How to spend your app retargeting budget: A webinar roundup with Remerge
Senior Content Manager
Nov 15, 2019
Retargeting is moving up the app marketing agenda. With user acquisition becoming more costly, app marketers are increasingly investing in app retargeting to improve their efficiency, and capitalize on the users most interested in their apps. But there are pitfalls where marketers can lose the plot, and our most recent webinar provides the answer to some of the biggest problems app retargeting can create.
Alexandre Pham, Adjust’s Director of Partnerships, EMEA, held an engaging conversation with Remerge’s Dr. Christian Liesegang, Chief Operating Officer & Co-Founder of Remerge, going into an in-depth discussion of current trends, such as user behaviors, retargeting waste, and how to tie it all together with Adjust and Remerge. Read on to learn the key points from the webinar, and you can also watch the full recording (or read it) further down.
Universal truths - Nowadays, “Universal Links are a must,” Christian says. It’s important to provide users with seamless user experience. That means, “no popup dialogues, no redirects to App Store.” Instead, your app links need to work hard, so your users don’t have to. Mastering Universal Linking is a complicated topic, and we outline differences between them and deep links in our guide here.
Click into place - There are a tremendous amount of new apps released every day. As these apps try to buy their way to success, fraudsters are attempting to take advantage, and steal from the advertising spend. As a result, “measuring post-click behavior is important to making sure that you optimize the campaign spend,” Christian adds. Focusing on the publishers that perform well is a second trick. Adjust’s Click Validation product helps you to filter out fraud by matching the click to impression data — and Remerge is one of the networks to help us achieve our vision of a more transparent ecosystem.
A captive audience - “We always want to have large segments — as much data as we possibly can. After all, we want to reach as many people as possible,” Christian says. However, in marketing, you always need to think about, "Okay, what is the right messaging for certain segments?" Adjust’s Audience Builder offers “a very good approach” for marketers to easily make segmentations work.
The full transcript
Alexandre - So let's start with the very easy question and if you could explain to us exactly what is app retargeting and how you deal with it at Remerge?
Christian - Sure. In a nutshell, retargeting basically takes place in the moment that a user is engaging with an app, for instance, in an e-commerce app looking at some products, shopping around, browsing, adding stuff to the shopping cart, which they would like to purchase at a later stage. Then, for some reason, they're actually leaving the app, so they stop interacting with the app. And this is actually where retargeting then kicks in. In that moment that a user leaves an app, retargeting takes over and we try to reengage the user with creatives advertisement, which are tailored and personalized to the demands of the particular user. In this example here, we have an ecommerce app, where user browsed at a certain product but never ever did purchase.
So retargeting actually is taking over. We are trying to find that user again in the ecosystem. And once we are able to detect that user and can show a creative to that user, we typically show personalized banners, which actually, for instance, could show the product that user showed interest in previously. And hopefully then the user is actually clicking on that creative. It leads them directly back into the app to that certain product, and makes them actually do the purchase of a particular product.
Okay, so this brings me to actually our first tip and trick and this is something that we hear obviously since iOS 9 and even more since iOS 10 as it became really part of the iOS ecosystem. So this is the universal link. I believe there are a lot of prerequisites when it comes to retargeting campaigns and specifically for iOS. So we will dig deeper into the universal link. And if you could tell us a bit more how you deal with it at Remerge, that’d be great.
Sure. In general, universal links at Remerge are hot topics since the beginning. They have been introduced a couple of years ago since iOS 9. They exist since iOS 10, they actually became mandatory to use for properly deep linking into an app. And what it does in a nutshell, it is a unified approach from Apple introduced it to the iOS ecosystem to have one link for a user to either provide a seamless experience to directly drive the user back into an app or if the app is not there for some reason to actually redirect into the mobile web space, for instance. So that we have, in the end, a guaranteed user experience. So whenever a user clicks on the link, it doesn't matter if that app is there, yes or no, it always leads to something for the user.
The interesting thing is from the flows that the app is actually detecting, the universal link is actually detecting if the app is still installed. If it's there, it directly leads you to a certain page within the app, which is shown on the upper right corner, or in case the app is not installed anymore the universal link has the capabilities to either fall back into a mobile solution, which is quite common in ecommerce and travel space. But also in case of gaming where you sometimes don't have a proper mobile solution available, you can also fall back into the app store and then try to reinstall it.
The good thing here is that all this is actually happening automatically. So a user doesn't need to do anything here. In that moment that the universal link is clicked, the redirection is done automatically and the app is opened seamlessly without any kind of user interaction needed. This in the end can actually lead to an increased app engagement with your clients by using the universal links.
Okay, so conclusion is, app developers who were running app retargeting before have to be implementing the universal link within their iOS app.
That is a very good point. Indeed, we are actually requesting this from each and every client to have universal links set up because that is the thing that has been vindicated by the iOS ecosystem. So there's no way to opt that out. And depending also, on topics that we're going to discuss after this year, we might see that it is needed to have a seamless app open experience for the user to also detect any kind of unintentional clicks from real clicks that are done by a user. What we see here on this graph is actually based on real time prospect data that we get from Adjust. So any client that is working with Adjust has the capabilities to share this prospect information with your retargeting partners so does Remerge.
And what we did here basically is we were looking at the number of successful sessions or app opens after a click and we also looked at the desired conversion events for the client. So, for instance, a purchase event. And what you can see is that we had kind of a flat curve on the left-hand side of the graph, and after we implemented the universal link or the client basically implemented the universal link, and we just set it up on our side. We could see that there was a tremendous increase in the reengagement rate, which in the end also leads to a higher conversion rate over time.
And what we typically then do explain this to a user is that it has a properly implemented universal link. It doesn't provide any possibility for a user who clicks on a link to actually opt out of that app open flow. So there is no popup dialog asking for confirmation. There's no redirect into the App Store. It is seamlessly leading into the app, giving the user a very good user experience and obviously also resides in an immediate session event which you can track as a successful app open afterwards.
Yes, I think this shows that having the ability to track the reengagement rate is very important, like making sure that the time between the click on the ad and the next session is fully taken into account during your optimization.
Yes. We are looking in particular into the app open rates and look also in the signals that we receive back from Adjust even the app open most successfully attributed to Remerge or not, by looking at so called reattributions and rejected reattributions. And what we can see here is...and that is one of the services that Remerge offers is that for our clients, we provide continuous link testing. That means whenever we get a tracking link, whenever we get a universal link, which should be used for retargeting, you're monitoring the performance in terms of reattributions and rejected reattributions for each client.
So what we can see here on the graph is basically the effect of implementing universal links, which lead to a better seamless experience for the user. And over time, you can see that the number of reattributions properly increase. Rejections go down, because actually we make sure that the app is seamlessly opening, which in the end leads to a very short click to app open time, which is key for detecting fraudulent behaviour. And so in the end, resides in the lower rejected reattribution ratio. So it means every click leads to a proper attributed session for the client, ultimately leading into better performance.
Rejected reattribution and overall fraud prevention is also something that we need to take into account when it comes to reengagement. I think a lot of app advertisers are aware that there is ad fraud ongoing on app installs. But this is also something that can happen within reengagement. So what we at Adjust look a lot at is specifically that time between the click on the ad and the next session from the user and understand if applying our distribution modelling, if this is part of a click span, fraud pattern or not.
So this is something that we put a lot of emphasis on. And it can also either detect fraud or simply also technical problems on the publisher side. So we can give this feedback to our partners for them to optimize as Remerge showed us there.
Another thing, another topic that we are trying to push on our side and with our partners is the click validation. So we started looking at, you know, on a daily basis the number of clicks and advertising ID that we're receiving. And there are simply too many clicks for the number of people and devices on Earth. We know that a lot of these clicks are fraudulent. And we want to fight that by pushing this click validation.
So simply, what we are asking our partners is to share impression data and link them to the clicks that they're generating for their customers. With this, it will be much more difficult for a fraudster to fake all the impressions that will lead to a click. It will just make it financially nonviable to perform fraud on Adjust clients. So this is what we want to bring. And we believe it's also going to bring much more transparency between our clients and partners so that in the end clients they have more trust in partners that are able to validate their clicks and prove that they actually displayed an impression before the clicks were generated.
Coming to checking these clicks, we want to also talk about this second tip and tricks, how can we make sure that we check our clicks and how this is done on Remerge side? Could you explain to us a bit more at how you're looking at reattribution signals and how do you apply this information into your algorithm?
Of course. So, first of all, Remerge does support the impression forwarding features that are required for click validation on Adjust. So that means Remerge is able to forward each and every impression to Adjust so as to actually do a proper click validation by verifying that the necessary impressions have happened in the right time frame before the click actually was tracked. That is one of the features.
The second feature that we're utilizing is again that we are monitoring currently all the app activity. So we are tracking for each click on an individual publisher basis for each retargeting campaign for each of our client individually, we are looking into the rejection ratio afterwards. So whenever Adjust basically gives us feedback that a certain click was rejected, we are currently tracking, okay, what was the responsible publisher there?
And then based on a decision made for each individual client depending on the technical setup, we draw automated conclusions to actually whitelist or blacklist a certain publisher. So what we apply at Remerge is a machine learning algorithm. The reason for that is that at the moment we observed that above roughly 6,000 new apps per day are popping up in the app stores. And we’ve been doing this manually checking and blacklisting and whitelisting publishers, it's almost impossible.
So the machine learning algorithm actually helps us here to track and collect the data that is sent back from Adjust and monitor the performance of each individual publisher. But you can see here on the graph on the left-hand side is actually how our machine learning algorithm starts to optimize after a certain amount of time. The baseline for that is actually that we need to collect as much data as possible. On average, it takes Remerge seven days to have a fully optimized algorithm set up for each individual publisher leading in the end to a lower spend on these publishers.
So if you notice that the publisher over time is not performing in terms of there's no proper app opens attributed to Remerge, there's no post-click conversions attributed to Remerge, that we start actually blacklisting and limiting the bidding behaviour on these publishers, ultimately completely stopping bidding of these publishers, which basically resides in a better distribution of the marketing budget by focusing on publishers that perform well and are not considered as fraudulent.
So if we on the one hand side, limiting the budget spend and focus on publishers that perform well, in the moment that we focus on publishers that do a good reattribution, we can also optimize the marketing spend by the client by focusing our bidding behaviour and bidding strategies on these kind of publishers.
Typically, as a side effect of blacklisting the fraudulent publishers, you can also see a positive impact on the good performing publishers, which is shown here on the graph. So at the same time frame, where we started blacklisting the fraudulent publishers, you can see that in overall using the same budget pacing as before, we generate a higher ratio of reattributions because we are focusing the money spent on these good performing publishers. This in the end, obviously also leads to a higher conversion rate for the client and it makes it easier for us to achieve the client KPIs.
Yes, I think that's a good point on that side to continuously optimize the publishers on which you buy your traffic from knowing that you guys buy everything through real time bidding, through exchanges. And this environment, even though you have full control on where this is displayed, you always have as an exchange more and more apps that are going to the pool and you as a buyer, and you are probably one of the biggest buyer for most of these exchange, you want to make sure that you're buying on the right placements.
So going to the last tip and trick, there is no reengagement campaigns without specifically targeting an audience. So I think what we could hear from you also is couple of example on how you experimented using specific audiences. So we're going to mention the Audience Builder. The Audience Builder is also an Adjust product, which allow our customers to segment their user data.
So they are tracking events with Adjust, they’re tracking several activities according to their user funnel. They will also want to create segments, buckets of users that have specific attributes. And what we allow is from these segments, customers are able to slice and dice the data with the attributes they want and also share these with the partners of their choice in order to create several segments and see how each of these segments behave after reengagement is performed. So we’re going to enter a case study that Remerge applied with a food delivery app.
Okay, thank you. Yes, we had a question a couple of months ago from one of our clients about, "Can we actually measure the performance of people based on the origin of the install?" So they went particular interested into a scenario to analyze, okay, how do people behave that have organically installed the app versus people that have recently uninstalled the app, for instance, or people that have not any interest anymore in actually using the app, so they have completely uninstalled the app. And obviously, we can do a lot of segmentation based on in-app events. In that particular case, we were relying on the Audience Builder setup because it was the easiest for the client to set that up.
Remerge has a good technical integration and automatically consuming this Audience Builder list for our targeting purposes. So the setup was pretty straightforward here. In the end, we had two audiences defined on the Adjust side, which were consumed by us. One audience was basically defining uninstallers, people that have recently de-installed the app or people that have reinstalled the app. And then we also compared this with users that organically installing the app.
And here we basically what we did is because we wanted to make sure that there is no bidding algorithm tampering with the bidding behaviour or evaluating the data in the wrong way. But we said, okay, we are mixing all these audiences into one retargeting campaign to make sure that the condition and the algorithm is behaving exactly the same for all these users.
And then internally, we could do a flagging of, okay, each user belong to a segment. And then we could use real time reporting that we offer to actually monitor the performance of these different segments. That was basically the setup. It was running for a couple of weeks. And we had some very good results on the different segments here.
So what you can see is basically is that we had in total for the test phase about 30,000 clicks. And we looked at the performance, okay, how many clicks were actually driven by people that have organically installed so have just the app currently on their phone. And we looked how many people actually have clicked that recently reinstalled the app. And we also looked at the number of clicks ever generated by people that have uninstalled the app.
And we see here at the moment that there's about 77% of the total ad spend for these clicks was actually spent by the installers campaigns, while the uninstaller took up 21% of the overall expense. So there was quite an amount of spent going into these kind of user segments that have recently uninstalled the app. The next thing that we actually looked at is what is happening after the clicks. So the click is one thing, but obviously, we also need to see are we able to bring these users back into the app. And here we could, again, see that especially the people that have uninstalled the app, even they might click on the app, it seems like they're not willing to afterwards reinstall the app.
So we saw only a percentage of 2% of people that have uninstalled the app coming back into the app. And this was also reflected then into post-click conversion. So it was not only interesting for the client to see how many people we can actually reengage and bring back to the app, no, they also wanted to see how many in-app purchases were actually driven afterwards. So in that particular case, food delivery, obviously, it was food ordering. So they also looked at, okay, what happens after the app opened?
What we could see and which is shown here on this table on the upper area is that we have a very high conversion rate of 1% for the installers and to that 5% for the re-installers. But obviously the uninstallers, they drove a lot of clicks, basically generated a lot of costs for our client. But in the end, the number of reengagements was extremely low. And even the conversion rate was the worst one of all the three segments that we actually compared.
I think that's a very interesting use case in this food delivery case because from a marketer perspective, when you are reengaging people who reinstall the app, I mean, I'm not so surprised that the conversion rate is, again, slightly higher, even though we could have more volume. Because in the end these re-installers, they uninstall the app, maybe for a reason, for seasonality, or whatsoever. But they reinstalled it knowing already the user experience that that app had, which probably explains also why you see such performance.
The example can be expanded at all sort of app verticals. And I think each marketer within their vertical could benefit from doing such test and understanding exactly how each of the segment from installers, re-installers, and uninstallers are performing afterwards and changing the messaging, changing the creative, and really digging to the why and push their users down the funnel.
Exactly. So one of the outcomes of this case study in the end was that we concluded together with the client that we should focus all our marketing activities actually on the organic installers and the re-installers and try to avoid targeting uninstallers because they’re actually eating up the budget and they're overall resulting into a negative impact on the overall client KPI. So that was one of the conclusions here that we said, "Okay, we are able to actually make it transparent. How is the performance of the different segments based on the audience builder list?" And this was in the end, leading to an optimized campaign setup where we agreed with a client, "Okay, it doesn't make much sense in the way how we do the marketing activities at the moment to target the uninstallers."
As you just said, in this particular case, there are two options. Either we completely ignore these users and don't target them at all, or really think together with the client about an alternative marketing strategy, how you reach these uninstallers. And that was one of the cases where we use the Audience Builder in a very simple fashion to get this resized pretty fast.
So in general, maybe to conclude that from a Remerge perspective is that talking about the three tips, we can take...definitely one of the things you should remember is that universal links is a must nowadays on iOS. It is important to provide a seamless user experience, no popup dialogues, no redirects to App Store so that the user actually has a chance to opt out. So that is already very good. Plus, it directly resonates with any kind of fraud detection utilized by Adjust, for instance, to make sure that every click leads to a legitimate session event afterwards. So both things work quite well in conjunction.
On the check your clicks topic, what we said here is also there is a tremendous amount of new apps popping up every day. So, of course, there's a lot of fraudulent traffic in there. And really measuring the post-click behaviour is important to actually make sure that you optimize the campaign spend and focusing really on the publishers that perform well. So here, again, choose a partner who's measuring what happens after the click, is detecting, "Okay, this is not a legitimate click." And that also can utilize this information to optimize your campaigns automatically.
For know your audience, of course, we always want to have as big segments as possible, we want to reach as many people as possible. But in general, in marketing, you always need to think about, "Okay, what is the right messaging for certain segments?" And Audience Builder list was here a very good approach for us to easily make segmentation, a client could easily follow how the segmentation works exactly. And we had a very clean setup on the campaign configuration to make sure, "Okay, this is where the uninstallers go. This is where the organic installers go." And then it was easy for us to measure and report back the performance, even in Adjust for each individual client to really show, "Look, this is the outcome you can verify in your Adjust dashboard."
Yes, thank you, Christian, for these tips. I think it just shows to our customers that there is obviously work before the campaigns are even running, making sure that the setup is right, universal link is set up, and you maximize your chance of leading your reengaged users within the app. There is also work to do during your campaigns on the ongoing optimization that is happening, checking your clicks, making sure that nothing is wrong on the publisher environment on the exchange that you buy from. And on the post-campaign or let's say the optimization side and moving forward, I think clients should keep experimenting on each of their user segments, making sure that they test different cases, different creatives and different overall scenarios that would push every single of their user segments however they segment and decide to segment them down the funnel and in order to reach their business KPIs.