TripAdvisor · B2C · travel review site
Sort hotels by each traveler's past behavior without calling it 'just for you'
Unlabelled version converted better; labelling it 'just for you' dropped conversion
What they did
What happened
- Stage
- scale
In their words
How Tinder Captures More Value With Tiered Pricing and Consumables — Ravi Mehta
Jacob Eiting… times, looking for hot people to go hang out with maybe date. Who knows? That's a very, everybody knows how to do that, right? There's less to explain there, you know what I mean? It's like meet this person, not meet this person attractive, not attractive, right? Or to you, whatever, stay away from the messy middle, you know what I mean? Know who you are. Are you the long onboarding? Are you the very, very short onboarding and go hard on either?
Ravi MehtaI'll give another really interesting example, and I think this all comes down to motivation and confidence. You want to keep a person highly motivated during the onboarding process and that you want them to be confident that what they're getting is what they want. At TripAdvisor, we had tested out a new algorithm, so it used to be that you would always see hotels sorted by their traveler ranking, so by their average review score. Instead, we personalized that based on your past behavior to show you hotels that you're more likely to convert on. We did some testing without doing anything other than just changing the algorithm, and we saw that people were actually converting much better with the new just for you algorithm, and then we rolled it out as just for you and said, okay, now your default is just for you. We've looked at your past behavior, we're showing you a set of hotels you're likely to convert on, and as soon as we branded as just for you, the conversion rate went down and the reason that the conversion rate went down when it was branded just for you versus when it was sort of a blind algorithm, is that travelers don't want you making decisions for them.
Ravi MehtaThey looked at that and said, TripAdvisor doesn't know what's just for me. They don't know anything. And so people converted less well, even though when we just did it for people and didn't tell them they converted better. And so there's a combination of both actually getting to the right answer for the user, but also creating confidence around that right answer. And that might be a longer intake, that might be a shorter intake.
Get tactics like this every Monday
New tactics from the week's founder interviews, each linked to where it was said.
More from this episode
- Offer three subscription levels and sell extra boosts on topFormer CPO calls it one of the best ways to price consumer apps; no revenue split given
- Keep dating free for most users so there are plenty of people to match85-90% never pay; the 10-15% who do fund the business
- Interview users spending $100-250 a month, then build a top plan for themLed to Tinder Platinum; no results shared
- Test many prices for each plan and add-on, country by countryHundreds of price SKUs live at once; no test results shared
- Sell a paid 'super like' that stands out from a normal likeSuper Likes are 3x more likely to get a match
- Let people start swiping in two minutes using Facebook login and photosOnboarding cut from ~20 minutes to under 2; brought younger people into online dating
- Replace a 3-screen checkout with a 20-question intake before paymentConversion up (host says 40%); 99%+ of people answer each next question
- Show ads to free users to cover their costSmall share of revenue; ads worth far less than a subscriber
- Start at $20-30 a month; if nobody pays, fix the product, not the price
- Give a short free trial that builds a daily habit, then chargeGuest converted without hesitation; a free-tier rival he uses daily never got paid
- Sell live classes, Zoom sessions and week-long intensives to your most engaged usersRevenue almost doubled; live offers now half of revenue
- Play rain sounds the moment the app opens, and in 15-second ads
- Set quarterly goals as a short why, 3-5 commitments, and a task listGuest says teams that adopted it found it works better than OKRs; no numbers
Tags: web, personalization, ranking, framing, trust