Tariq Shaukat
Analyst · Cowen.
Sure. John, I'll take both of those. On the international expansion, our focus is really two-pronged. Part 1 is continue to expand into new markets where we see opportunity. Part 2 is where we have already expanded to go deep -- and deeper and deeper into those markets. The job is not done when we launch in a country. And on both fronts, we feel very pleased that the playbook that we've got is working really well. So if I take the go deeper piece, as an example, we're still seeing rapid growth in Germany in the DACH region, the German-speaking regions in France and other parts of Western Europe that we have launched in. And we're expanding both to more segments in our core cities and to new cities. That's also true in India, where we're seeing very, very rapid growth as well. And a lot of that is, if you will, geo expansion inside of India as well as continuing to go deeper in the key cities like Mumbai that we're already in. So that playbook, we think, is working well and is allowing us to continue to gain market share. And to refresh everyone's memory, the historical focus has been in Western Europe South and Southeast Asia, Indonesia, Philippines, Singapore and parts of Latin America, particularly Mexico and the southern parts of Latin America. As it comes to new market launches, we're really focusing this -- at least the first half of this year on the other markets in Western and Central Europe that we believe have a strong opportunity. So -- and I think Whitney mentioned Poland, Italy, Portugal and Nordics as key markets we'll be focused on. But again, it really is in combination with these go deep efforts that we've got as well. Moving on to the recommendations engine, we are absolutely expecting this to be a global rollout. We -- as you recall, we obsessively test our products to make sure that they work. We do that with different segments or in different geographies. So for this machine learning-based recommendation engine, we started it in certain markets in Western Europe, are seeing great results, which I'll speak about. And so now we are actively training the model in other geographies so that we can bring it globally as we use it for the Best Bees offering as an example that Whitney mentioned. We would expect over time that to be a global offering, not just a Western European offering. And the results that we're seeing, as you mentioned, are really quite substantial. We are seeing double-digit increases in match rates when you're presented with this sort of super compatible recommendation versus our already very good recommendation engine that we've got, the non-machine learning centric model. So we think that we're really adding a lot of incremental value on top of an already strong product, and that's what we think will allow us to monetize this further.