Guide to Methods and Calculations

In this article, we will cover the basics of retargeting, namely:

1. What is this?

2. What games is it suitable for?

3. What is the principle of its operation?

4. Which users does it make sense to bring back?

5. How to evaluate the results?

Let's get started.

What is Retargeting?

Retargeting is the process of re-engaging a user into an app for subsequent monetization within that app. In our case, the app is a game.

I think everyone has encountered retargeting in various delivery services, transport rentals or taxis. In such services, it may look like an offer of a discount for the first order after a long break.

Sometimes retargeting doesn't directly benefit the user. In mobile games, when a company is just starting to think about re-engaging users, this is often the case. Such tests don't always show high efficiency, but they give the company an idea of ​​how ready the product is for retargeting.

There are also indirect signs that the product is ready to launch retargeting campaigns.

Do we need Retargeting?

Destruction of hopes

Retargeting is not suitable for every product by default. It cannot be perceived as a magic pill that will solve your problems (if you have them, you should solve them first and then think about retargeting) or will allow you to scale your profits 10 times. According to various sources, a good indicator for Retargeting is a situation when 10% of all marketing costs are directed to this channel.

Also, when thinking about Retargeting, it is important to understand what the user will do in your app when they return. In games, for example, you should not expect high results from Retargeting in hyper-casual projects, where the user has enough content for two weeks on average. In such a configuration, retargeting is still possible, but it is unlikely to bring in much money. Most often, games with a user lifespan of more than six months are suitable for retargeting, when they can leave the game before getting acquainted with all the content.

Possible benchmarks for Retargeting tests

If the project has Lifetime at least 180 daysthen you can already think about the fact that Retargeting can become your growth point. However, this indicator alone is not enough.

According to various estimates from different platforms (Google Ads, Moloco, Remerge), the minimum threshold for launching Retargeting is when the audience you want to return exceeds 50 thousand users.

Since you clearly don't want to return all users, you can focus on 1-2 million unique installations over the entire life of the project.

Further, everything is too individual for each project, and depends on the set of metrics (LTV, CPI, Retention). Here the field for experiments begins.

What is the principle of operation?

The principle of operation is as similar as possible to what advertisers use for the initial user acquisition. An advertising campaign is launched with certain creatives, but the main difference is that the campaign is launched not for all users, but only for those that we have allocated for it (this will be discussed later).

Audience highlighting works differently on different platforms. Some use custom lists with advertising ID or IDFA (on iOS, despite the lack of IDFA for a large number of users due to ATT Consent, there are still users with IDFA that can be highlighted), depending on the platform you are running the campaign on. Some have you completely hand over the information to the platform and set rules based on pre-defined criteria (usually days of absence from the game, completion of certain events, and income).

Next, you choose the principle by which the campaign will work:

1. Launch the campaign as a regular initial acquisition campaign and evaluate it as a regular campaign.
   I will not dwell on this method, since I have not encountered such advertising campaigns in practice. I believe that it is unpredictable in terms of payback, since we do not know how much the selected cohort would have brought in if we had not launched retargeting. For example, I have seen situations where additional costs led to the fact that on average we received less than if we had not launched retargeting.

2. Conversion Lift Test, A/B Test and other names that appear on different platforms.
   The essence of this method is simple: divide users into test and control groups. The test group is shown ads, and the control group is not. Then, for a fixed period (for example, two weeks), the campaign is monitored. Sometimes the test is stopped if the indicators are significantly worse than the targets set before the campaign launch. Some of the most basic target metrics are CPR (Cost per Re-engagement), CPSS (Cost per Session-Start), and ROAS. Once the test is complete, its evaluation begins, which we will discuss later.

What time to run Retargeting?

For retargeting to be successful, the user needs to be motivated to enter the game. And it is at the moment when he has this motivation that retargeting should be launched. There are four main periods suitable for launching retargeting:

1. Long events. You may notice that many games have various long-term events (Easter, Christmas, Halloween). It is suggested to launch retargeting campaigns on these dates, since new content appears for all users during these periods, and there is no need for developers to spend additional efforts on adding new mechanics. The mechanism is quite simple: a creative is created in the style of the event with a call to return to the game.

   

2. Short events. Their essence is the same as long events. The only difference is that they are launched for several days.

3. Deep link awards. This is the best approach for in-app retargeting campaigns because it allows you to have full control over how they are launched. The control is that you give out different rewards to different users and audiences. Proper analysis can help bring back a large number of users by offering each one a personalized set of rewards.

   

4. Adding completely new mechanics to the game. The principle is the same as with events, but the term is set based on the payback of the campaign. That is, when we understand that we have already returned all the users who are interested in this mechanics, and there is no point in trying to return anyone further.

Who to bring back?

This is the most difficult question, as there are many ways to understand it:

1. Number of days absent from the application. Obviously, if a user logged in yesterday, he is more likely to return today than a user who last logged in two weeks ago. Therefore, for the first user, the costs may be useless, but for the second, they are quite justified.

2. User region. Different countries generate different amounts of revenue in games, and mixing them into one audience can produce strange results.

3. Number of days in the game before leaving. Here the calculation is that the user could be loyal to the game if he played it, for example, for two weeks, but something distracted his attention, and he stopped playing. We remind him that he played our application.

4. Payer/defaulter. The approach is simple: if a user has already made purchases in the app, then the likelihood that he will pay again is higher than that of a user who has never paid.

How to evaluate?

In fact, the evaluation algorithm is very similar to the feature evaluation method in an A/B test, but slightly modernized for marketing needs.

1. We calculate the audience size in the test and control groups.

2. We count the number of users and their income.

3. We calculate the coefficient of audience division into test/control (k).

4. We calculate the increment of installations.

5. Installs Increment = N(test) – k * N(control).

6. There are then two methods of evaluation:

   1. Assume that users who returned from the test on their own have the same ARPU as the control, and consider their ARPU equal to the average ARPU(test).

   2. Consider that they have different ARPU and that in the test we attract higher quality users.

Example

During the allocated period, the number of users to whom we showed advertising was 200,000 people, and the number of users to whom we

left under control – 60,000 people.

The number of installations from the test group was 12,000 people, from the control group – 3,000 people.

• ARPU for d30 test = $1.

• ARPU for d30 control = $0.5.

• Spend = $1,000.

• k = 200,000 / 60,000 = 3.33 (the coefficient of division into groups is conditional).

• Incremental installs = 12,000 – 3.33 * 3,000 = 2,000.

Next, we calculate the income increment based on the installation increment:

This methodology is most often used to optimize high-level events such as launch and install (more precisely, re-engagement), since in this case the platform’s task comes down to practically one thing – to return the user.

Second methodology (similar input data and metric notations):

• k = 3.33.

• Revenue control = $1,500.

• Revenue test = $12,000.

• Incremental Revenue = $12,000 – 3.33 * $1,500 = $7,000.

• ROAS d30 Incremental = 7.

This method is most often used for value optimization to take into account the fact that the platform's goal is to attract high-value users, not just bring the user back to the game.

Conclusion

You need to know how to work with Retargeting, but you also need to understand that this is not a “Money” button. It is also worth understanding that it is not suitable for all projects.

If you want more about marketing analytics of mobile games, you can subscribe to my channel: https://t.me/analyst_in_jungle