We caught up with Andy Owen-Jones at Travel Technology Europe in London, to talk about bd4travel and its impact on the digital and legal changes in the distribution space of the travel industry. Here’s what he said:
TD: How did bd4travel come about?
AOJ: The theory is now that 98.5% of online digital marketing efforts are wasted, even with AB testing. You take variant A and variant B and you use the winning configuration for your site, so by definition your site becomes more and more suitable for selling. However everyone is different and at various stages of the sales funnel, so you start off by looking for a family holiday in Majorca and you end up renting a villa in Crete!
On a shop floor, you buy what you want rather than by forced special offers. We set out to understand what things are driving customers’ decisions and how much they want to spend. We’ve been working on a AI model for the last four years and now I think we’ve got probably the best model in the market. We use that for our clients to help them see, in realtime, what their clients are after.
Our big clients are Expedia, Travel Republic and Holiday Check and a variety of others; these are large businesses regularly receiving more than 100,000 users per month so we have enough data to make things work.
TD: How does the software work?
AOJ: It’s important to remember that our OTA clients have never seen their customers! When we first met one client, they had a cardboard copy of a person in the corner of the office to remind them that it was people they were selling to. We have built a way of visualising your customers in real time; we’re not making personas that you then set business goals for.
One of our algorithms can work out how much they are likely to spend and how elastic that is. The second one is their stage in the sales funnel: looking, planning, ready to book, or client. The third one is their propensity to churn – not come back for at least three weeks – and the fourth one is what they actually want: is it family friendly, with Wi-Fi? Or with a pool and adults only? We pick up a couple of thousand data points for each person.
We looked across standard e-commerce models to see what we could bring into travel and didn’t see many things that worked. For example, if somebody buys a book they might buy another one two weeks later. In leisure travel you’ve got maybe one shopping basket per year. We started working on an AI-driven model to infer what people want; our goal is to say that ‘For this person, in this context, at this time – what is the best guess we have? Can we understand that even before they do?”.
TD: So how does this benefit the consumer?
AOJ: If you take 1200 hotels in Majorca, pretty much all of them will have a pool, so put in a filter called ‘Pool’ and you will still have 1200 hotels. If you then sort by price it doesn’t tell you anything, so we go very granular. We can see the hotels they have looked at, their ‘DNA’, for example price-sensitive, and what source they came from.
We can show a deal specially chosen for you on the landing page: if we think you are sporty, we will show you sporty themes. If we think you are a honeymooning couple we can send a time-sensitive voucher, where if you book in the next 20 minutes, you will get a free dinner on the beach.
Within 50 milliseconds our goal is to put the website’s deals in the right order for you, given your behaviour. Then we will pull out relevant activities that are accepted or declined – you can basically ‘Tinderise’ travel!
Because we know the DNA of a hotel we can suggest alternatives nearby which are similar – or maybe even more suitable, just in a different destination, almost like Netflix’s ‘Suitable for you’ content. Now we’re getting more esoteric accommodation, how do you surface the most relevant? Most algorithms are crunched on volume and not being interesting.
We’ve modelled 600,000 hotels, plus cars and destinations. The more data we use, the better our recommendations become and we can do some quite interesting guided things. For example let’s say somebody is likely to convert. Just for that individual, you can create a personalised promotion if they book in the next 20 minutes.
We can also tell a tour operator how many non-availability issues they have had in real time and you can set alerts if it hits more than, say, 8 per second. Then we will swap out that provider and swap a new one automatically.
“We want OTAs to have the same level of visibility and control you would expect in a shop – backed up by infinite information.”
It’s fair to say nowadays that 85% of people won’t convert on your website. But everybody that comes on to our sites will build a profile. If you’re going to go after people with advertising money, why not target the qualified leads who are ready to buy?
You can easily eliminate 60% of customers and spend your money on the right people. Triple your bid! We are currently experimenting customers with how much to bid for truly qualified leads based on their behavioral profiles and it’s really bringing down the cost of acquisition.
TD: What are the implications of GDPR (General Data Protection Regulation) for bdtravel?
AOJ: With GDPR, it’s a pain that it has to be put in place but to be honest the industry has misused data for years, for example remarketing terrible products. I would not give consent for that. But what we see now is an opportunity; like credit scores used to be hidden by banks, now they are really useful – we want to make people’s profiles more transparent.
The theory is that users don’t know what they are looking for until they start searching, so how can we give them the tools so that they can work it out better? There’s a retail theory that you should give customers something good, something better and something best, like with Tesco Finest, but there’s no way of doing that in travel and we’re just throwing prices at people.
GDPR means we have to get a user’s consent – but what value proposition are you putting across to earn their consent? In the past, we’ve been terrible with the data; we used your data to sell you a toaster when you actually wanted a frog! By working side by side with people, we’ll give them a return on sharing their data.
