Google Ads for Mobile Apps: Choosing the Best Optimization Event

They had previously been just focused on installs. They were like, "If we can just get cheap installs, we're fine." Um, and we figured out the mechanics that like a cheap install will still be profitable for us. Which is true. Again, like that's not wrong. They they that works. Um, however, one of the things that we talked about with adfunded customers or um, apps, as most people know, you make your money by showing more ads, right? So, to show more ads, what's another important part is like people opening the app more often, right? You get a chance to serve another impression. And so one approach that we took was okay I know we're you can see here like mostly optimizing for first open events um across all campaigns but then we switched one to a Hi everyone welcome to another episode of Tangenti 1011 I'm Roman from Tangent

and today I'm joined by Ashley. Hi Ashley. Hi. How's it going? Yeah fantastic. Fantastic. We've already done one video with Ashley and it was an absolute hit. Uh, and we've invited her to talk again about the Google ads. What are we going to discuss, Ashley? So, today I think we're going to be talking about how to pick the right optimization event. Um, you and I had talked in the past about how many people are maybe not optimizing for the right event or how people go about figuring out what the right event actually should be. So, that's what we'll get into with some examples of it today. Sweet. And, uh, why did we invite you to talk about the Google campaigns? Let's see. So actually

you and I know each other from previous days. Um I spent almost 10 years at Google and for six of those years I led the app ad sales team. So my team um supported most of the like the larger gaming marketing teams on Google's app campaigns UAC product. And about oh I'm trying to think now maybe 16 months ago I actually left Google. It's been yeah a little over a year. I left Google and um found myself doing consulting work for a lot of app companies who needed help figuring out Google. And so now the company is Candid Consulting. It's sort of merged into both consulting for apps but also running more of an agency for Google campaigns specifically. So I feel quite qualified to talk on the topic. Yeah. Yeah. Yeah. Yeah. Even in in the last video people demanding to have more v more videos with yesterday. So great

that's why we are here. Okay. I I know you've prepared some slides. Uh please share your screen. Uh full disclosure I haven't seen the slides only the first one. So if uh let me know if I'm asking the questions and the they'll be answered later in the slides. Just let me know. Right. I think probably not. I mean, feel free to ask any questions in real time. Um, and so yeah, I mean I I told you I kept it intentionally really really light on slide content. Um, I like to talk a lot more. So, um, yeah, we'll get into it, but as we said, we're talking about picking the right conversion event to optimize app campaigns. So, I think there's a few things to keep in mind when doing that or when you're approaching that. And so, I would say there's sort of like three things that are really important when

you're determining, hey, what's the the right event for me to pick and how should my campaign um really like what should it be focused on? And the first thing that you need to pay attention to is volume. So, if you're optimizing for something, you need to make sure that whatever you're optimizing for is able to produce enough volume for Google to actually learn. Um, and so I usually estimate that to be about 30 inapp events per day. So, I know like of course everybody wants to optimize for like the highest value user, but sometimes those highv value users, there's just not enough of them that occur, especially if you have like a limited budget that you're working with. Um, yeah. Any questions on that piece? Yeah. Yeah. Will you give us like some examples from your experience? What were the events like like that if that's not highest value? Yeah, we'll get into those. Can you see? Yeah. Yeah. Yeah. I

already asked the questions at the end. Examples. Um the other thing I think that's really important for people to understand is the speed at which those events actually happen. So I said that typically you want to pick an event that happens within 24 hours of the install. And ideally something that doesn't happen more than 72 hours um of the install or after the install. And this is so that the again Google's algorithm can actually learn with some you know speed and um quickness so that it's able to kind of figure things out quickly and pivot from there if it's not working. You know, you don't want something that's happening on like day 30 and then basically if if that event happens 30 days after the ad impression has been served it then it has to go back into learning mode. you know, almost 30 a month later, in some cases even longer if if people are picking something really far out. So, the speed at which a conversion happens is really really

important. Um, and then the last one is obvious, but I it's obvious, I think, to me, but I don't think it's always obvious when I like get into things with people is that you need to make sure that you're picking an event that actually has a strong correlation to what you care about, right? Is that a purchase? Is that time spent in the app? is that retention in the app like are you able to show that like if somebody does X ultimately it results in Y. Um and so that that's a really important piece and that one I think probably takes more time and analysis to figure out what that correlation is. Gotcha. Will you have examples of bad events? I mean I don't want to say that they're bad. They could just be optimized. Let's put it that way. And I think each event and and we'll talk about this in some of the examples like they can serve their own purpose. Um and so sometimes you need to I know you want to talk a little bit about like rorowaz for example

rorowaz is great like I think yeah wouldn't wouldn't we all love to optimize for rorowaz but sometimes those campaigns uh specifically with Google can't always scale all that well. So even if you do want to optimize for RO, I've seen a lot of other advertisers who need to sort of supplement it with something that's not rorowass. So like more of a CPA campaign. Um and so again like these different event types can serve different purposes and functions. Gotcha. And also another downside, this is actually something that I learned recently. you need to have uh Firebase if you want to send out inapp uh ads uh so for for purchases and for um adf funed that's true oh for both for both yeah ah okay so yeah you need to have

and if you don't have firebase uh yeah and and what about I I mentioned firebase what about the events in general uh is this a best practice or to to use Firebase or Yeah. Yes. Google's always going to tell you that. And I will say like I have seen I actually for the first example that that I'll show you that client had for a long time been bidding on their MMP events and um we made a case to actually switch it over to Firebase events and we did see a significant uplift in the number of conversions that we were able to drive. And so I think it's just important to remember that like Google gets a lot more data when you use Firebase. Um, and so if you're comfortable, I I've honestly I haven't run across anybody who doesn't have it at this point. Like it's kind of just like the standard now, whether people like it or not. Um, and

so Google gets a lot more data when you are using Firebase and that helps inform their algorithms and therefore the performance tends to um improve. So I got you. Yeah. And the maybe the last one since we touched the topic of AOS uh any uh requirements for a campaign when I'm running on AOS. So any like budget requirements or volume requirements or like I can start it at any point. You can start it at any point technically. Um but the budgets do need to be pretty high. So yeah and the learning period can be quite long. So I think that's the tricky thing with rorowaz is that it does ultimately I I have found for most advertisers perform well. But the challenge is is that it's a it can it can be a high price tag to like barrier to entry, right? Just to get in and allow the

campaign to learn. You have to be comfortable with it not producing results for a period of time like 3 weeks. So, if we're talking about, and I would say budgets have to be pretty healthy, too. Um, I would, you know, I think Google would probably recommend like $1,000 a day, which now we're talking about almost 20,000 that potentially is just gone because the system needed to figure things out. I see. Yeah. I think that's a drawback to it. And then the other one, as I mentioned, is the the scale. Sometimes it just doesn't it doesn't deliver enough volume for advertisers and so they have to resort to other campaign types. Gotcha. And on the contrast the events do do they require same amount of time or I assume less on the which one? Sorry like if we are doing event optimization. Oh no those ones don't require as long.

I don't I don't think that they have as high of a um budget requirement, but then again, like you're not necessarily optimizing for returns when you're optimizing for um an event specifically. So, it's a little bit of a trade-off. I see. Yeah. Um but yeah, I'll give you an example, the next case. So, I'm going to give a little background here before we I'll show you the results. So I have one client that let me take a step back. I think one of the really important things is sort of this um the correlation piece that I talked about here and really understanding like what do I actually care about for you know what somebody doing in my app. And so in this case, I had a client and for their particular business needs, they actually cared a lot more about just driving a higher number of monthly active users. To them,

that's what mattered. And when I first came on to the um onto the account, they were they were just optimizing for like an onboarding complete. It is a subscriptionbased app, but there you can do a lot within the app without needing to purchase a subscription. And so they were just trying to get more people in. But in talking to them being like, okay, but you actually care about monthly active users and being able to retain those users for a longer period of time, started to kind of question whether or not the onboarding complete was the best action, the best conversion event for them to be optimizing toward. And so we did a a hefty analysis to make sure that we were touching on all three of these things. Could we find another event that drove meaningful volume? Did this event happen relatively quickly that we could use it as a conversion event so that Google could learn from

it? And did this conversion event actually have a stronger correlation to that end result which is the monthly active users. Um we in this case we actually used mix panel. The client uses mix panel and we went through a lot of different events to actually see okay like can we see that we we ultimately ended up set settling on like retention to see for monthly active users was like okay if somebody's sticking around for a longer period of time they're retaining the monthly active should go up right that's that was the hypothesis there. Um, and so we spent a lot of time narrowing down a lot of different events and crossing them out basically and saying like, oh, could is this event like for example, this is a a messaging app. And, you know, we would say, oh, if they, you know, invited a friend to message, was did that check all these boxes? Did it have enough volume? Was it happening relatively quickly? Was there

a strong enough correlation? you know, if somebody um you know, like started a conversation, if somebody made a phone call, you can call like somebody started a video call. Like we went through all of these different ones trying to figure out the right mix of all of this. And ultimately what we landed on was we saw that people who started a conversation with somebody else had a higher retention than somebody who just like completed the onboarding. Okay. So I see I see. So, and you had the data like the mix panel dashboard always in front of you and you could uh kind of see like okay this event volume is this this event volume was that so you always had a dot in front of you. This is yes that was a critical component being able to figure it out. So based on that click over to the next slide. This is what we saw which is a pretty significant uplift in the retention rate

of the users. So again we used retention like we couldn't necessarily use monthly active users. Um that would be a really hard thing to optimize for. I don't even know how you would do it. But also as we take way too long right like to figure that out. So um what we ended up doing is you can see the previous retention rate is the period of time where we were optimizing for onboarding complete um and then the retention rate here and then retention rate um on day seven is what was after we switched it to conversation started and you can see the uplift in the retention the percentages are quite meaningful in this case. I do think it's important to understand though like yes certain costs did increase because of this. So you do have to change your mindset a little bit when you switch the action. So the cost per action went up. Um I think previously I'm trying to think we were at like maybe $3.80

for this is a US-based campaign for the um onboarding complete. for Android and when we switched it over to conversation started it bumped up to like 440 I think. So we did have a pretty healthy increase percentage-wise in terms of what the cost per action was but we're getting higher quality users. So you have to kind of like reframe your mind um in terms of costs a lot of times too. I see. Um, one of the other, I guess, like challenges that we had with this one was like we we were doing okay in terms of volume, but I wanted it to be higher. In this case, I was actually trying to get closer to like 50 inapp events per day. And so, we did have to increase the bid a little bit in order for us to get there. But, we've been able to recover the total volume in terms of like installs um, and users coming into the app by increasing the bid.

Gotcha. Yeah. So again, just I think going back and asking yourself like is the event that I picked actually the one that like is going to be predictive of a meaningful user to me later on. Gotcha. Any other questions on that one? I'll give you another example in a sec. Um yeah, th this one is pretty clear. Yeah. All right. The next one I feel like feels like common sense. Um it did to me when I brought it up to my client. But at the same time, what we were talking about earlier is like you'll see in the ne next example, we actually are still running quite a few combination of um event types like different campaigns bidding toward different events because they have different functions. So I'll show you what I mean in this sense. So this next one, the client is uh like 100% adfunded. So, they had been

approaching things. Um, I had a lot of stuff kind of grayed out here, but they had previously been just focused on installs. They were like, "If we can just get cheap installs, we're fine." Um, and we figured out the mechanics that like a cheap install will still be profitable for us. Which is true. Again, like that's not wrong. They they that works. Um, however, one of the things that we talked about with adfunded customers or um, apps, as most people know, you make your money by showing more ads, right? So, to show more ads, what's another important part is like people opening the app more often, right? You get a chance to serve another impression. And so one approach that we took was okay I know we're you can see here like mostly optimizing for first open events um across all campaigns but then we switched one to a D1 retention

event. So we're just up again this helpful to understand D1 retention. We originally had it D2 D2 or D3 retention and that was just too long for the system and also the number of events had dropped. Not not below that 30 threshold but like it had decreased quite a bit. So we went back to just day one retention. So these are users who are still using the app um 24 hours after the install. And you can see here, I wish I had like a little cursor, but um in that last column on the right, the drastic difference of 7-day retention just by switching it to optimizing for D1 versus a first open. I see. Twice more basically. Yeah. So, and then again, like I don't have the LTV numbers on here. I can tell you that the LT numbers are much much stronger um on the D1 retention events, but again, I think it's going back to the prior

example. You can see here too though that it's a lot more expensive um to run a D1 retention. The CPI on this is like, you know, almost $2.70. when you look at their largest campaign, the one um that drove almost 100,000 installs, you know, it was almost a dollar cheaper on a CPI basis. So, we did have to kind of get out of that mindset of optimizing toward a specific CPI because the CPI was just going to be so much higher on the retention event. I see. Yeah. And for the retention event, I I assume you were sending it from Firebase. Yes. Yeah. Is it a default event in Uhhuh. So I I also had cases like this when customers ask us to integrate uh day one retention as a custom event in order to send it to an ad network.

Yeah. Right. So that's that's the case. Yeah. So, this had to be a custom event that was created um in order for us to leverage it. Yeah. Um the other thing that Go ahead. Sorry, Roman. Yeah, no worries. Uh I I'm also wondering, did you discuss like other options like 10 ads view, interstitial view or rewarded ad view, something like this? You you absolutely could. I've seen a lot of games do that. In this case, this is not a gaming app. they're just serving straight banners. So, um, but yes, I have seen people do that where they actually say like, you know, if it served three interstitials as a as an event, you know, optimized for that, they have strong correlation with those users being more valuable to them. Gotcha. For sure. Um so again like that's a great example of like finding the right you know the right types of events again the right volume the speed

at which they happen and then trying to select those to optimize. Um for this customer you will see like if you think about it there's not an infinite number of users who will stay you know will keep the app or open the app again after 24 hours. So for them, this particular client, the total number of installs is still really meaningful to them. And so we can't rely solely on the D1 retention events. So that's why you see, and I I kind of blurred out there, there's different campaign types in here. They they kind of all do different functions, but that's why you see multiple um install campaigns still running because they supplement the volume. I see. Yeah. Um, this one was a cool thing. We tested this, oh, I don't know how many months ago. Maybe seven, eight months ago on like one, this client in particular has

like 12 apps and we tested it on one. Slowly we saw the volume picking up and like now it's the primary strategy for all of their apps is focusing on retention. Nice. It's been cool to see the shift. Yeah. Um, the last example I have is kind of a it's kind of a random one. I don't see it happen all that often to be honest where somebody is optimizing for two events at once. Um, yeah, within the same campaign. Is it campaign? What? Yes, I know. I see it. I have run into it a couple of times. Um, I'm here to tell people why they shouldn't do that. I'm intrigued. Okay. Okay. So, let's look at the next example. Let me come back over here. Okay. So, in this case, this this client, you'll see there's two different campaigns. The first one that they were

running before I I came on to it was optimizing for people subscribing to the app and then also purchasing gift cards within the app. Um, and as soon as I saw it, I was like, "No, I don't like that." Because those two users represent two different values to you, right? In this case, like somebody who is subscribing is far more valuable to you than somebody who just purchases a gift card. Um, because they don't get to keep all of that money, right? They they give it to the to the user. So I think the other thing that was important to understand is like they were like oh but actually at the time they were like well it's actually performing okay. They were looking at this column the cost per participated in app action and they were like no it's you know it's hitting our target and I was like well if you remove those 188 gift card purchases

your number your actual CPA is like way higher. And I know that seems obvious maybe to most, but if you're starting at this and maybe you're new to marketing as a whole, then this is like, you know, maybe a new concept for you. So, the strategy here was like keeping it simple. We want to focus on one thing at a time um so that we can really see like the true results. And obviously, you can see here we switched it over just to optimizing for subscriptions or yeah, for for start trials. Um, and you can see that the the cost the true cost because we're only optimizing for one event was significantly lower than the campaign that was optimizing for two. So, should we never optimize for two as I'm just trying to understand like is there why you what I mean again like in in Firebase you can combine them. So you can create a custom

event and say, you know, I want to optimize for people who start a trial and make and make a gift card purchase. I wouldn't, you know, that would be really a really small group of people, but you could do something like that. But to optimize for two separate events, I think the system is like it's all over the place, right? It doesn't know what to do. It's like, oh, I got a gift card purchase. Great. They got a subscription. is like, "Yeah, but again, those users aren't, you know, they don't the same weight." Um, so I can't really imagine too many cases in which you would optimize for two standalone events, but you could certainly be like a instead of a or uh situation, you know, subscription or gift card, it could be a subscription and gift card as the example. I see. Gotcha. And that's it. Those are the three examples that I have to talk through today. You showed me

fantastic examples from both Android and iOS. Is there a difference like when you are like choosing the event, how the Google treats those any? Good question. So, um, yes, Android you have a lot more flexibility with, right? Like you don't have to deal with, um, scan and you you obviously just get so much more user level data that the system can work off of. So, with Android, I'd say you just have a lot more freedom there. Um but and then on iOS I think your one your budgets typically have to be higher and again Google is modeling these conversions. So that's a whole other can of worms that we don't we won't go into but I would say the speed at which something happens with after the install is more important

on iOS than it would be on Android. Yeah. You know, I I gave the example of like 24 to 72 hours on iOS. I would say you your event has to be within 24 hours. Same. Yeah. um any scad network hacks whatever I I've started like I don't know two years ago or so like people oh I was also one of them like oh store pass it as a binary value uh in the first part of the conversion value pass uh one thing then use three other digits to pass a different thing how is I I don't have any hacks on on scan and to be honest I find myself using it with a lot less frequency. Like it's a good it's a good check if you're

running on multiple networks just to say like okay how are they how are they performing against each other but the data in which you get is so unuseful that like I I still have to if if you want to keep your Google iOS campaigns running or meta or whatever you're running on it's not going to show you down to like the asset level which is how I make campaign optimization. So even if it's like, hey, this is producing whatever the CPA is great on this. I'm like, I don't know which one to dial up and which one to dial down. So I still honestly I still use Google's modeled conversion data as not necessarily like my source of truth of like what happened, but I use that data to inform my decisions on how I'm going to manage the campaigns because it's more useful data. I see. You know. Yeah. Yeah. Yeah. Gotcha. makes a lot of sense. Um, I have

one more question, but I'll leave it in the comments. I'll encourage everyone to leave their com their questions in the comments. We'll ask Ashley to reply just to keep the comment section active. Um, thanks a lot, Ashley. Um, I hope I'm sure everyone like this video, so give us a thumbs up and subscribe and till next one. All right. Thanks so much, Roman. It's always great chatting with you. Pleasure is mine. See you. Bye.