Creatives and new funnels in iOS app advertising
The problem with data aggregation at low volumes and launching a web2web funnel for apps.
Foreword
We launched an advertising campaign for a mobile app and tested various creative approaches. Initially, we decided to take a systematic approach: to create personalized ads that displayed zodiac signs and a brief horoscope for each of them for the coming weeks.
In our targeting, we selected users by date of birth and zodiac sign to show them their personal horoscope. For example, if it is October, we show ads with a horoscope for Libras, in November — for Scorpios, and so on. Thus, the creatives were prepared for the entire year ahead.
After the ad launch, the metrics looked normal at first glance: the app was being installed, and the cost-per-click and cost-per-install metrics were acceptable. But there were no in-app sales. We began to analyze the situation and discovered a discrepancy in the data: internal analytics showed that payments were indeed being made, but Facebook and Google were not seeing them.
Since the app was only available on iOS, this became a separate problem — it turned out to be very difficult to set up conversion tracking correctly. Apple is reluctant to share user data: events are transmitted in aggregate form, with forecasts and delays. This is critical for small budgets, as algorithms do not receive enough signals for training.
This made it difficult to test creatives. We could only see the primary metrics — CTR and cost per install — and then everything turned into a “black box.” We worked intuitively, without being able to determine exactly which creatives actually led to payments.
The situation in Google Ads was slightly better, but there were still difficulties with transferring conversions from iOS. According to some reports, the new Firebase SDK update should improve the situation, but in practice, the problem remains: the small number of events means that predictive models do not work consistently, and some conversions are simply not transferred.
Dirty creative with clickbait isn't always bad actually
We tried many options, but the unexpected solution came by accident. We decided to test the simplest, even crude creatives — with pain in the text, crying girls, and bright emotional headlines. Surprisingly, it was these “dark” creatives that worked: they brought in a lot of trials.
While it is not certain that this approach will be effective in the long term, it allowed us to “wake up” the prediction model and obtain at least some data on conversions. After that, we returned to more balanced formats — testing infographics, videos, calm texts, and visuals.
However, the problem with transferring conversions to iOS advertising accounts remains. If the app has a large number of events, the system can correctly predict and transfer the approximate number of conversions. But if the budget is small and there are few events, advertising becomes difficult to analyze, and testing creatives turns into a game of blind man's bluff.
web2web воронка
I would like to note separately that we have recently observed a trend: the web-to-web advertising model works better. Even if you have an app and most of the action takes place in it, the main services and products are delivered through an alternative app on Android or iOS, but there is a way to avoid paying a percentage to Google Play and the App Store. Many people do this, and I recommend it. Of course, you have an app, and you will need to develop a special solution for this. If your service is provided through a website, this does not apply to you. If you already have an app and want to save money and pay a lower commission on in-app store payments, then a web-to-web funnel is the only thing that can help you right now. If you haven't done it yet, contact us immediately — we will help you with this. We already have extensive experience in this area.
That is, you set up advertising and promote the website. Next, as a rule, there are quizzes, as you need to get the person to pay. There can be various options. Quizzes are very popular now: you create a website, roughly speaking, a long landing page with different steps, which does not scroll from top to bottom, but opens a new screen when clicked. This way, the user scrolls through the landing page not from top to bottom, but like a book or magazine. They click on a button, answer a question, or interact with one screen. After that, the screen flips, and the user sees a new screen with new information and new questions.
Personal ads
iOS conversion tracking
Interactive quizzes
web2web funnel
In general, this interactive quiz model works very well to warm up the user as much as possible and make them reluctant to leave the site. At the last stage, when you offer to pay a small amount for a trial, they are highly motivated to do so, as they have already done a lot of work, answered many questions, and read a lot of text. The quiz seems as personalized as possible, tailored to the user and interacting with them. For example, ChatGPT, depending on their answers, substitutes personalized texts based on these answers, and so on.
In other words, it is the most “human” quiz possible, proving that you are not a fraud and that everything really works — a good solution that can be implemented.
And in general, it doesn't have to be a quiz — it can be a regular website with a standard structure: how we can help you, why choose us, what we offer, what are the benefits, try it, here's the proof, and so on. At this stage, the person also makes the first trial payment, or at least registers, for example, by leaving their contact details. Then your sales department or funnel starts working after registration via email or messengers, and the customer makes a trial purchase of the application. After that, they go to the app store, select the application, install it, and do not pay anything in the application itself. The user enters their username and password, or the email and password they chose when paying for the trial version, or just their email, and receives a password automatically. That's it — they can now use the app. In other words, the main payment was made on the website, you don't pay anyone a commission for this, and the conversion and use of the app proceed as usual; you just don't make any payments in the app store itself.
Stop paying commissions to app stores
On the one hand, you could say that this is a bit of “cheating,” but on the other hand, you are not breaking any laws. Think about it: how much money do you pay for advertising on Google to promote your app, how much for SEO optimization in Google Play app stores? And after all that, you still have to pay a commission for the paid service purchased in your app. With this standard classic approach, you give Google almost 50-60% of all your earnings. That is, you earn money, spend it on advertising, advertising brings in new customers, you get these customers, and then you pay Google again for what the customers paid you. And after that, there is no clear answer as to who is a “cheater” and who is not.
That's why the web-to-web funnel for applications is a great topic. We highly recommend it, and if you need help setting it up, please contact us.
