Open source is key, says Lastminute.com CTO

Lastminute.com has earned its place as the UK's dot-com darling by the sheer fact that its business model actually stood up to the market reality-check that killed of so many of its peers.

But although it may have earned its business spurs -- the small matter of profitability aside -- the company doesn't have the same reputation as a technical innovator as its US counterparts such as Amazon, eBay or Google or the UK's Tesco.com.

The incredibly over-the-top interactivity that Boo.com inflicted on the Web -- associating technical innovation with business disaster in the mind of investors -- probably has something to do with Lastminute's low-key tech reputation to date.

But that could all be about to change. Previously head of technology for Walt Disney Internet Group, Chip Steinmetz has been quietly overhauling the dot-com favourite's architecture since his appointment as CTO almost a year ago.

Re-engineering the site around Linux and open-source technologies such as JBoss Web middleware is just one part of a plan to cut page-load times and improve the user interface. The wider picture is to boost Lastminute's image as an e-commerce innovator and services provider while attracting more partnerships such as the one recently announced with TheTrainline.com.

Steinmetz believes that the average travel search on Lastminute is as complex, if not more so, that anything Google or Amazon has to cope with -- which is why his department has so many PhDs keen to work on its unique computer science problems.

ZDNet UK spoke to Steinmetz about the trade-off between usability and speed and the importance of Linux and open-source to his mission.

You joined Lastminute.com around a year ago. Are you happy with the technology platform you inherited and how do you want to develop it?
Last summer was about trying to handle sales growth but now we're concentrating on making the site more reliable. When you're throwing a Web site together at the beginning you might not take all the energy to architect and engineer it as you would have liked if you had realised it was going to be that successful.

But the basic platform you can keep adding to for some time. On the hardware side we bought IBM blade servers. You can keep scaling those up pretty fast and a lot more easily than having big farms of PCs like Google and Yahoo have. It's an advantage for us being a little later in the game than them. We have the chance to take advantage of some new technology.

Were you happy with the page-load times when you arrived?
Last year average page load across Europe was about 12 seconds, including people on dial-up. It is measured by a third-party company Gomez Services with about 1,000 nodes around the world. If you looked at Expedia at that time they had a page-load time of about nine seconds, Travelocity was about nine and a half seconds, and we put a bunch of infrastructure in place and we got it down to around four seconds' average page load.

Did that hold up over Christmas? Some analysts put your load times at 20 seconds during that period.
Generally I was pretty happy over Christmas, actually. Relative to the Christmas before I got here, 2003 was orders of magnitude better and faster. It's a lot faster than most travel sites -- a lot faster than Opodo for sure.

Those slower pages were probably oriented to products we didn't pay as much attention to like gifts. So over Christmas a lot of people ordered gifts and we didn't optimise that; everything else was fine. We had plenty of extra capacity -- 50 per cent extra -- we could have handled. So that is available for growth into the summer high season when we can use it all up. We concentrated on areas that were high transaction volume first and made sure we could handle those, now we have to address everything. All the gift stuff and all the things on the site that we call lifestyle products -- there are implementations going in this month to improve those too.

Is a lot of your job about managing the balance between the usability and the speed of the site?
We have lots of dynamic content and all that can add a lot to page-weight. We have this constant trade-off of how to make the site pretty and interactive but also responsive. So we're spending a lot of investment on a new generation of Web interfaces. Over the next few months we're looking to add a whole extra layer to the architecture. That will involve everything from better colours to better fonts and easier to use navigation.

Travel is a lot harder to build sites around than books. Books are static. With flights and hotels the price changes at different times and different intervals. And things can get sold out from under us, where as books don't usually get sold out from under you. Our problems are pretty difficult. So to give a really compelling user experience with recommendations and personalisation is really a hard problem in travel.

What parameters do you use to test the usability of the site -- are you using focus groups etc?
We actually hired Jakob Nielsen's consulting company. They did a study across the whole site and gave us a presentation of all their findings last week. Boy, was it great. It's very difficult internally to see that kind of stuff anymore; you get used to things, so to have this third party is great. They had a recommendation which we took to heart, which is we should do a lot of testing with real customers. You have got to have your customers give you real feedback.

I am actually pretty upbeat about the improvements. This is not a massive investment to make it look great -- why not? This is the biggest e-commerce site in Europe, what's bigger? I mean really, as far as people doing development here. I mean you've got the outposts of Amazon, Google, etc. but as far as transactional volume I think we're the biggest. So if it is just this much work too make the site look great competing against other US dot-coms then why don't we do it? Plus I think it does drive revenue.

Do you have to argue that in terms of funding - that any user interface improvements will have a sales comeback?
I don't think this is a big investment at all. I think this is in the course of normal work -- just following rules that are next generation design issues e.g. just using the right font so not everybody has to have perfect 20/20 vision.

Don't you think some conservative board members might think, "If it's not broke then why mess with it?"
I think you could get into that situation some day. Remember the cycle we are in now we are racing with two or three of our major competitors. Consequently in every product area we are improving the site so it's not much work to say "Guys, make sure you follow these basic human interface rules and make them the rules of the future".

There were some technology partner announcements between Lastminute and Microsoft last year around .Net. How much of the site is actually engineered around .Net?
We primarily stayed with open-source technology. Almost everything we do is in Java. So it's all J2EE. Almost all the operating systems are Linux, we do have some Solaris on our databases and we use open-source middleware, which is Jboss. Almost all our development tools are from the open-source community. The guys love to deal with open source and get involved in the community. If we want to make changes to Jboss -- we usually get the changes overnight. Try doing that with Weblogic or Websphere or whatever. The trade-off is that you need guys with a lot of expertise. But since what we have to do the rest of the time is very high-tech then we have to have those guys anyway. So it's no additional investment to us but to people who aren't building a big e-commerce platform it's an extra cost.

Have you ever considered Linux desktop internally?
Sure. Some of the guys have. But we have so many connections to suppliers plus we're a big email user and getting full email connectivity would be difficult at this stage. We email all the time -- we send email confirmations of electronic tickets, so all that has to work well, that would be my own worry about converting from Windows to a Linux desktop. It would mean having things like Ximian working seamlessly without us having to go through too many integration issues. It is almost there and some people are doing it but I think they are mostly internal projects that they have a compelling reason for. We might be able to save some money over time but I think it's something to look at on the horizon.

With a lot of these relationships Lastminute has with like Trainline, Tesco.com, Iberia.com -- is a lot of your time spent with integration issues?
Well, yeah, 20 percent. Integration is at multiple levels though: there's connectivity, we might get a data feed from somebody else or we're feeding them. In Trainline's case we give them our inventory too. It's a two-way Web-services or XML feed but the data's different so you have to normalise all that. We have very sophisticated communication architecture. Probably more sophisticated than any other travel company, which is an advantage because once you've figured out how to do it, you can connect new people more easily.

Do most of your partners have the Web-services or XML expertise to contribute to the relationship?
I think in about 80 percent of cases they will have in-house expertise but in 20 percent of cases we will help them or accommodate whatever they have.

Do you feel that Lastminute is a software or services provider now as much as a travel portal?
Oh definitely. With the hook up with the Trainline, we wrote an XML specification that was a 100-page document of schemas, back-up and recovery, etc. They started writing to the specifications so it's exactly like you're a software provider. And a lot of the guys I have hired have worked in software companies and have built real products.

Click here for Part 2 of this interview.