Travel news for the trade community

The best-funded AI company in the world is now delivering results for OTAs – we take a look at the product in action

Product generates €284,000 in additional revenue in under 6 weeks

Imagine waking up one day to find that your website has magically become 26% more effective at selling travel packages. That is, effectively, what Sentient Technologies – a company that uses artificial intelligence algorithms to optimise websites, whose founding team developed the technology that became Apple’s Siri – are trying to achieve.

A recent project led to an projected RPU (revenue per user) lift of 26% and a 1.6% lift in conversions for one major operator, from 2% (1.3M users) of the company’s entire traffic. This generated €284,000 in additional revenue in under 6 weeks. I caught up with the team via video call to take a look at their software.

“We offer intelligent commerce solutions and web optimisation to help with results for clients such as OTAs; these include website structure and filters, navigation, price sorting, even colours and branding,” said Paul Jarratt, Sentient’s head of marketing communications. “We started, 10 years ago, as an evolutionary computational system for a hedge fund but now we have also moved into travel.”

“For most travel operators, their business goals are explicit: leads, bookings or revenue. To optimise a website for one of these KPIs, you could require, say, 20 different ideas in almost infinite combinations. We use artificial intelligence to give marketers the ability to analyze data and make decisions on a much larger scale.”

Paul continued, “The big problem we face is that less than 1% of data is analysed. AI gives us the ability to find meaning in the data and to scale our efforts faster. Things move so quickly nowadays it’s hard to keep up if you manually manage it.”

“In a traditional AB test, you create a new hypothesis that you implement and see if it improved your results and then you repeat until it gets better.” This is applicable ad infinitum, of course.

Things move so quickly nowadays it’s hard to keep up if you manually manage it.

“Instead of testing one concept, we can give all of our ideas at once: a number of things, from moving product filters on the search box, or we can put the cheapest flights at the top. There are tens of thousands of combinations and we can improve conversions and continually test and harvest for generations of improvements to get, sometimes, up to 30 to 40% better.”

How the magic happens

The Sentient team took me for a test drive of the product via video screen sharing: sales engineer Sam Nazari showed me that users can change the copy of their buttons in a heartbeat, to, say, ‘Book a vacation now’ instead of ‘Purchase’, or change the background images, or use a hex code picker to change the colour of their buttons. As users visit the site, the system promotes layouts and channels that perform well.

“For traditional AB or multivariate testing, only 1 in 7 tests is a success,” Jarratt explained, “But for AI-driven optimisation, usual results include 1-2% revenue increase in over 100,000 visits and up to a 30-40% increase in revenues for some clients’ conversion rates including for a well-known tour operator.”

At the system dashboard, you can see reports and live tests. Nazari tells me “We can optimize the whole funnel for a conversion or add a specific conversion, such as a customer arriving at a confirmation page, or a click or even a manual trigger via an API,” (delivered via ReactJS for the techie types).

One single line of JavaScript is enough to create all the reporting and site interaction. It’s fully integrated with Google Analytics out of the box to allow for granular reporting on metrics such as time on page.

“It’s different to AB testing because you can test various things simultaneously. For example, CTA and copy and logo. There’s a 10% chance of having the control as your testing candidate and you usually run a test for 30 days although of course, huge OTAs can test a lot more combinations in the same time because of their large user numbers.”

“After around 3000 conversions, four generations of evolution are complete and you start to see significant changes. For reporting, we usually have about 8 weeks or 450,000 unique visitors for 36 unique variations. You might get 10 million combinations in an eight-page sales funnel,” Nazari confirmed.

The evolutionary algorithm prunes the poor performers and optimises the site automatically – and there have been some surprising results. “One previous client with green and blue branding ran a test and one of the top three performing ‘Buy’ button CTAs!” Nazari smiled.

Is this the death of branding guidelines as we know it, then? Would you be happy to say goodbye to the blue and green you loved so dearly for so long, in order to have a garish salmon button?

I certainly would be happy if I could see a 40% uplift in bookings!

You might also like