🚀4000%+ engagement lift from early tests – the superpowers of in-app messaging – with Andy Carvell(Co-Founder at Phiture)😎

Our guest today on the Mobile User Acquisition Show is Andy Carvell, the co-founder and partner at the Berlin based mobile growth consultancy Phiture. Before founding Phiture, Andy led the retention team at Soundcloud, where he helped drive some massive experimentation-driven growth. He is also known for his Mobile Growth Stack, a framework that I’ve found impressive in both its comprehensiveness and simplicity.

In today’s conversation we talk about in-app messages, which can sometimes be underappreciated as a retention strategy especially compared to PNs and emails. Andy outlines why they work, when they work – and talks about the tremendous impact they can have. He gives the example of a 4000%+ uplift in engagement that they helped drive even when they were untargeted. I’m very excited for an in-depth view into an underrated but very powerful strategy that this interview with Andy provides.

KEY HIGHLIGHTS

💬 What are in-app messages?

⚡️ The typical use cases for in-app messages – and how they are a part of a multi-channel strategy for Phiture.

💯 The 4000%+ engagement lift that Phiture saw when they tested completely untargeted in-app messages for promoted books for Blinkist.

💪 Why in-app messages are powerful – and how to measure their impact – both positive and negative.

😰 How over time in-app messages can get fatigued.

💰How to think about users’ propensity to purchase or lapse – and to think about predictive messaging.

💡 Some of the more surprising things that Andy has seen with in-app messaging.

🏁 For an app that has never tested in-app messaging, where Andy recommends beginning – and what he sees as low hanging fruit.

KEY QUOTES

The immediacy of in-app messages

With in-app messages, not only are you only able to reach those active users, but there’s also another dimension to it which is how these are actually triggered, typically you would trigger them on a certain event.

That event could be app open or it could be that you pop up or display this message right after a user has done a certain action or when they land on a certain screen.

How to quickly improve conversion rates

You can really iterate very quickly on your conversion messaging, your value prop, the benefits that you are pushing, and get a very measurable and quick feedback loop in terms of what’s working, and really improve your conversion rates.

The power of attention

I think you have to think about what an in-app message really is at heart. As I said, it’s a small interaction with the user. It’s actually overriding the user experience while the user is literally got the phone out and they are looking at the screen, so you already got their attention. These other channels, of course, notifications and email, you are hoping to somehow get their attention at some point. There’s a sound effect; hopefully they check out their messages, or maybe they are going to read through their email inbox later, but these notification channels, they are about trying to redirect attention. With in-app messaging, you already have the user’s attention. With that great power comes great responsibility of course.

Maximize the opportunity

When you get somebody’s attention, you could then direct them to another part of the app; you can deliver them a very segmented, very personalized message; but at the end of the day you are getting up in their face and you are showing them something, so it better be valuable.

When to deploy reengagement messaging

If we contact them three days later, they are less likely to come back than if we contact them after one day. We contact them five days later, they are even less likely and that will always be the case, where you see that drop off, but what we wanted to look at was: what is the right frequency to try reactivation or reengagement messaging?

In-app messaging is easy to set up

And then in terms of low-hanging fruit, where to start testing your first messages, I really advise working on the onboarding flow, you can do this thing called adaptive onboarding which I’ve already alluded to where you just target, figure out what are your key features, your key experience in the app, things that you would want any new user to go and experience, that you see are tied to better activation rates.

FULL TRANSCRIPT BELOW

Shamanth: I’m very excited to welcome Andy Carvell, cofounder at Phiture to the Mobile User Acquisition Show. Andy, welcome to the show.

Andy: Thank you very much for having me on the show.

Shamanth: Absolutely. I am excited to have you also because I feel like our paths have crossed virtually so often – and certainly your writings are very much held in very high regard not just by myself but so many people across the mobile space starting from your Mobile Growth Stack to everything else you’ve written since then, this is something I am very excited for today.

Andy: That’s very kind of you to say so, thanks for the kind words. It’s always great to know that people are getting something out of the stuff that we publish.

Shamanth: Absolutely. And today we are going to talk about in-app messages and some of the wins you guys have unlocked using in-app messages. To get started, can you tell us what in-app messages are?

Andy: Yeah, absolutely. I am super excited to talk about this topic. I’ve been talking about in-app messages most chances I get lately simply because I really believe they are the golden channel for mobile right now. So exactly what is an in-app message? It’s essentially a small user interaction unit that’s delivered from a server side platform, typically on mobile engagement platform, such as Braze, Leanplum or CleverTap.

And the way it looks and the way it behaves is really down to the marketer that’s deploying it, and the capabilities of the delivery platform itself. Some platforms allow more flexibility with the formats, but typically we are talking about popup or a modal window that pops up and displays a message. It could have an image, typically some text, and one or more buttons. It could be a full screen interstitial, could be something that just slides up like a little slide up or pop over.

Yeah, many different kinds of formats really, but the common denominator is it’s something that’s essentially overriding the standard interface of the app. And you can make them look however you want, particularly with some of the platforms do allow you to build out essentially what’s a small HTML5 WebView, and then you could include CSS and JavaScript and really make them look completely native actually.

Shamanth: Indeed. So for a lay person, would it be fair to say these are the mobile equivalents of popups?

Andy: I think you can, you can simplify to that. I think that’s the bit that’s probably easiest for people to understand. They are segmented triggered popup messages.

Shamanth: Yes, got you. When you think about ways in which you can target users of an app, users who’ve already started using an app, via lifecycle campaigns. In-app messages will typically impact users who are already using an app as opposed to, let’s say, PNs and emails that you can use to target people who may not be actually inside an app.

So I imagine, that’s an essential difference between how in-app messages work versus PNs and emails which are more well-known, more mainstream we could say. So with that context, what are some of the typical use cases or strategic goals for which you would recommend using in-app messages compared to other forms of retention strategies like PNs and emails?

Andy: Yeah, again, really good question. It’s really important to acknowledge the different dynamics of these different channels that you can use when building out a retention strategy or a CRM strategy. So just like push notifications and emails, these things are typically deployed through an engagement platform or marketing automation platform, which means that you get proper measurement of impact around what you are doing.

The server side deployment aspect allows you to iterate quickly on new ideas and deploy things without an app store release. And you could target, either all users, but you are right, absolutely, just the only ones that we are going to see are inside the app, that’s a key difference, versus push an email which are channels which you can use to bring people back into the app.

And you can segment down to an individual user level. And one of the key differences, before we get into the use cases, would be

with in-app messages not only are you only able to reach those active users, but there’s also another dimension to it which is how these are actually triggered, typically you would trigger them on a certain event.

That event could be app open or it could be that you pop up or display this message right after a user has done a certain action or when they land on a certain screen.

So you can show them in a very contextual moment at a key part of this journey. So yeah, back to your question, you are right, you can only reach those active users, although what you are able to do is chain them.

One of the approaches we use a lot at Phiture is, to use a multichannel approach, so we might use a push notification or an email to, to get the user’s attention, bring them back into the app, and then the segmentation would work basically on popping up that message, when they open the app based on having them just click on that push or email campaign. So you can chain these campaigns together and leverage the strengths of the different channels.

Then, in terms of the actual use cases, the things that we’ve seen working really well, I mean, they are pretty numerous. But, let’s say, some of the ways that you can really leverage in-app messages that almost always works great: onboarding, you can really supplement your existing product onboarding by segmenting users who haven’t got to activation point or haven’t experienced or tried out certain features by popping up a message and actually taking them straight to the feature, educating them about that feature, for the users who haven’t tried it.

Upsell, I think is the key one, for any monetization efforts, like, conversion to subscription. Everyone’s trying to do subscription apps these days.

Content recommendation, I have an example from Blinkist. We worked with Blinkist recently to help them implement some in-app messages to do content recommendation, when they were recommending a promoted book that week which wasn’t even segmented actually, it’s going to be a quick test to go to the whole user base, just to see how valuable that would be to be able to push one piece of content of like a promoted book that week, and there was a 4000% uplift in users who read that particular book after receiving that message. That I think is, yes, super impressive, but also, you have to understand the dynamics of the messaging channel we are basically getting right in everyone’s face with a popup that says, “hey, check out this book,” so a lot of people are going to click it.

But especially when it’s new, there’s probably also a little bit of halo effect there. But I think what was also super interesting from that experiment was there was a 6.5% overall increase in engagement with any book. So that was a really interesting dynamic that we saw that by just pushing any content whatsoever we saw overall uplift, on aggregate uplift and engagement across the whole user base. So it seemed to just stimulate people to go and explore books in general even if they weren’t interested in the one that we promoted.

Finally, couple of other use cases, I could go on all day to be honest. But surveys, I think is a really big one, particularly since, if you are building slightly more advanced in-app messages, and you are using JavaScript for example, you can ping results of surveys straight back into your CRM system, which means that you are enriching that user profile with a response to a question, and you are doing that in real time. So you can immediately start segmenting it.

So a particular use case which use a lot at Phiture would be that we deploy, an intent collection survey: typically, it’s always a good idea to learn more about why your users are there, what are they hoping to get out of the app, and you can ask them that, a simple multichoice question on the very first section of the app. We can store that information and then immediately start customizing them into, for example, different onboarding tracks, based on what they are there to do or, if it’s a content app, then what kind of genres of content are they interested in. Some of that stuff you might eventually build natively into your onboarding product, but we find that this is a really quick way to test, and rapidly prototype advanced onboarding flows that are super personalized.

Shamanth: Yeah. Those were some very impressive numbers especially what you described with what you guys did with Blinkist, I am wondering if that sort of uplift is perhaps because of an in-app message is just on a very virgin channel, right, I mean, if you are looking at PNs, I have 10 different PNs on my phone. But if I am in the app, as you said, it’s very in your face, it’s hard to miss this, just no competition for it. Do you think that explains why it’s so effective?

Andy:

Shamanth: Yeah.

Andy: This is why designers often hate the in-app messages. Usually there’s tension in the team; this is usually between the product or the design team, because the growth marketing team that sees the value of this channel.

And, of course, you can measure the impact of that, so you can iterate towards higher value, higher conversion rate. We routinely see insane conversion from in-app messages just exactly because of that.

And secondly, yeah, you have that responsibility to measure the results carefully and not just looking at the results you want to see, how many people are clicking on the button and how many people are doing the action that you are trying to drive them to, but also the negative effects like are people uninstalling the app? Are people churning out of the app after having received these things?

Shamanth: Indeed, indeed. And, as you said, great power, great responsibility. And you touched in this conversation earlier on some of the segmentation that’s so crucial to making in-app messaging work, but even though it sounds like that your win with Blinkist was just even without any segmentation, you did say…

Andy: Yeah. It was a basic first test and still the impact was huge.

Shamanth: Segmentation is so crucial. I would imagine, part of what you want to achieve via segmentation is understanding the probability of, let’s just say, somebody lapsing or, and just reacting to it accordingly or somebody who is on the fence about making a purchase and showing them a tailored pop up, in-app messages, right?

Andy: Yeah.

Shamanth: How do you think about this sort of predictive messaging?

Andy: I think it’s the next level, right, it’s where everyone is trying to get to. There’s a lot of hype around machine learning, churn prediction models and, propensity scoring in general. It’s also a topic which I really love, and I’ve worked on quite a bit even in my days at SoundCloud, we were working on various propensity scoring models and pre-calculation stuff. I think it’s really fertile ground for that next level of personalization, and personally I’ve been wary about anybody who claims to have built this on a generic level because I’d love to see it.

I’d love to see something where you can basically just plug in a few variables, and it generates your propensity score, for the user to do X or Y. I’ve seen a lot of products claiming that they could do this with the vague references to AI and machine learning but I haven’t seen this working in this generic way, and also I’ve seen a lot of good companies with smart engineers and data scientists struggle to build propensity or prediction models, which are much more accurate than just flipping a coin. These are hard problems to solve. I am not saying that it’s not possible to get there, but yeah, it’s tricky.

So yeah, back to your question: I think you raised a couple of specific ones there, so propensity to churn, propensity to lapse, right? So I am all for being pragmatic: we got pretty far at SoundCloud when I was there, and running a retention team.

Just with a fairly basic model which wasn’t AI prediction or anything like this, we just looked at the whole user base and we mapped out: if we would contact people and try and get them back into the app, how many days should they be inactive?

If we contact them three days later, they are less likely to come back than if we contact them after one day. We contact them five days later, they are even less likely and that will always be the case, where you see that drop off, but what we wanted to look at was: what is the right frequency to try reactivation or reengagement messaging?

Is there a point where there is a step-change where, if you leave it later, that a lot of people just don’t come back. At SoundCloud we saw about five days was a pretty good point, just on average, right, but that’s not a propensity scoring system, that’s just how we figured out how to do a very generic interaction timing.

Now, you can do scoring on that stuff. Yeah, I think you can really do it down to a user-specific level, right, because ideally, you want to be looking at the natural usage frequency of specific users, maybe probably you can cluster them into a few different segments, if you don’t want to do user level segmentation, but you want to look at that user’s history, maybe they’ve been around for a while, or maybe they only come once or twice a week, and maybe that user, they just active on weekends for example, maybe they are working really hard during the week, maybe there’s another user who’s unemployed and they are on it,, almost 24 hours a day. And you don’t want to necessarily treat those two users the same, so something that actually looks at that history and then tries to project that forward is, that’s how these models typically work. But yeah, and then you can come to a better understanding of when is the right time to interact. But it’s hard to do that modeling in a way which works well.

And then, with purchasing, I think, there’s another way you can look at that. Again, without going into a super deep machine learning approach. At Phiture we built an intent model: we are looking for any signs of intent to purchase, which you can break down into a few different levels of intent. Have they ever clicked on a feature which is behind the paywall: have they investigated things which are actually paid features, just in their course of using the app? And if they have done that, they’ve at least shown some interest in stuff that you are charging for. If they started a free trial and then dropped out, that could be a good indication that there’s a high propensity to purchase, you just haven’t got them across the line yet.

So I’d argue, you can get pretty far with that logic, before going into a deep machine learning model. Not that I’ve got anything against machine learning. We are also building ML models here at Phiture, so I know from personal experience they are hard and I’ve also seen a lot of companies take a lot of time into them and not get an awful lot of value out of that.

Shamanth: I am curious, what have been some of the most surprising things you’ve seen when you tested in-app messaging.

Andy: So with in-app messages, specifically, I think we’ve already touched on the crazy impact that we see it’s really been pretty impressive. I remember even from SoundCloud days, and we see quality conversion anyway, for all the reasons I’ve just described, you can get in someone’s face while you’ve got their attention, you will get a lot of people taking the call to action, you will be able to drive people to, to a place this in the app that they may be never were before.

Shamanth: Yeah.

Andy: But, I guess, another thing which I’ve been consistently impressed or surprised by perhaps is how willing people are to answer questions, that’s a full-on survey with lots of questions in it, or just a mini-survey like the intent collection that I mentioned earlier where you might just ask them a single question with a multiple choice.

Shamanth: Yeah.

Andy: People actually really like being asked for their opinion about things. An app doesn’t often ask them for some input from themselves, it’s mostly they are just consuming content from the app, and people respond very well to it. I am from the mobile app development space, I used to, I used to make mobile video games that was my background, so I am very much familiar with the concept of a techie company or even the lone developer, sitting in a dark room, coding something away.

Content recommendation always works really well, yeah, I mentioned the Blinkist example. That was very untargeted, that was that first test into that, I think they are building out more targeted recommendations now. Yeah, finally, I think one thing which is maybe not surprising exactly, but something which you have to learn when you are using in-apps, this idea of having trigger points. It’s not just that you target the segment and then everyone in that segment receives it. You’ve got a segment target, so, let’s say, I am targeting, I don’t know, new users on their third day, right. So the only people eligible to receive that campaign are people who are seen in the app on their third day. But then I have a trigger point as well, so let’s say we trigger the message when they add something to the shopping basket: add-to-cart event. And so my reach is much lower on that message because not only am I seeing just users on the third day, but it’s actually a much, much smaller subset of those users.

Shamanth: Right.

Andy: Only those who hit the trigger point. And I think when you are doing, calculating, for example, how long it’s going to take for an experiment to reach significance or how many users are going to be exposed to a certain experimental treatment, it’s important to be able to bear that stuff in mind, because that’s very different from a push notification or an email that you are blasting to the entire segment.

Shamanth: Yeah. If your reach is so small, then there’s just no point, right, if it’s just so down funnel, yeah.

Andy: Yeah, and also, it’s unpredictable, because you don’t know how many people are going to trigger that event on a given day, and you might have historical data that should be able to give you an idea, but still, there’s an element of unpredictability about this.

Shamanth: And for an app that has never done in-app messaging before, where, where do you recommend they begin, what should they look for in terms of infrastructure or what would their lowest hanging fruit be for an app that’s never done this before?

Andy: Yeah, good point, so infrastructure, I really don’t advise trying to build it yourself.

If you are really on a budget, I think even Firebase supports in-app messaging now. If you are really on a budget and you are just starting out building a simple app, I think, it’s a good place to start. And it’s still a bunch of functionality which is quite helpful to developers.

And then in terms of low-hanging fruit, where to start testing your first messages, I really advise working on the onboarding flow, you can do this thing called adaptive onboarding which I’ve already alluded to where you just target, figure out what are your key features, your key experience in the app, things that you would want any new user to go and experience, that you see are tied to better activation rates.

Literally just target users who haven’t done those things and educate them about them. Ideally deep link them to those features or those screens with the CTA button. And if you just do that, you can use the built-in message templates which probably don’t look super native for you, but you can still probably get impact above and beyond what you would have without running those campaigns. And then, a next step would be to build custom templates. I am going to give a little plug now for us: at Phiture we’ve built an in-app message studio which basically allows people to build super native looking templates, we call it Blender, and that’s, that’s super helpful for folks who, who just don’t want to get into the nitty-gritty of coding HTML and CSS. But yeah, you can just build them yourself, you just need some frontend experience.

Shamanth: Indeed, and we will link to Blender in the show notes and the transcript for people to check out as well. Andy, this has been such a privilege to have you on the Mobile User Acquisition Show, I want to be respectful of your time. Where can people find out more about you?

Andy: Thanks Shamanth. Yeah, you can find me on LinkedIn, Andy Carvell. You can check out Phiture which is the mobile growth consultancy which I co founded which is www.phiture.com, P-H-I-T-U-R-E dot com. There’s information about Blender on there as well.

Shamanth: Wonderful. Thank you so much Andy for being on the show.

Andy: Thank you Shamanth, it’s been a pleasure.

A REQUEST BEFORE YOU GO

I have a very important favor to ask, which as those of you who know me know I don’t do often. If you get any pleasure or inspiration from this episode, could you PLEASE leave a review on your favorite podcasting platform – be it iTunes, Overcast, Spotify or wherever you get your podcast fix. This podcast is very much a labor of love – and each episode takes many many hours to put together. When you write a review, it will not only be a great deal of encouragement to us, but it will also support getting the word out about the Mobile User Acquisition Show.

Constructive criticism and suggestions for improvement are welcome, whether on podcasting platforms – or by email to shamanth at rocketshiphq.com. We read all reviews & I want to make this podcast better.

Thank you – and I look forward to seeing you with the next episode!

KEY HIGHLIGHTS

KEY QUOTES

The immediacy of in-app messages

How to quickly improve conversion rates

The power of attention

Maximize the opportunity

When to deploy reengagement messaging

In-app messaging is easy to set up

WANT TO SCALE PROFITABLY IN A GENERATIVE AI WORLD ?