honcho at Open Axis Group) in April about this problem. He had an interesting perspective on the issue. Per their research at F9, they could get to a three 9s (99.9%) accuracy if they had a $1 tolerance level for the GDSs. For a 1 cent tolerance level - the number dropped to just over 80%. (Prof note - this was before the merger with Midwest and Republic so I expect this has gone down significantly since then).

In my analysis of the problem the complexity of the solutions and the loose adoption of rules have shown that there are a lot of right answers. These are not false positives they are actually right and valid. These range from elements like the tax tables used, the order of the calculations and the translation of currencies. In several tests we have run this can mean up to 20-25% differential in pricing. More noticeable at the lower total price level than at the higher priced tickets.

With increased complexity of taxes (The US is going to implement another increase in September for visitors) and the introduction of ancillary services and the somewhat shady use of taxes masquerading as really airline fees (like fuel surcharges) - this is not going to get any better.

And don't expect any valid benchmarking to happen any time soon. All attempts thus far have resulted in failure or some self serving results. Each GDSs for example claims their pricing engine is the best.

A little bit of history, when the Professor used to work at one GDS in the 1980s and 1990s - we saw that our pricing in comparison test results sucked vs the others. We had a low tolerance for failure and demanded that our engine adhere strictly to the rules. We found that that the other players didn't quite do that. When we loosened the rules and matched them - the resulting improvement in our pricing engine was dramatic. No one was actually doing wrong but manipulating the rules resulted in lower fares and prices for the consumer.
The airlines who owned us were not very happy about it initially but in the long run were ok with it. The alternative was paying more fees to the other GDSs and losing opportunities for sale.

So the answer is this is a war of biblical proportions. And you can see why Google with its patina of global trust wants to have at this problem.

Anyone serving up fares in this market is probably quaking at what Google can do now with ITA's search/pricing engine and certified data sources. I bet the GDS old timers are now kicking themselves for their short sighted view of not responding to the airlines need for a solution that ultimately allowed ITA to exist.

But remember my opening statement. It is about the definition of subjective criteria.

So the fairest of them all depends almost invariably as to who is looking in the mirror....


No comments: