If you operate a subscription product on mobile, SKAdNetwork has really torpedoed your measurement capabilities on iOS. While previously, you could look at early trial conversion data and then review performance on a cohort once the trial period was up to evaluate the number of new subscribers and revenue that was generated, you are now completely missing real revenue data by cohort. You don’t know what campaigns are driving lots of subscriptions and which are driving trials that result in no revenue. There is no perfect solution and you are not going to get back to your previous precision of measurement, but with some smart configuration, you can get much closer.
Trial Conversion in the First 24 hours
Before getting into the meat of the issue, it’s important to get an understanding of how many of your trials are actually being caught by SKAN. In other words, look back at cohorts from 30, 60, and 90 days ago and calculate the percent of trials that were triggered within the first 24 hours. If you are pushing your trial early in the funnel, this number should be 90%+, however there may be exceptions. Second, try cutting this data by any dimensions that you think might affect the speed at which a user might complete the conversion. Some examples might be region, user age, or device type. If there are meaningful differences in the rate at which you’re capturing trials in SKAN for different groups of users, it’s important to know that upfront, so you can consider how this affects the accuracy of measurement for different campaigns.
Trial to Subscription Conversion Rate Correlations
Once you have a sense for how reliable and consistent your trial conversion measurement will be in the first 24 hours, it’s time to start thinking about how to estimate your conversion to a paid subscription. If you’ve been measuring and optimizing your marketing campaigns based on the assumption that all cohorts convert at the same rate once they’ve started the trial, you are likely allocating budget inefficiently.
Look at the first 24 hours of your user journey, both before and after the trial, and perform a correlation analysis between event completions and conversion to a paid subscription. You’re looking for the events with the strongest correlation to the eventual paid subscription conversion so you can tag these events in SKAN and better estimate each cohort’s projected revenue value.
Don’t Be Afraid to Ask Questions
Are you gathering any demographic data from users by asking them questions about their age, gender, household income, or other potential indicators of likelihood to pay? We are in a new world of privacy restrictions in advertising, but that doesn’t mean customers won’t willingly give you information about themselves if you ask. If you aren’t already, try adding one question into the early flow of your onboarding experience in an AB test. You will likely find that you get a very high response rate with little to no impact on retention. If that is successful, try adding another. You can then analyze to see if specific groups of users have meaningfully different payer conversion rates.These can then become events in your SKAN conversion values.
Build ROAS Projection From Your SKAN Events
You should now have events triggering that tell you how many trial conversions you have per cohort as well as one or more additional events correlated with subscription conversion. The more events you have, the more granular and precise your ROAS measurement will be. Now, you can use whatever tool you use for product analytics to segment cohorts based on these events, and measure the average revenue per user for each group. This can be done on whatever time window makes sense for your scale of spend. If you have significant spend, you could do daily, otherwise weekly could make sense. Also, as long as you are running each marketing campaign in a single country, you can segment by country (or region if you do some grouping). Once you have average revenue calculated, you can map these to the counts of conversion values that you are receiving from SKAN for each campaign. Combined with your cost data, this will allow you to calculate your ROAS per campaign.
New Considerations with SKAN 4.0
- Once you have your fine grain conversion values configured, look at your average value per user for these values and determine how you want to map your low, medium, and high coarse grain conversion values
- Your 2nd and 3rd postbacks should now be able to contain your actual subscription conversion event (depending on your trial window). This will provide slower feedback both to you and to your ad network, but will be a good check against your estimate from the first postback
- I would recommend setting a lockwindow for your first postback right after 24 hours to ensure the timeliest data, unless there is a key indicator of subscription conversion that frequently happens just after 24 hours
Every business is different, and there is no one size fits all implementation, but hopefully this provides a good starting point to work from. Good luck with improving your granularity and accuracy of measurement.