This is the continuation of a series of articles spun out of tnoozLIVE@TTE, recorded live at the TTE event Olympia, London. More clips to come! To learn more about how to bring tnoozLIVE@ to your event, please email Ella Sopp.
Personalisation has been the buzzword for some time but we're starting to see some examples of it coming to fruition in travel. Linda Fox, tnooz managing editor, sat down with bd4travel co-founder Andy Owen Jones at TTE to discuss some of the trends. Here, he discusses how some of the larger companies are adopting a test and learn approach, getting better at traveller intent and applying elements of personalisation.
Linda: Morning we are back with tnoozLIVE@ TTE this morning. I've got Andy Owen Jones with me from BD4Travel and we are going to talk about personalization and some of the trends in the industry. Andy thanks for joining us and can we start with BD4 travel and the kind of milestones that you've seen, in the last year or so.
Andy: Apart from trade shows, which when you're working in travel technology, the trade shows always seem like quite a milestone.
Linda: They do feel like a milestone
Andy: It's because you always have to deliver new things and develop them. I think the biggest things we've done are work with larger customers at scale, so clients that we've been working with for a while like Expedia we've been working in more places with them. Our relationship with Travel Republic has got very close and we are able to deliver more and more products to them based on the different approach that we bring. I think what we've seen is as people have got more understanding of AI in the market place we can bring stuff out of closet that we have been doing for a long time and say "actually this deep learning stuff does work. It's not something to be afraid of, it brings you real benefits". Then, people start to swing their strategy around and focus more on ok if I knew what everyone wanted what can I do. I think that's really changed the dynamic of things so, I think big milestones have been large customers just getting bigger.
Linda: And I remember talking to you in the past and that was always the frustration for you is just getting someone to try it and then seeing how it went , so as a sort of first moving in with personalization in travel, what are you seeing in terms of the kind of, the biggest challenge, the biggest frustration still for you?
Andy: I think when you work in an OTA or a tour operator and you have an online presence the best thing that you can do is set up a test and learn approach and I think everyone who works in that kind of environment knows that the whole point is you try things out and some of them work and some of the don't. I think it has been a while to get our customers used to the idea that you've got to try things, and if you don't try something you'll never learn. Those who have started to understand that and started to understand that if you have got absolutely excellent data, you can learn about things and you are able to move far faster than those who don't.
What I think my biggest frustration is clients who want to have an out of the box thing that's delivered and they don't necessarily want to learn they just want a product that ticks a box and then move on. The one's where it works really well, we learn together and you go down a journey of saying ok let's take it away from averages and segments and take it down to the level of the individual. If you knew what every person on your site wanted what could you do? That's something which we're right at the early stages of learning it so far and the clients we've worked with who understand that we have superb relationship with and we learn together. We build new products together, we design things together and we get deeper and deeper into what it means to address a customer's user experience in a good way.
Linda: So, for me one of the biggest elements of that personalisation is the profile, the customer's actual profile and yet someone was saying to me earlier today that we are still only able to capture about 30% of the customers profile, is that your experience as well?
Andy: It depends what you mean. I don't think customers know a lot of the time what they want and so yes we can only capture certain things about them but we can often understand what they want before they do. So, now we don't know 100% about them but I think if you go to the shop you don't necessarily know I want that particular jumper or I want those particular trousers, it's when you compare things and look at the prices and start to try things on that you start to see I want that one. Does that mean you don't know enough about me if you can't predict that in advance, maybe, but as people search for their holidays and start to compare things we can tell more and more and more about them.
I think the key to personalisation in the future will be, can we give them information back to say we've looked at what you're looking at, we understand what you're after, here are all sorts of interesting ideas for you. I would say the maturing in terms of understanding about personalisation is not we know exactly who you are and we know your past history therefore we can predict this is what you're gonna do. It's we've got an infinite world of travel, we understand what you're comparing, we can help you have an insight into your own desires and then we can surface products that are going to be interesting and stimulating to you. So, yes I don't think we can know everything about someone, and I don't think people want to give you all their details but I think we can learn so much about individuals and feed that back to them and say "look, there are so many great hotels or so many great trips you're never going to find if you rank them all in price order and just go down them from start to finish", but by understanding what you're comparing we can start to say to you "look actually, we think for this holiday don't try that, try here and for this hotel, don't try that kind of thing, try this".
And, I don't think you need a 100% profile to do that, you just need to be better than guessing, putting it in price, putting it in these things and that is something that the nice thing with artificial intelligence or machine learning is you can learn with every interaction, whereas if you're trying to say let's have their profile and if this person does that then do this as a kind of business rule, you'll never get more than 30% of their profile and even if you do it won't bring you any benefit.
Linda: So when I talk to companies like Expedia everything now seems to be about intent and is that the kind of development we're going to see this year, when it comes to personalisation?
Andy: Very much so. We have four main sets of algorithms that we work with, one of which is where are they in the funnel? Are they looking for inspiration? Are they ready to book? Or, are they just planning? I think if you know they're intent at that phase you can say do I want to keep them interested or do I need to help them through the booking process as fast as possible? Then, when it comes to New York hotels, there may be 1500 hotels in New York, 1350 of them will have a pool and a veranda so there's no point understanding that they want a swimming pool, everyone who goes there wants a swimming pool. What you need to understand is, for this particular holiday, how are they weighing up criteria and so I'd say there's no such thing as a concrete intent, there's always a contextual intent and a weighing of alternatives.
For me intent is one part of the puzzle, the other part is how good is your content. If you can't understand what that hotel is good for or what that tour is good for, or what are the attributes of all your products, there's no point in understanding a customer's intent because you're learning shows them the same thing anyway so the two things go very much hand in hand. What does that person want and how good is the content I'm putting in front of them and I think that's what most shops would call retailing outside the travel industry. They would say let us try and understand the customer, let's offer them something that's good and something that's ideal for them and then something that's cheap and give them some choices, building in margins, building in all those kinds of things but really giving people the chance to choose.
So, I think the concept of intent is very useful but out of context it's as meaningless as trying to say what is my score and it doesn't really matter what your score is it's how you use that to help people make decisions so Netflix films scores are different for me compared to my daughter.
It's 98% suitable for me, 30% suitable for my daughter and vice versa. She likes all the action movies, I like all the romance movies, you know that kind of thing. We would end up having to make that decision very contextually where that person is in their journey, what kind of holiday they're after, where they are in their choice process and so on and I think intent on it's own is not a very viable concept but combined with context and combined with product understanding it becomes very useful.
Linda: So just a final question, somebody said yesterday and it was in context of chatbots that people know that what's being delivered them is being delivered via machine they're much more forgiving than if it's being delivered by a human, do you agree with that?
Andy: For now that's probably true in terms of chatbots, but I think it's partly because people have been trained to get used to clunky interfaces. I think what you'll see with chatbots is as they learn they will become as good as people in terms of answering rudimentary routine questions and they'll be far more patient than people are.
Where I'm not sure, I think it'll take a bit longer is in voice activated stuff, because people are prepared to put up with things which are slightly more abrupt or where they can say I don't know the answer to that but I'm not sure people will use that channel very much. They want someone who can get to an answer.
In terms of personalisation itself, I think people don't realise that the machine is doing the personalisation a lot of the time. If you get editorial content that's nicely chosen it can be very good. The big difference for me with editorial content and machine content is you can't learn very often from editorial content, you can learn once a day or once a week, with machine content you can learn a million times a day so it's intrinsically going to be better for routine jobs. It should learn faster so even if it's clunky now it will be better going forward. I suspect people will be more unforgiving to machines than they are to people, because machines should have infinite information, infinite learning opportunities and people get use to the machine having the right answer and if it doesn't, people will get impatient.