💻 Thinking out of the black box to win with Google UAC – with Natalie Drozd (redux) 📦

Our guest today is Natalie Drozd, app marketing expert. Her specialisation lies in App Store optimization, and growing apps through organic and paid marketing.

In our conversation today, Natalie breaks down her strategies for unearthing the hidden levers in managing what can often appear to be a very opaque black box – Google UAC. She explains how to manage bids, creatives and audiences to drive performance even when there aren’t explicit ways to do so. I took a ton of notes from this episode that we were able to implement directly – and I highly recommend this episode as a masterclass in making sense of Google UAC.

As a note: As some of you may know, we had had Natalie on the show a few weeks ago. For various reasons we had to remove that episode – and re-record. This is the revised/finished product – enjoy!

ABOUT NATALIE: LinkedIn

ABOUT ROCKETSHIP HQ: Website | LinkedIn | Twitter | YouTube

KEY HIGHLIGHTS

🎰 3 levers you can influence on Google UAC

🎚️ High bids do not guarantee high value users

📍 Placements are key to performance.

😎 How you can use bidding to influence placements

🔃 The tradeoff between broad and targeted audiences – and how they are impacted by placements/bids

🔬 Breaking down the 3 UAC placements

🎛️ How you should choose your mix of placements depending on your goals of ROAS and scale

⚖️ How to use bids to prioritise outcomes

🏇 The competition aspect of placements

📹 Using YouTube for brand-building

☯️ The differences between passive and active marketing and the impact on conversions

🌱 Why YouTube is the place for growth

🎡 The product lifecycle informs ad goals

🧲 The stability of IAP-based campaigns

🎟️ Google’s minimum event volume threshold

🤹 CPA campaigns on Play search and Play browse; CPM campaigns on AdMob

🎞️ How to run video only and display only campaigns

📽️ Running video-only ads on YouTube for less competition

🏁 Different campaigns for the same event will compete with each other

📡 Google’s machine learning uses signals from your app

🗄️ Machine learning is not AI

🔝How to bid for maximum value

🎄 Seasonality impacts performance

🌎 The many factors to consider when targeting worldwide users

🏘️ The way to structure ad groups correctly

🛍️ Choosing the right event for the region depending on market size

🎚️ Event volume is critical

🔬 How to test new creatives in UAC

🗣️ The value of YouTube comments; even the bad ones

👍 All engagement is good engagement

🛒 Testing campaigns in non-core geos

📅 Why you shouldn’t make changes to campaigns more than once a week

🍂 Don’t forget about weekly seasonality

🗺️ How to structure worldwide campaign

📝 How to track changes to campaigns effectively

💅 100k: the number of changes that Google is making to your ad

🤑 Google is invested in marketer success

➕ Turning Google’s knowledge everything into a positive

KEY QUOTES

Your bids will influence placements up to a point

Your bids can influence where your ad will be shown. And if you bid higher, it doesn’t mean that you will get more valuable users, it will mean that you will get two more competitive options; which is more valuable for Google, not for you.

Why YouTube conversion rates are lower

When you’re playing a game and you need to get a rewarded video to get one extra life, you don’t want to buy anything; you just need to watch this video till the end to get your extra life. When you’re on YouTube, listening to music, or just watching a show, yeah, you’re watching the show, you are not here for ads. That’s why the conversion rates are lower. That’s why the prices are higher, but the quality is not usually as good.

Different event for different campaigns

And secondly, what I don’t recommend is running different campaigns for the same event in the same location, even though people usually say that they are excluded because you’re the same advertiser. From my previous experience, it didn’t work. If you run, for example, 3 campaigns for the same event with different bids, only the campaign with the highest bid will work and you will not get very different placements.

How to identify different cohorts

For example, you have subscription, you have in-app purchases, you have ad monetization: those use case scenarios are so different that you see that those cohorts do not intersect. In that case, you can separate and run campaigns for in-app purchase or for ad-rewarded watch.

The machine learning needs data to work

That means you cannot have 100 ad groups and only 10 installs per day because the system will not learn. What I always try to keep in mind, Google is machine learning not artificial intelligence. If you don’t give it enough data, it will not learn. If you give it enough and create good circumstances, it will work well.

The cost of machine learning is tied to events

And that’s why it’s so difficult, for example, for games with difficult monetization strategies with an in-app event of hundreds of dollars to optimise, just because it’s not easy to find users who are ready to pay for it. Unlike on Facebook, you don’t have to have audiences, or on Snapchat where you can also get audiences that are localised. Here, it’s on machine learning and that’s its drawback; that it learns, but it’s learning has a cost. And the more expensive your events are, the more expensive your learning is.

How to structure ad groups

I try to use semantic separation, breaking it down into different use cases, or tone of voice, or even the messaging I have.

I look at Google like a tree, and every ad group is a different branch. And you can offer different things in there, use different keywords in text assets, and also images or videos with different main ideas or main call to actions used as well.

It really helps because, according to what I learned, Google optimises its audiences in the campaign at an ad group level. And that’s why we can see that sometimes CPI can be different on different ad groups. That’s because for different creatives they find different people. Surprise surprise.

Testing creatives

So what I do is to test new assets in install volume campaigns on small budgets, because they are cheaper. And you can see if any creatives can get impression volume, because that’s extremely applicable to videos. From 20 videos, usually only one or two get impressions, based on my experience.

If I see that this video works well, I can add it to an ad group in the performing campaign.

Campaign changes only once a week

What I try to do is I make one change in one country, and I dedicate myself fully to this location. The next day, I’m working with a different country group. Let’s assume one day I work with the US only: checking all the creatives that are going out; is the bid enough; does the budget need any changes. The next day, I’m working on Latin America, then I’m working on Europe, then in Asia. That means that all the campaigns have enough of my attention, and I give them enough time to optimise and to check if the new creatives are going to work or not.

100k campaign changes

Google runs different creatives on its own; it doesn’t just change placement, it changes text. Sometimes it may happen after 100,000 changes per day, and you don’t even know about it. And that’s why you have to be patient and to leave it at least for a week to let the machine learning figure out if it can find a good audience or not.

FULL TRANSCRIPT BELOW

Shamanth: I’m very excited to have Natalie Drozd on The Mobile User Acquisition Show. Natalie, welcome to the show.

Natalie: Hi, I’m so happy to be here. And I hope that everyone will find this podcast valuable and interesting.

Shamanth: Excellent. Yeah, we are thrilled to have you, certainly because every time I’ve spoken to you in the past, I’ve certainly learned a lot and taken a ton of notes on how to run Google UAC. You certainly bring a very nuanced understanding of what a lot of people think is just a black box. And as I’ve learned from talking to you, it doesn’t have to be a black box. So that’s perhaps a good place to start. So is it really a black box?

Natalie: I don’t think so. Thank you for such a sweet introduction. My life motto is never stop learning, and knowledge is not useful if it’s not shared. That’s why I love to learn, and I love to share my insights, because then we can multiply the knowledge we have.

When it comes to Google UAC campaigns are usually considered as black box because there are not that many things you can influence. But if you dig a bit deeper, it turns out there are things you can influence and that can help you to get better results. So I would not call it a black box, but it’s not as straightforward as, for example, running Facebook ads or ads on any other social platform.

Shamanth: Yeah. And we will dig into what levers you do have an influence over, in controlling the performance. A good starting point would be bidding. So what impact does bidding have in influencing where your ads get shown?

Natalie: This is a good start because, when we think about marketing, we always think about money. Bidding is where the money is.

So when it comes to bidding, we all know that you can bid for installs (CPIs), you can bid for actions (CPA campaigns), and if you have Firebase, you can do ROAS targeting. But the structure is usually the same in all of those cases, because the higher the bid, the more different placements you will get. The impact that gets you to higher value placements doesn’t mean that’s really good for you. But it means the competition there is higher.

When we take a look at different placements where your ads can be shown in different inventories that Google has, we will see three main different parts:

– One of them is Google Play, which is Google Play search and Google Play browse. When you’re looking for an app, and you will see the first two slots are usually ads. Or when you’re just browsing, scrolling, looking through different apps you have or similar apps, there is also a ‘Recommended for you’ ad section. This is also a very valuable part.

– The second one is AdMob. This has a huge, as Donald Trump would say, amount of users, because that’s all the ads on websites, on different apps, and anywhere on the internet with all the applications that have monetization through ads.

– And the third one is YouTube. This is the flagship Google product, because they are always trying to show it as one of the best things to have, with a separate network.

Those three are different things, and they have different use cases. When you have bids, your bid actually helps you to think and it helps you to prioritise where you will show. YouTube is the most competitive one, and if you bid high, you will probably get views on YouTube, either YouTube video or any auto-generated things. We will talk about them later. But the price is higher because competition there is higher.

When you go to Google Play search or Google Play browse, you’re only competing with other apps, because that is the only thing you can find in the App Store and the Google Play Store, so to say.

When you are running ads on AdMob, there are a lot of so-called cheap impressions. But here you’re competing with other, for example, ecommerce sites which try to sell you things. It also has apps and websites. So the competition is higher because the amount of players is higher there.

And when it comes to YouTube, it’s even more difficult because you’re competing not only with apps, not only with websites, but with a lot of brand campaigns. And as we know the top Fortune 500 companies really care about their branding and they will spend millions of dollars, just to make sure that people know who they are and what they do. And what is even more important is that, on mobile devices when you run UAC campaigns on YouTube because they lead to placements, you’re also competing for desktop brand awareness campaigns because people usually don’t care if it’s a desktop or mobile version when they run brand campaigns.

Coming back to your question,

your bids can influence where your ad will be shown. And if you bid higher, it doesn’t mean that you will get more valuable users, it will mean that you will get two more competitive options; which is more valuable for Google, not for you.

Shamanth: Right! I like that characterization, because you’re saying, if you’re only on the app stores, there’s just other apps, and very likely you’re competing only with apps in your genre, right? If on Play Store browse and Play Store search, that’s only your genre, on AdMob it’s wider, and certainly YouTube is much more expensive because the big brands consider it premium inventory. So that’s not necessarily the most valuable for you; it’s just what’s the most expensive.

So why would it make sense for an app to even look at YouTube?

Natalie: Brands want to have ads on YouTube, because it sounds cool. That’s my answer. And this is the place where all the big brands have their ads.

One small distinction between passive marketing and active marketing: active marketing is when a user is already searching for an app that needs to satisfy his or her need. This is the place for Play search and Play browse, and it’s also Apple search as an Apple. For a person who’s already in the Play Store, it means that he is looking for an app, that’s a huge signal that this person wants something to solve their problem.

While AdMob and YouTube, it’s passive marketing, because people are usually disrupted from their main activity.

So why do people want to get into YouTube? Because it really helps you to drive brand awareness, and at least awareness about what you do. You get a lot of impressions, and if your goal is to make sure that everyone knows about you, that’s probably a good place to be. But you cannot expect really high results in terms of ROI or LTV. Why? Because you target a lot of users, and you don’t know if they really need you or not. But they will definitely know about you.

Shamanth: Right. And it also sounds like YouTube and having a really broad campaign might make sense if you’re looking to scale, because Play Store browse and search have relatively limited volumes. If you really want to reach large volumes of people, then you have to get to YouTube.

Natalie: Exactly. Because as we know, people who are actively searching for any type of app; search is only limited to them. You do not just sit on your Play Store or App Store searching for apps, because you don’t need them all the time. It’s why the amount of queries is relatively low.

But people who are on YouTube are using it for really different use cases each month. If you’re really satisfied with your campaign—and you see that the main placements are Play search and Play browse, which are attributed as display network and search—and you really want to scale, YouTube is a good place to go. Because it helps you to inform new users about your app, and maybe create this demand for your app. Because you can just pick the demand with Play search and Play browse that already exists.

Shamanth: Right. So with this in mind, right, if you want the highest quality, and volume isn’t important: you keep the bids low; you’re on Google Play search, Google Play browse. And if you want more volume: you bid high; you get more expensive entry into YouTube.

So with that understanding in mind, how do you recommend structuring campaigns?

Natalie: First of all, you need to understand your ‘why’: why do you need to run the campaign? You need to get more users; you need to get valuable traffic; or you need to scale. Depending on different parts of the product life cycle, you will be on different stages.

And what I usually do is I do have campaigns for just main actions—these are usually in-app actions. And that’s for campaigns that are usually Play search and Play browse, because what they do is they try to find valuable traffic and they are usually very stable. Because the placement is the same, we know that the amount of users that search for a certain category is relatively the same, unless it’s a highly seasonal topic. For example, apps for holidays, Christmas, different gifts. That’s why CPA campaigns with main placements on Google Play search and Google Play browse are totally okay if you want stability, and if you want things to go smoothly.

On the other hand, if you need to grow—this is also very important. If you want to get on YouTube, you either need to bid very high, or run an install volume campaign, because it will get everywhere. And cheap impressions are more likely to be gotten on AdMob; for YouTube impressions on Google Play search and Google Play browse are very expensive. That’s why CPMs there aren’t very good.

If I were to get into YouTube, first of all, I use videos, and secondly I use install volume campaigns to get the impression volume I need.

Shamanth: Yeah. So thinking about this, it’s: have a low-bid campaign to get high value traffic, and a high-bid campaign to get YouTube traffic. Do you find that to be a good structure?

Natalie: Not really, because everything depends on what events you are optimising for, what type of billing you use, and also what type of campaign you have.

For example, you can run video-only campaigns on YouTube, and to do so you have to ask your manager if you could run a video-only campaign, or just send an email to support—if you don’t have a dedicated manager. They can click on something in the backend, and you will be shown only on the video placements. And in this case, you will not be competing with the campaigns on Google Play search and Google Play browse, because there is no video and your video will not compete with your app icon or title there. Running normal campaigns for valuable users on Play search and Play browse, and running video-only campaigns to get impressions on YouTube or AdMob with your videos or static images.

Shamanth: Right. So this can be competition, if you say let me bid high on one campaign, and let me bid low on another campaign; those could just compete with each other.

Natalie: They will definitely compete against each other.

Shamanth: They will compete with each other. Got it. So you pick one as your key goal. You say, hey, I want value, then you just run the low bid one. Or if you want volume, you just want to run the high bid one on YouTube.

Natalie: Not exactly. What I wanted to say is that you cannot have different campaigns for the same event.

For example, you can run an event purchase campaign to get impressions on Google Play search and Play browse, and thus valuable users. And you can run an install volume campaign for the same location to get impressions on YouTube.

Or, for example, if your ad has different monetization strategies.

Shamanth: So you treat each behavioral event as a different kind of user, and you target them. And within each event, you don’t set up multiple campaigns.

Natalie: Yeah, it makes sense. Google uses machine learning, so it uses signals from your app. It’s not artificial intelligence, it cannot think for you. And that’s why I also don’t recommend using events from the same funnel.

For example, if you have event registration, and then even purchase: everyone who purchases, also does the registration because he or she cannot do without it, in many cases. That’s why it’s not useful to do it. But as I mentioned, if you have separate user journeys, this is the case where you run different campaigns.

And yeah, install volume campaigns always work. And they usually do not compete with CPA campaigns. So if you have only one funnel, and one user journey, use install volume campaigns to get impressions, just be sure that the quality of traffic will not be as high as in CPA campaigns.

Shamanth: Got it. And still staying with bidding, if you have to bid conservatively—you said, if you want to bid for maximum value—do you bid your actual CPI? Let’s just say you want $2 CPI and $10 cost per subscriber or purchase. So do you bid $2 and $10 in those campaigns, or do you bid higher than that? Do you bid lower than that? How do you think about that?

Natalie: Well, it really depends on your past performance. What I try to do is to bid 10% higher than I want to get, because what I saw in my campaigns is that Google really tries to get below your target CPA. And if Google stays there for a week or so, then it’s priced to scale.

It doesn’t always work; it doesn’t work now, during this holiday madness. That’s what I would say. But in normal life, when there are no crazy events taking place. That’s how Google machine learning works from my observation. It’s not useful to do it these days when there are a lot of advertisers with huge budgets bidding for Black Friday or for the holiday season, but wait until February or March where nothing extraordinary is taking place. That’s how in normal conditions, the bidding mechanism can work.

Shamanth: Sure, and I like how you characterise bidding as a way to be informed by the different user funnels and user journeys.

Just to build on that, let’s just say an advertiser has a relatively simple user journey. Say this is a subscription app, or this is a match-3 game with the first three IAPs being very standard. There’s no multiple monetization streams; it’s fairly straightforward in terms of the user path to monetization. What does the ideal campaign structure look like, if they are worldwide? Let’s assume they are worldwide, how many campaigns do you recommend having? How many ad sets do you recommend having? Obviously, the ad set depends on the different apps, but how should an app start to think about it, if they have, let’s just say a very simple monetization structure?

Natalie: First thing I want to do is I don’t run worldwide campaigns, because you have to keep in mind that LTV may differ from region to region. But if the game is very simple, let’s assume that LTV is the same everywhere, and there is only one language—that’s also important thing to keep in mind—there’s one campaign for the event purchase or if it’s a match-3 game, that would be, for example, rewarded videos, and one campaign for install volume. That’s small, but it’s enough.

And if you want to get more assets, this is what creative groups are created for. I would suggest, if you want to get different ads, or different text assets or videos, you can always add more ad groups. Just keep in mind that it has to have some events.

Shamanth: What’s the ideal volume of events?

Natalie: I keep the Google hygiene of 10 events per ad group. I know it’s not a statistical value, but at the same time, it is enough for Google to learn and to have users to optimise for. So 10 in-app purchases, for example, if you’re running for in-app purchases; or registration, which is even easier. Or if it’s rewarded videos, it’s much easier because when you record every conversion, then you will have a lot of events to optimise for.

Shamanth: Right. And presumably, if you need 10 events per ad group, you set up 5 ad groups if you can get 50. You set up the number of ad groups accordingly.

And how do you decide on how to structure the ad groups? Let’s just say there’s a purchase-optimised campaign or a subscriber-optimised campaign. You say, right, I need to have three ad groups in here. What’s the structure of what goes within these ad groups?

Natalie:

I try to use semantic separation, breaking it down into different use cases, or tone of voice, or even the messaging I have.

I look at Google like a tree, and every ad group is a different branch. And you can offer different things in there, use different keywords in text assets, and also images or videos with different main ideas or main call to actions used as well.

It really helps because, according to what I learned, Google optimises its audiences in the campaign at an ad group level. And that’s why we can see that sometimes CPI can be different on different ad groups. That’s because for different creatives they find different people. Surprise surprise.

And that’s why, if you add match-3 game on one, you would have just simple text like: “Play this game.” and everything which is gameplay related. A different one would be about use case, like: “Are you bored at home?” And a third one would be more about your brand, if you’re a popular match-3 game. So that’s why different users will see the message that appeals exactly to them.

Shamanth: Yeah. Is there a risk, in that case, that one use case has very high LTV or very low CPA, and that use case gets all of the traffic?

Natalie: It definitely can happen, because first of all, if you have three ad groups at the very beginning—day 1 or 2—they have the same amount of impressions, but then they start to optimise. And secondly, if you, for example, add a new ad group while the campaign is ready running, this new ad group will be prioritised at the beginning. And that’s why it can change the whole situation, for better or for the worse, because it finds new users.

So what I’ve tried to do is not to create too many different ad groups, because it will steal so-called traffic from other ad groups. But if I see that some of the ad assets are not working well, I will just change them in the oldest group. And it will also refresh the mechanisms of this old ad group with new changes, and will get new impressions.

Shamanth: Got it. And which brings me to asking you about how you think about running creative tests in Google UAC? Because you did express the kind of constraints that are present; the chief constraint being you can’t have two campaigns, and you also cannot isolate creative level performance very clearly.

So how do you run creative tests, if let’s say you have a new concept to test? You set up, let’s just say four ad groups; one for each use case and they’re all performing well. And then use case A: maybe you wanted to show gameplay, or you want to show the subscription mechanic. You have a new concept that you want to test. How do you think about testing a completely new concept?

Natalie: Well, it’s a good question for different approaches. The easiest one is just to create a new ad group in your campaign. And as I warned, it may take all the traffic, and it may be more expensive than the old ones or less expensive, it’s different. So it’s tricky and a bit dangerous.

So what I do is to test new assets in install volume campaigns on small budgets, because they are cheaper. And you can see if any creatives can get impression volume, because that’s extremely applicable to videos. From 20 videos, usually only one or two get impressions, based on my experience. If I see that this video works well, I can add it to an ad group in the performing campaign

in a different one, because Google saves—especially when it comes to videos, not about text assets—the previous history about it. This is also very interesting on YouTube. And I know that was a very provocative topic that people asked how videos on YouTube are shown and how they are prioritised.

The thing is, if you get YouTube placements, you cannot actually comment under the ad. And if you have YouTube unlisted video, no one can see it. But Google has bugs on YouTube, and even though your video is unlisted, in some use cases, you can leave comments. And it can say good or bad things for you, but any comment on the video is more valuable than no comments. And what that means is that it really sends very valuable signals to Google that this video is good, even though a user may write: “Oh such a boring app.” or something like that.

Another thing is that sometimes big brands run video campaigns with videos that are not unlisted, but which are included in the brand YouTube pages. One of the best examples is eToro: that’s a cryptocurrency wallet. They ran a huge campaign, and it worked so well that people were Googling for it, just to leave a comment under that ad. You can find it under “eToro YouTube commercial”. You will see how many comments are there and how many of them refer to an ad.

So if you run YouTube ads, if you want to get to YouTube, the video has to be really high quality, and it has to make the user want to write something under your ad. This is exactly what happens on Facebook: if users are engaging with your post, then it means that it will get more impressions than the one which is just there. Sorry, for moving away from the question but this is a very important thing.

Shamanth: Yeah, the social proof matters as much on Google, on YouTube, as it does on Facebook. And it’s not immediately obvious to a lot of people. That makes a lot of sense. So what you do recommend in terms of running creative tests is setting up a new ad set, setting up an install-optimised campaign, running videos there. Whatever gets the most traction is most likely to get social proof, and most likely to perform, so you graduate that, so to speak, to a main campaign.

Natalie: Right, and also a different strategy that can be used is, for example, you may run a campaign at a location where the CPIs are low, but you can get a lot of impressions. This is a really popular case for testing creatives in India, and then putting them in different countries.

I would run English-speaking tests in India, because the user behaviour of people who speak English in India is really similar to people who speak English in other parts of the world, when it comes to games. So when it comes, for example, to business apps, or fitness apps, that may be different, because the user behaviour alters, but games are the same. So this is a good thing to do. And if your video is working well and it has a lot of impressions on the Indian campaign, very likely it will also get high impressions on other campaigns. Again, this is my experience, maybe it will not work for you guys. But in my case, it works.

Shamanth: Absolutely, we do use a very similar approach on Facebook, which is to run creative tests in India or tier 3 countries and whatever wins you graduate that to core geos. So I think that does make a lot of sense.

Natalie: I like India, because there’s a huge amount of population. This is also a very good place for starting Android apps, because a lot of devices, it’s easy to find bugs, and it’s easy to optimise and see if it works well or not.

Shamanth: Yeah. A number of games I have worked with have soft launches in India. Interestingly, I believe the Philippines is another option that works well for these purposes. Certainly, a large English-speaking population and relatively low CPMs, and relatively low CPI. So I think that all makes sense.

With Google UAC, something that does happen is the disruption of performance every time you make a change, every time you add new creatives. So when you are running these creative tests, moving them into main campaigns from India or Philippines, how do you ensure there’s minimal performance impact?

Natalie: That is also one of the main Google’s rules for hygiene: not to make changes more than once a week. I strongly recommend that because, as I mentioned before, it is all machine learning, and if you make too many changes at the same time, the system receives too many different signals, and it doesn’t know what to optimise for.

We also have to keep in mind the weekly seasonality. People, during workdays, behave in one way and then differently during the weekend. And that’s why if you made some changes, and even if you see that it is not working, I strongly recommend waiting because you have to see the whole cycle, not only one or two days.

Usually the first week of running a new campaign is not a good indicator, because what usually happens is that, as it learns, the performance becomes way better, especially if you are trying to run a completely new campaign in a new country. At these times, you really have to wait. But if it’s only creative testing, one test per week is good.

Shamanth: Right, that makes a lot of sense. And speaking of focusing attention on different geo tiers, how many tiers do you recommend that an advertiser have? Let’s just say they’re worldwide—their market is throughout the world—let’s just say they’re a subscription app. How many tiers should they have? How many languages do you typically recommend that they target?

Natalie: What I try to do is run one campaign per language, and one campaign in the same time zone. It’s important because Google also has 24-hour seasonality, and if you’re running a campaign which is worldwide, it cannot clearly optimise at which part of the day it has to show ads to whom. That’s why, for example, sometimes it works to have a Canada and US campaign, because their purchasing power is similar and timezones are the same. Something else for Europe, something else for Australia, something else for Japan or Korea; because firstly languages, and secondly time zones, so everyone has the ads at the relevant time.

Shamanth: Right. So you don’t really combine, let’s just say, Argentina and Spain, even though they have the same language?

Natalie: No, for two reasons: different time zone and different purchasing power.

Shamanth: Right. But if you have to have one language-based campaign, would you recommend one campaign for every country in Europe, let’s just say? Or how do you think about that?

Natalie: No, I don’t think that’s a good idea, because in this case, there are a lot of small countries and their campaigns will learn very, very slowly, I think the campaign in English will definitely be good.

Also, we have to keep in mind that, if you run campaigns and you get Google Ads placements on Play search and Play browse, the ad will show your localised product page. So you will get the most valuable users in the language they use.

I love searching for those cases, because my phone is in English, my Google account is in Polish, and my native language is Ukrainian. So I just love seeing Google ads in four different languages, because they also always try to give something in French or German.

Google uses the languages that are on your phone. That is a very important thing to keep in mind. For example, in the US, there are a lot of people who have Spanish and English, and that’s also why this is a good audience. Sometimes you can run, in one location, campaigns in two languages, and it will work well. If you see that there is a huge amount of population which speaks a different language. For example, Polish in the UK is a thing, because there are a lot of people from Poland who live in the UK. Or, for example, India with different languages there. If you want to target people who speak Hindi, you would run a Hindi campaign, and one English-speaking campaign.

Shamanth: Is there a population size or audience size that you think might be suitable to target? As context, if you’re looking at Europe: maybe just running in Spain—that’s 40 million people—you said that may not be optimal. But what’s the ideal audience size or population size per campaign?

Natalie: I can say there is no optimal size, because the other thing you have to keep in mind is the amount of the events you are targeting for.

For example, you could run a campaign in Spain, and if you have a hundred events per day, you’re a lucky person. But if you’re running a campaign for Spain with in-app purchases for $200, it’s not enough to have to have the whole of Europe to make the machine learning algorithm learn who are those users who are ready to pay $200 for a game. So it’s not about the audience size; that’s why you cannot choose an audience size in Google Ads campaigns. It is all about the events you’re sending. The tricky part is that the more you send, the better, but at the same time, a lot of developers don’t really like to send too much information. But I think Google already has a lot because it has Google Play console, it has creatives, and that’s why it knows how your app works. It knows your retention rate, so just take it as it is and try to benefit from it.

Shamanth: Yeah. So what I’m hearing you say is within the constraints—same time zone, same purchasing power, same or similar languages—have as wide an audience as possible, so that you maximise the power of the machine learning.

Natalie: Right. You have to know the rules of the game to play it well. And this is, I think, the most important thing with Google, because there are so many different approaches, but unlike with Facebook marketing, there are not so many good guides on how to run campaigns. That’s why you really need to take into account everything you know and test it; and don’t blame the machine learning for your mistakes—if they happen.

Shamanth: Yeah, right. And this is certainly a game that you seem to have mastered at a level that not a lot of people have. So I’m very, very appreciative of that.

Natalie, you’ve shared a number of ways, number of levers that you can use to change campaigns and impact campaigns. How do you recommend that people keep track of a lot of these changes? Let’s just say the creative tests, or bidding tests, different geo tests: how do you recommend people keep track of a lot of this?

Natalie: There are a lot of things happening, that’s why you definitely need to somehow keep track of what was changed and when. I have two different options:

1) I just send messages to myself about different campaigns, when and what change was made. So I have a very interesting monologue, mostly in Slack.

2) I use data visualizers. And in those cases, you can also add timestamps for different campaigns, to see it from different perspectives. So you can see, for example, a timetable for the last month and add different timestamps for when the bid was changed, when the creatives were changed, and monitor how CPI was going up or down, changing during this time.

It actually requires a lot of, first of all, self discipline, to keep track of all of the changes because sometimes, it may seem very insignificant to you. But it turns out if anything went wrong, for example, you uploaded a creative which violated the policy, and that’s why everything else became really very, very poor performing. If you don’t know that this creative was uploaded this day, we’ll be looking for a very, very long time to figure out why the price changed.

Shamanth: Right, and it becomes so critical to know and understand what changes you’ve made; just because of the way the machine learning works, even the smallest change can trigger huge impacts on the algorithm and how it performs.

Natalie: Definitely. The changes we make from this side, as marketers, are only a very small portion of all the changes that the machine algorithm does from the backend. You don’t see them, but know that Google uses, first of all, auto-generated creatives.

Yeah, you have to be really patient. And you really have to keep track of what you do and be responsible for your actions. Because sometimes tests are successful; sometimes they’re not, as long as you can prove that this test was important, and why and how you did it. And if you did it right, it will definitely pay.

Shamanth: Certainly, Natalie, as you have expressed in this conversation, I think there’s so much that marketers can do within Google UAC. Just there’s so much intelligence that they can bring to how they operate, what can seem like a black box. This has been incredibly instructive Natalie, and this perhaps a good place for us to start to wrap. Before we do that, are there other things that you feel that marketers should keep in mind?

Natalie: First of all, I think, be patient. And don’t stress too much if things are not going right, because not everything depends on you.

And second, reach out for help if you need anything. Managers from Google are usually very responsive; Google is top for a reason. And if you have a dedicated manager, that’s wonderful. Do ask them for help. If you don’t have one, support is not the best option, but at least it can give you guides and manuals for what you can do.

And you can also reach out to fellow marketers, who can help you if you have some struggles, because it can be a win-win game. If you as a marketer get good results and Google UAC get good results on their side. We always have to keep in mind that the main goal of Google UAC is to make money for Google. And Google is only making money, if we as marketers have good results. So that’s why everyone is interested in running good campaigns. Because if we are not successful, we will not spend money any more.

Be proactive, be patient, and reach out for help if you need anything. This is something that helped me a lot, and I’m very grateful to my Google managers, and also fellow marketers, who communicate about running UAC campaigns. And that’s what I recommend to you because it was my way to understanding the Google black box better.

Shamanth: Certainly. That’s perhaps a great place for us to wrap up. Natalie, before we go, how can folks find out more about you and everything you do?

Natalie: You can find me on LinkedIn: Natalie Drozd. I think there will be a link at the end of the podcast page. And yeah, if you have questions, just find me. I’ll be super happy to discuss anything, because by learning and by teaching, we become better versions of ourselves.

Shamanth: Excellent. I appreciate that. And I so appreciate your sharing, Natalie. Thank you so much for being on the Mobile User Acquisition Show.

Natalie: Thank you very much for this chance to speak and to share my knowledge with you. Thank you Shamanth.

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!