Mobile Growth & Pancake #1: Growth Through AI with Lomit Patel

On the first episode of Mobile Growth and Pancakes we are joined by Lomit Patel, Vice President, Growth at IMVU, to discuss the role of AI in mobile growth, his growth team, and the status quo of IMVU.

On the first episode of Mobile Growth & Pancakes, Storemaven’s new podcast, our host Esther Shatz is joined by Lomit Patel, the VP of growth at IMVU, the world’s largest avatar-based social media app. They discuss why the best UA strategies must balance acquisition and retention, Lomit’s use of AI and wholistic growth teams.

Lomit shares how he joined IMVU at a time when the growth graph was heading in the wrong direction, yet with his deep-set passion for AI, he was able to implement automated systems that enabled his growth team to break new ground.

Check out all the other episodes of Mobile Growth & Pancakes here

Connect with Lomit and IMVU here: 

Timestamps:

00:47 Introduction to Lomit at his role at IMVU
01:42 Key KPIs and metrics for growth
03:00 Strategy for growth through AI
05:56 Lomit’s growth strategies
12:50 The team driving AI, eliminating biases, and hiring the right people
17:18 Apple, IDFA, and adjusting to change
24:42 Focusing on Android early on
26:39 Quick-fire questions

You can listen to the full episode here:

Or listen on:

Apple Podcasts
Spotify
Google Podcasts

Key takeaways from the episode: 

  • IMVU is the world’s largest avatar-based social networking app. Lomit shares that they are seeing a surge in activity during the pandemic.
  • To drive growth, Lomit’s team is responsible for the entire customer journey from user acquisition to retention to monetization. Lomit has seen both the benefits and downsides of different teams focusing on different metrics, so instead, he suggests having a single team responsible for all growth metrics.
  • The best growth strategies must balance two concepts: user acquisition has to always be considered together with retention. Focusing on one whilst neglecting the other will not lead to sustainable growth.
  • Leveraging digital tech platforms and figuring out how to integrate these  together enabled IMVU’s AI growth.
  • IMVU worked with companies like Snapchat, TikTok and others in their alpha and beta stages helping them figure out how to monetize users with different ad models. 
  • If Lomit were to start again, he would focus more on user retention than acquisition: it is more important to keep users than to get new ones.

“Ultimately, it’s not about how many you bring in. It’s about how many you keep.”

Lomit Patel

Main Transcript:

Esther Shatz: Welcome to Mobile Growth and Pancakes, a podcast by Storemaven. We break down how and why mobile apps grow. In each episode, we invite a mobile growth expert onto the show to break down a specific mobile growth strategy, how it worked, why it worked, and what they would do differently. I’m your host Esther Shatz.

Esther: I have with me today, Lomit Patel, who is the vice president of growth at IMVU. It would be great if you could just introduce yourself real quick to the audience.

Lomit Patel: Yes. Hi, Esther. I’m so excited to be here with you today. Just to let everybody know, my name is Lomit Patel. I head up growth at IMVU, which is the world’s largest avatar-based social network app. It’s like a role-playing game where people create avatars and create these virtual worlds where they can meet people from all around the world. It’s really popular. We have millions of users that play on our game today.

We’ve been growing rapidly before, but the shelter-in-place has definitely helped us. We’ve seen an abundance growth, especially on the organic side. Primarily, like a lot of other social and gaming apps, it provides people an outlet to pass time. With our game, it’s more about continuing to have that human connection, right? It’s been doing really well, fortunately.

Esther: Always good to have some good news in times like this. I’d also love to understand, when you talk about growth, what are the key KPIs that you’re looking for? What sort of metrics do you define as growth?

Lomit: That is a really good question. When it comes to true growth teams, you can’t really work with vanity metrics, right? For us, the team metrics were always optimizing towards our ROAS and cost to acquire a customer.

Esther: Just to give us a little bit of sense of what kind of scale we’re looking at, can you give a ballpark range on the MAUs, DAUs, something like that, number of installs just to give us an idea of what we’re looking at?

Lomit: IMVU has been around 16 years and we’ve had probably north of 250 million registered users. Currently, in terms of MAUs, we’re probably north of seven million a month right now. In terms of installs, I would say we don’t really focus on installs that maybe some folks do. We’re always optimizing towards downstream events in, primarily, events being new pairs and revenue. In general, we’re getting– in terms of budget, we’re spending– Depending on seasonality, it could be millions or several million a month.

Esther: Okay. Perfect. I want to jump right in. You are, I’d say, quite a recognized expert when it comes to AI and growth through AI. I’d love if you could talk us through what that means a little bit, how this strategy actually works for you, how you utilize it day-to-day.

Lomit: For us, the exciting thing about working growth is that the company is really looking to us to drive growth forward. The way we define growth is my team is responsible for the entire customer journey. It’s from user acquisition to retention to monetization. Instead of just focusing on any given part of that funnel, our responsibility encompasses the entire user funnel, which is great because then we’re always focused on ensuring we’re driving the right quality of users that are ultimately going to lead to our best lifetime value customers.

Esther: That’s such an important point. I think you see it a lot in companies where you silo off different departments. You have one person who’s responsible for bringing in the traffic and they’re measured on typical UA KPIs. You have somebody else who’s working on optimizing for the endgame. Every step along the way, we have– If you’re looking for different KPIs and you’re focused on different things, you don’t have that communication to say, “Hey, great. You brought in incredible users through this campaign, but none of them are doing anything for us within the app.” Yes, amazing that you guys are looking at it in a more holistic way.

Lomit: I would add what you said is really important because I’ve seen the benefits and the downsides of just having different teams focusing on different areas. I think, ultimately, when it comes to growth, it’s really good to have one team that’s actually responsible for the entire metrics. That doesn’t mean that the entire team is going to be able to execute across the board, but they have to become a Sherpa who’s basically managing on prioritizing how the other teams end up supporting them to try and achieve those metrics.

Esther: Can you talk us through how it looks when you have– What does it mean practically, this cross down the funnel, full optimization?

Lomit: What it means is that the conversations could be a lot easier or harder, depending on how your numbers are going. [chuckles] For us, what it means is, ultimately, we have a pretty big say when it comes to having a seat at the table. Every year when the business is setting goals as far as, “Hey, what’s going to be our revenue target? How much are we going to grow? How does the product roadmap need to be influenced to support that growth?” My team has a pretty good say, not only on the strategic side of how the business needs to be going but also in terms of on the execution and tactical side. In terms of how the other teams ultimately end up setting goals that ladder up to supporting the overall growth goal, which I think is really important.

Esther: Is there a specific strategy that you are specifically proud of that you and your team have put in place?

Lomit: What I would say, there’s obviously different strategies that work for different businesses when it comes to growth. The best strategy is always a balance of user acquisition and retention. You don’t really want to just focus on one and neglect the other part of that. What I’m proud of is the fact that we’ve always had– and I guess it really comes down to how we’ve measured success and responsibilities that we encompass, which is the entire user journey, which means that we get to focus on both of those sides.

The way we’ve been able to do that, that has been different from the way I’ve previously done it. Other companies could have managed growth for over 20 years at a number of different startups. Over here, I’ve been here coming up to about four years now. One of the things that excited me about joining IMVU was I joined them at a point where growth was actually going in the wrong direction because it used to be like a desktop app.

When I joined, that’s when they were moving into mobile, but they never really spent much money. For the most part, there’s a lot of skeptics about whether mobile is going to work or not. The good and the bad side of that was that I was coming into a situation where nobody really believed that mobile was going to really save the business. At the same time, what I knew was that IMVU had a lot of great user data.

We get a lot of user data. In mobile, we don’t get a lot of time. People’s user attention is somewhat shorter, so you have to be able to react in real-time. The way that they were doing user acquisition before was the typical way where you have a bunch of analysts, data scientists. You can do spreadsheets, download data, look at what’s going on, and then go back and do these manual changes across different ad exchanges and partners.

Long story short, they wouldn’t react in real-time. The strategy that I ended up implementing was really helping to implement AI and automation into how that role would fit into growth for us. It was all about being really leveraging the digital tech platforms that were potentially out there and figuring out how we could put a couple of platforms together and make AI work for us.

Example being, AppsFlyer is a mobile measurement partner. For the most part, IMVU had a culture of trying to build everything. I said, “No. We don’t want to build our own attribution system for mobile.” We brought AppsFlyer for that. The other part of that was we worked with Leanplum to do a lot of automation and CRM, and then we ended up creating our own customer data platform where we were able to integrate all of our data, cross-mobile, and desktop in one place.

We have a unique identifier, so it’s easier for us to track the cross-platform usage with unique customer IDs and unique email addresses. Now that we were getting this data, the way we ended up growing mobile was because I had relationships with a lot of these partners like Google and Facebook. Actually, we’re able to work really early with Snapchat before anybody was really advertising.

We were one of the beta partners to help them figure out how advertising work four years ago. People talk about TikTok now. We were working with TikTok over two years ago, helping them to figure out how to better monetize all these users. I would say one of our secret powers has been to partner with companies in their alpha and beta stages, where they’re trying to figure out different ad models like Google UAC. Everybody uses it now, but we were using it way back.

The long story short was I’ve always been a big proponent of artificial intelligence. I had the ability to really implement it here at IMVU. The way I was able to implement it was to really learn how these other AI systems work across different partners, is by being part of their alphas and betas with Facebook and with Google and with all of these different ad networks or DSPs that had AI implemented in some way, shape, or form.

I really came to really understand, what were the data events that really mattered? How would they really leveraging AI or machine learning to try and optimize their algorithms to try and drive scale? What I came to realize was, ultimately– and this is good because every business has to do what’s best for them. Their algorithms were set up to try and help advertisers be successful in their platform, but it was done in silos because they had never had a holistic view into how they really compared to Google.

They never had an idea how to compare it to Facebook. It had an idea how Google compared to other campaigns we were running in Google, but it never had the holistic view. What we ended up doing was to try and build a layer that really sat between us and all these different partners who are spending money. With that AI intelligent machine that we ended up leveraging was ultimately using real-time data signals, which is the same signals all these other machines work on.

Instead of looking at it in silos, it was looking at it holistically. It was basically making decisions in real-time around bids, budgets, and creatives to really figure out how much we should be paying at any given time across all these different ad partners because every ad partner is an ad exchange at the end of the day. The way I got that inspiration was really studying the finance industry and looking at how trading desks work because, ultimately, they have to buy and sell.

If you’re a trader, you can’t just make money when the market’s going up. You got to be able to buy and sell and make monies on the margins. That’s the way I see how sophisticated paid user acquisition folks work. The only difference been is that you were kind of given fixed budgets to different partners when you could have more fluidity by having it more flexible. Because at any given time, certain partners and exchanges are going to be more efficient just based on supply and demand.

That, I would say, was the secret sauce there for our strategy, which was to be able to be agile in terms of how we were allocating budgets. We were optimizing ultimately towards a lifetime value user. We were getting all of this data. On that data, we were making predictions on how much any given user from any given campaign was really worth for us based on where they were coming from.

Esther: That’s amazing. I think, first of all, I love the finance analogy because it’s so true. You set specific budgets for Facebook and you set specific budgets even on a campaign level. Forget even on a platform level and it’s very hard to react. Nobody reacts to that real-time. You can’t do it on a human level. It’s just too difficult. That’s awesome. I’m curious when you switch over to the role of AI, obviously, the role of somebody who’s in charge of user acquisition and somebody who’s traditionally been involved in growth anyway, their position has to inevitably change. What does that look like when you’re not actively managing the bids and actively reviewing your campaigns in that way?

Lomit: The good and bad is I’m not sure if other people have run into this problem, but living in San Francisco, especially when the economy was great, obviously, it’s a little challenging right now, it’s really hard to hold on to folks that are really working in user acquisition. Ironically, most of my best user acquisition folks were getting poached by Google and Facebook. [chuckles] There was a little trade on how did it work. The good and bad part is people at that level are generally going to move jobs for like $10,000, give and take. Even though they don’t realize, half of that ends up going back into taxes anyway. [laughs]

Esther: [chuckles] Yes. The secret of the raise, it only gives you a little bit more. [chuckles]

Lomit: That’s right. Basically, one of the things when I started, I had a much bigger team. What I ended up doing was once I started transitioning to AI, I started bringing in the right types of people that would really be able to embrace this, not fight this because that’s half the battle is. You got to bring people in because for this technology to work, I had to bring in people that were able to think strategically and were able to not want to be getting in and doing all the manual tasks and processes around analyzing data and wanting to go and change and turn the dials.

I wanted people to set the machine off to do a lot of that work. In order to do that, we had to make sure, for example, one person has to be really smarter in data because, ultimately, you need someone to look at the algorithms and make sure there’s no biases or anything like that happening and that we’re not over-indexing or we’re so obsessed on optimizing towards an outcome where we ended up driving users that look good in the short-term but not good in the long-term.

I tried to hire one personnel who was really strong on the data side. I hired one person who’s really strong on the creative side because we ultimately ended up building a creative team in-house because creative testing became a big facet of how a machine works. Because part of it isn’t just finding the right users at the right price. It’s also getting much smarter on personalization in terms of, what’s the right message that would stand out for us versus our competitors?

A lot of that really comes from increasing the iterations or creative testing. To give you an example, any given month, we’re probably testing around 5,000 to 10,000 different variations of creative. A lot of that really comes from creating videos in static. With static, generally, we create templates where we have feeds that are changing different call-to-action, headlines, and images in real-time based on who our AI is trying to target.

With videos, it’s a little bit harder to change in videos. We’re changing different themes around different videos and causing that data back to our creative team to really use that data to come up with new iterations of creatives that we continue to keep feeding into the machine, and then ultimately having people that are strong on the soft skills like communications, leadership, relation-building.

A big part of our people, my team, excel with the machine part is really reaching out to different partners. For example, Google and Facebook, and ensuring we have those relations in place so we can continue to get into their different beta programs that are out there, as well as identifying new channels for us to be testing. Internally, we have to continue to champion what we’re doing so that other teams continue to support us on what we need.

A big part of that is making changes in the product, making sure that our data doesn’t break. Because as you know, every time there’s SDK updates that happen with partners, that can generally happen if you don’t overlook that, so ensuring that there’s really good checks and balances along the whole process and to all the different systems that go into supporting our AI intelligent machine.

Esther: Yes, that’s amazing. Also, you touched on a point that I’m super interested in, which is you do have these algorithm changes or platform updates or what have you that some may be a little bit more minor, some more major. Obviously, I’m thinking about Apple’s announcement about the IDFA. I would imagine it has a pretty extreme effect on everyone. Have you thought at all about how do we accommodate that? How do you have an AI at such a scale that you guys have it at, not fall prey to these kinds of changes?

Lomit: What Apple announced shouldn’t have really been a surprise to people primarily because they already have been talking a lot about user privacy for a while. That’s been part of a position that they’ve taken and they’ve introduced limited ad tracking. I believe it was probably September 2019, but it was about a year ago. I can tell you what we did on our end because one of the things I tried to do is always plan 12 to 18 months ahead and then try to anticipate worst-case scenario and plan for that.

A good example is I, like most people, never saw this COVID-19 thing coming. This has been a crazy disaster and, unfortunately, impacted a lot of people in a lot of negative ways. What I did plan for was two years ago that there was a recession that was going to come. Ultimately, the stock market– I’ve been around long enough to live through a couple of recessions that I know what goes up ultimately has to come down and correct itself.

That’s where our AI really helped us because back in March when a lot of shelter-in-place started to take place, because we have a machine that does a lot of these is automated in terms of how it allocates and optimizes our campaigns. When the costs started to come down naturally with a lot of advertisers pulling out, our machine just started to scale up our spend because, for the most part, we have an open budget as long as we’re hitting our ROAS and our cost to acquire customer goals. That was one example where, humanly, we might have taken a different approach and step back and maybe lost the opportunity that we were able to spend 60%, 70% more budget during the last-

Esther: Oh, wow.

Lomit: – in Q2. What happened was, ultimately, we were front-loading up our second-half-of-the-year budget into Q2 because things were so much more efficient for us to acquire users.

Esther: Competition went down so significantly?

Lomit: Yes. The other thing that we did preparing for Apple, we didn’t know when they were going to make this announcement, but we wanted to make Android to become an important business for us as well. Generally, iOS, obviously, has the best quality users for us. We always try to prioritize feature updates on iOS versus Android. There was a bit of a gap between feature parity between those two platforms.

In the last 12 months, we hired more people on the Android side to try and close the gap on feature parity primarily because our Android app at the time was lacking a lot of the retention and the monetization options that our iOS had. We had that in place and that’s really helped us to continue to spend more on Android. As a result of what’s happening right now, one thing that’s going to happen that we know because our machine is so highly dependent on ROAS data. The benefit right now is, with iOS, it’s all based on deterministic data because with an IDFA, you can really identify who’s coming from where and how much they’re spending.

What we’ve started to do now is work on more probabilistic models based on AppsFlyer data, as well as work on web-to-app data because we have cross-platform users. The first thing we’re starting to work on is trying to build some more media mix models so that we can leverage some more indirect data signals to try to figure out how to attribute for iOS. Yes, iOS is going to get challenged. If we don’t get that real-time data, it’s going to have challenges. One way to mitigate that is our machine is naturally going to spend more on Android because I don’t anticipate Google making any changes this year anyway.

Esther: Not yet. Not for you. [laughs]

Lomit: Yes, but what I will say the big difference between Google and Apple is that Google’s debt business is highly dependent on advertising, right? That’s the core– Well, Apple’s business isn’t dependent on advertising, so they can take a different approach to how they want to be positioned in the industry. Apple’s business is based on service. Services are becoming a big part of their revenue.

If they completely cut the legs of advertisers to grow their apps, that’s going to indirectly impact their share price down the road. I’m a proud Apple shareholder. I know, ultimately, companies need to find a balance of, what’s the right thing to do, but what’s right for the business? I feel if people haven’t done this, but more people are going to start focusing on Android, or at least try to figure out how to monetize users better on Android.

Beyond that, try to figure out with other data signals, and let’s be more probabilistic and deterministic to figure out how that can come into play. One other option I know a lot of people are talking about is the SDK ad network. Between you and me, at least from what I understand on that, I don’t think that would really work well for us from an AI perspective primarily because we rely on real-time data signals and that’s not real-time.

The second part of that is I think it’s limited to maybe 100 campaigns and we’re wanting tens of thousands of campaigns. Do we need to cherry-pick which of the hundred campaigns that we want to– I feel, ultimately, it’s going to be– it’s like any puzzle, which is what I enjoy about our industry. There’s never a dull moment. There’s always some curveballs that are being thrown at us to keep us on our toes. Right now, it’s a matter of taking a couple of different pieces and trying to figure out before that announcement happens.

Even when that announcement comes into play near the end of September, I think it’s generally still going to take a couple of months before more people are going to upgrade to that. It will give us enough time to really figure out the right weight and the right model to attribute for it, but it’s not going to be completely accurate. That’s where AI is really helpful because it helps you ultimately refine your algorithms to work on data where it’s not going to be like 100% accurate, but you need to figure out, what risk are you willing to live with, like 60%, 70% accuracy to make those decisions?

Esther: Yes. I feel like in our industry, there’s always going to be some level of inaccuracy. Even ROAS as a metric, it’s such a hard metric to figure out how to calculate because you can’t look at a user’s entire lifespan and take that and call it. You have to make some level of sacrifice. It sounds like you’re ready and ready for Apple to bring it on and you’ve got your steps planned out, so that’s great. I have one last question about your AI and the strategy at IMVU, which is, let’s say you could flashback four years. Start doing it all over again. You’re faced with that same challenge. Growth is going in the opposite direction. You’re trying to move to mobile. Is there anything you’d do differently?

Lomit: I would say, for the most part, what we’ve done has worked out well for us. What I would do differently is instead of creating that huge gap between iOS and Android, because our apps ended up becoming– the feature difference was pretty significant at any given point because we were just focused on one app at the expense of the other. I really don’t want to be in a situation where that happens.

We’re fortunate that we started to prioritize Android early enough, but imagine folks that haven’t right now. That’s going to cause a sea shift in terms of how they allocate and hire resources to try and get their Android app. I think the key thing is- and we’ve done this pretty well, I would say, for the most part- is always try to have a diverse approach to whatever you’re doing.

Don’t be relying on any given channel, any given partner, or any given employee because, ultimately, those things can shift or change at any given time. If you’re so highly dependent on that one wearable, then it can put you into a tailspin. Try to be broadly diverse as much as possible. Don’t be highly dependent on just working on Google or Facebook. Try to diversify your user acquisition channels.

Don’t try to just focus on UA. Try to focus on retention because, ultimately, it’s not about how many you bring in. It’s about how many you keep. The third part, and I know more people are doing this, is identify the right key metrics that really drive long-term success. Registration sounds nice. Millions of installs could be cool. Ultimately, what does that mean for revenue and what does it mean for paying customers?

Esther: 100%, very well said. Okay. I’m going to finish you off now with our quick-fire round, quick questions that I ask everyone. The first one, you’re ready?

Lomit: I’m ready as long as I don’t get shot down if I get it wrong.

[laughter]

Esther: It’s not a pop quiz, but we’ll see what happens.

Lomit: Okay. 

Esther: If you could give just one tip to an aspiring growth marketer, what would it be?

Lomit: I would say know your numbers.

Esther: Your favorite mobile growth resource?

Lomit: I would say, obviously, I love reading about different ad tech solutions that are out there. The other thing is just continue to read as much as possible in the industry because things are changing so much. I would encourage people to listen to this podcast, but listen to different podcasts that are out there because it’s really funny. You could listen to the same people talking about the same thing with 20 different opinions. Ultimately, the broader knowledge you get on any given topic by reading around it and studying around it, the more you’re able to develop your own point of view on it.

Esther: Who is the person in Mobile Growth that you would most want to take out for lunch and why?

Lomit: I would say I love to take everyone out. I love going to conferences. Right now, the thing I miss the most is just being stuck at home for the most part. If anybody wants to go to lunch with me, I would love to do that.

[laughter]

Esther: It’s an open invitation.

[crosstalk]

[laughter]

Lomit: It’s an open invitation, yes. I don’t know about you, but I’m tired of these Zoom lunches, right?

Esther: [laughs] They don’t compare. They really don’t compare. [laughs] Okay. It’s on topic and it’s our most important question. What is your favorite flavor of pancake?

Lomit: I’ll be honest. I’m not a big pancake– but I will say that I have kids, so I do know about pancakes. It’s normally like chocolate chip.

Esther: Okay. Amazing. Thank you so, so much. That was awesome. For people who want to find out more about you, your thoughts, learn more, where can they find you?

Lomit: I’m very active on LinkedIn. Anybody who reaches out to me, I’m happy to connect with folks. I’m always posting a lot of stuff about what I’m reading and what I’m learning in the industry. On LinkedIn, Lomit Patel. I also have a blog where I write more articles. My blog is really easy too. It’s pretty much my name, lomitpatel.com. If anybody’s interested in the book that I published recently, that’s available on Amazon, Lean AI. Definitely check that out. For me, I just love being part of this industry because things are always moving and changing. The one thing that’s constant is just the relationships that you build, right? You always want to continue to nurture and give back. I wouldn’t be here without other people giving me time to learn and grow as well.

Esther: Lomit, thank you so much. That was incredible. I learned a lot. [laughs]

Lomit: Thanks for having me, Esther. I’m really excited to be one of the early ones on the show and I know it’s going to be extremely popular. I’ll be listening in and cheering you on.

Esther: Thank you. Thank you so much. All right. Stay safe.

Lomit: You too.

About Esther Shatz
Esther is a UX nerd with a natural aversion to sunrise, morning people, and birds. After consulting on website UX for clients like Disney, Walmart, and Skype, she moved into the world of mobile and now obsessively searches for new App Store tricks to share with her clients and random passerby. Hobbies include coffee, hula-hooping and making soup.