David Goulden: In terms of — Lloyd, let me just remind you that this is a year that we still expect to be leaning into marketing and merchandising because we believe there’s recovery in the travel market available to us. And I think that our growth rates relative to the market demonstrates that we’re making progress there. If you remember, we started off the year by saying we expect our marketing and merchandising investment to be roughly the same as it was as a percentage of gross bookings, same as it was in 2022, we still expect that to be the case. Remember, we said that 2022 was a year which we look out of leaning in and making more investment relative to gross bookings that we did in 2019. And that story for us hasn’t changed.
We believe that we want to continue to lean in this year to continue to gain share in a recovery marketplace. And the fact that we saw some improvement in Q1, as I mentioned in my — in the answer to my question to Justin, what to do with the kind of shape of our booking profile and the booking window and getting more summer bookings with higher transaction values than we would expect it to get from a mix point of view in Q1 and actually why the ROIs came in better. But we still plan to believe in the industry. And as I said in Q2 and for the full year, we still expect that investment in marketing and merchandising to be as it was in 2022.
Operator: Your next question comes from the line of Brian Nowak from Morgan Stanley.
Brian Nowak: I have two. The first one is on the direct mix of traffic, the percentage of the business is direct. You’ve made a lot of progress on that over the last year or so, a couple of years, Glenn. If you break it down by region, where have you made the most progress in increasing the percentage of the business that’s direct? And how should we think about how high that is? And what are the regions that you have the most runway to sort of drive that percentage of the business that’s direct up longer? And then, the second one on your forecasting. You guys have a lot of data and you’re very good at forecasting. It seems like demand really came through better than you thought throughout the quarter. Which regions sort of drove that outsized demand versus your forecasting?
Glenn Fogel: So we don’t break down the direct mix by region at all. But I can repeat things I’ve said in the past about how important direct mix is to the long-term increase the value of the franchise because of things we talk about with our connected trip. They’re getting people in through the app primarily, having them understand all the different things we’ll be able to offer to them, give them a real personalized thing that will make them want to always come back to us. The more we learn about them, the more they come back to us. So it’s a flywheel effect happening there. Again, I won’t break it down by region but my goal is to have every region have as many people as possible using the app and coming directly. That’s the first thing. In terms of how we varied against our forecast, I don’t know how much. David, why don’t you get into that?
David Goulden: You can see that — obviously, we did significantly better on the top line. And I think the place to go to — I’d say it was fairly balanced across all regions. Obviously, the price we sold on booking window expand the most was in U.S. and Europe. So that’s where we got a higher share of summer bookings than we expected to get in Q1. But you see also Asia did very well. And in general, also exceeded our expectations but we didn’t have the booking window phenomena there. So I wouldn’t call out any particular region but I will call out some of the differences in what we saw across the regions that contributed to the overachievement.
Operator: Our next question comes from the line of Doug Anmuth from JPMorgan.
Doug Anmuth: Glenn, I just wanted to revisit on AI. Can you just talk about some of the advantages that booking would have in leveraging generative AI versus what other external or potentially new travel services might provide? And do you feel the need to protect the data on your platform to keep it from being used for training broadly across many large language models?
Glenn Fogel: Thanks, Doug and very interesting questions about that and I’ll break it out into — you start off with AI, then we went to a specific subset of degenerative AI. Let’s just talk in general about AI first and how important this is into our business for many, many, many years, more than a decade perhaps, a machine learning model that really helped us do a better job making sure that what we’re presenting to consumer, for example, is really something that they have a higher percentage chance they’re going to convert on that. All sorts of ways that we use, very sophisticated machine learning models in many parts of the business that have helped us get us where we are today. Then of course, we come out in the fold, something that is — the somewhat a lot of people were not that aware of this generative AI, large language models and see what could be done with them.
And clearly, anytime there’s a major shift in technology, everyone thinks like, well, is this going to be disruptive to the players who are already doing well? Is it not? And I said in my prepared remarks, how I feel very confident of where we are in this because of the work we’ve done in AI in the past, the number of people we’ve had working on, the amount of capital we have, our collaborations with universities. I believe that we are going to be benefiting greatly from this new type of technology in many different areas and some we haven’t any thought of them yet and some things are going to be easy, perhaps increasing the productivity of our developers which we believe is hopefully going to achieve some very good results in the hopefully not so distant future, to things that perhaps are further away but ways that people interact with us in that connected trip vision, a way that you really do are able to recreate that human travel agent into something that’s actually an automated player but is that does so much better than the human being did in the past?
Now in terms of your very important question about the data and how do we protect it and not litigate it out and be used by others, that is a very important thing. And everyone from our attorneys to people in our development departments, people who are really working with some of these large language model foundational model people, we are looking very closely. Have they already been using our data and should they be — and I think there’s going to be a lot of interesting regulations in this area that nobody knows the answer yet, too but how will people become and say — if they’re data has been used in the past for training purposes or not. And I think it’s a very open question that nobody knows the answer to yet but we will be very interested in the results.
Operator: Your next question comes from the line of Kevin Kopelman from TD Cowen.
Kevin Kopelman: Yes. First question, I wanted to ask about ADR’s big area of interest. With all the strength you’re seeing in Q1, could you just — and quarter to date, can you touch on how ADRs have trended generally compared to what you’re seeing as of the last call.
David Goulden: Yes. So when we gave you some data on the call, I me
