On April 26, Apple released iOS 14.5. In the new version of iOS, requirements for the privacy of user data have been seriously increased. Now developers can use the attribution system developed by Apple — SKAdNetwork.
SKAdNetwork is a serious compromise between concern for personal data and care for the needs of developers. On the one hand, it is important to allow users not to tell too much about themselves and their in-app behavior. On the other hand, it is important to allow owners of games and apps to evaluate the effectiveness of advertising campaigns. AppMetrica helps developers interpret Conversion Value model values.
What is Conversion Value
The Conversion Value is an integer between 0 and 63 that helps app developers gauge the effectiveness of attracting traffic by user behavior. It followed the IDFA (The Identifier for Advertisers, a tool used in previous versions of iOS to associate a user with the ad from which they clicked) that has ended.
Our SDK can increment this number until it reaches a value of 63, or until it can change. Apple then submits the number to the ad system using SKAdNetwork.
A number between 0 and 63 is the only piece of data that Apple shares. This approach to sharing user activity data with us is a big limitation, but it does provide important information for the advertiser.
Apple transfers a number from 0 to 63, because exactly this range of numbers fits in a segment of 6 bits long.
Important Conversion Value features
- The Conversion Value is increased if the user has performed the target action.
- If the user has not performed the target action in 24 hours, the measurement stops and Apple stores the number value previously sent via SKAdNetwork.
- Within 24 hours of the end of the action chain, Apple returns this information in an anonymized form.
In the first case, the user performed 41 conversion actions on the first day and none on the second day, and AppMetrica got a value of 41. In the second case, the user scored 63 actions in three days, and the counter stopped. And in the third case, he consistently gained one action per day, and after a few days AppMetrica automatically stops measuring, so as not to wait for the result for another two months.
Let's see how Conversion Value works.
Imagine a mobile game called iDumplings. In it the user quickly taps on dumplings, scoring points. To get to the next level, you need to score twice as many points as the previous day.
The developer of iDumplings advertises his app on Yandex and wants to know what level users reach in the app. The target action is to fulfill a game condition, to go to the next level.
User Vasily clicks on the ad, downloads the game and runs it on his iPhone.
On the first day Vasil gets 6 points. Conversion Value gets a value of 1. The next day Vasya clicks on 22 dumplings, which is more than enough to get level 2 of the game. Conversion Value gets a value of 2. On the third day, he clicks on 50 dumplings: +1 to the level and +1 to the Conversion Value (3 in total).
On the fourth day, Vasily's son was born, and he had no time for dumplings. He didn't play the game, the conversion action didn't happen. The timer stopped, the application sent Conversion Value - 3 to Apple. Apple, using SKAdNetwork, sent the value to Yandex Mobile App Ads.
Over time, statistics on Conversion Value and other users were gathered, and the developer saw that on average, players who came from the source ads reached level 4.
You will see the data only at the end of the chain
The Conversion Value feature is that conversion data will reach the advertising system via SKAdNetwork only after the chain of actions is interrupted. This is important to consider for advertisers who have a lot of actions in their app.
For example, if users swipe potential partners by the dozens per day in the app, you shouldn't set the Conversion Value to 5 photos. This will cause the Conversion Value to be reached almost always, and if the user likes 400 photos at once, the goal will be reached immediately.
It is important to estimate in advance how many target actions app users take, so as to set the step for CV by number of events roughly similar to the average values.
If a user in your app reads, for example, two articles a day, and you want to pass "How many articles the user has read", the CV value could accumulate for about 30 days until it reaches the maximum. Thankfully, AppMetica has an option to force you to interrupt the Conversion Value chain in such a case. For example, you can find out how many articles the user has read in two days. This is also important when you change your conversion strategy.
The Conversion Value might have a situation when the value is reached in 1 or 2 days. (For example, if we set 1 photo and the user "likes" all of them). In such a case the Conversion Value metric will "pick up" all users, and it will be extremely difficult to identify the value of a useful traffic source.
Tip. Choose an average value for your application as your conversion step size.
Measuring conversion value
AppMetrica offers several options for measuring conversion value. You can select them directly in the AppMetrica interface.
Conversions. This model allows you to find out which actions the user has and has not taken. You can select up to 6 different events, and for each of them the report will show the percentage and number of users who have reached it.
You can use both regular and revenue and e-commerce events in each of the 6 slots.
To enable event value measurement, go to settings from the User Acquisition report, under Bundle ID select the applications you are interested in, and under "what to measure" select up to 6 events.
Note that you can "mix" events (for example, take three from events, two from e-commerce and one from revenue). Moreover, you can specify a measurement period to make the Conversion Value chain shorter — and thus draw conclusions about the effectiveness of the advertising campaign sooner. For example, if you're optimizing traffic for in-app registrations.
Number of Events. Find out how often a user performed the event you're interested in. The report will show the total number of events and the average per user.
You can track their exact number up to 63 times, or you can increase the maximum by changing the step size. Eventually AppMetrica will convert this into understandable data.
To activate value measurement by event quantity, go to the "Event Quantity" tab and select the event whose quantity you will measure in the "Event to Count" section.
Then set the step size. We recommend that you set the step size greater than 1 for frequent events. To estimate the step size, you can see the average number of events per user in the event report. Don't forget to specify the measurement period. For frequent events, it is better to specify a measurement period of a few days, to avoid a long data chain for users who do it less often than usual.
Revenue. Estimate the amount of purchases, ARPU, number and share of paying users (based on revenue or e-commerce events).
Get the revenue per user from purchases. The report shows revenue, number and share of paying users, ARPU. Revenue in all currencies is converted to the selected unit, the maximum possible value is 63 units.
For example, if 1 unit of measurement is $0.5, then the minimum fixed income is $0.5, the next one is $1, and the maximum for the period is $31.5. Transactions made during the measurement period are summed up.
To activate value measurement by revenue, go to the "revenue" tab and specify the revenue event type or e-commerce purchase. (the first for in-app purchases, the second for e-commerce). Specify the currency and measurement period. To estimate the unit of measure, you can check the Average Revenue Per User (ARPPU) in revenue reports for in-app purchases or in the Purchase Analysis report for e-commerce.
The step size
The step size determines the number of events until the next step, by default it is 1. It can be increased when measuring events that the user performs in large numbers.
For example, the target action in your app is viewing a photo. It's quite frequent, in the first session users view about 30 pieces on average, and in the first two days — several hundred times.
But the maximum conversion value is only 63, which means it will be the same for most if you count it accurately. To get around this limitation, you can count not every event, but, for example, every 5. Then the maximum will be tracked 315 actions (the first five photos viewed — value 1, then another five — value 2, and so on).
And if you measure revenue from in-app purchases, the average revenue per user in the first three days ranges from 50 cents to (in rare cases) a few dollars. To get as close to the reality of the calculation as possible, set for 1 unit $0.5. Then for each user the income will be calculated as follows: $0.5, then $1 and so on up to $31.5.
Choosing an application
You need to apply a separate Conversion value measurement scheme for each application.
If you see the warning "A different conversion scheme is attached", it means that your application also uses a different APIKey on which the measurement is already set up. The same conversion value measurement scheme will work in the one application.
Taking care of users' personal data often gives app developers and marketers a headache. But with AppMetrica, you can literally "squeeze everything" out of your current attribution model.
We've hidden complicated math and tricky models from you "under the hood". Just advertise apps, and we'll take care of the rest!
The AppMetrica team