Uncovering where users first heard about your app and how they came to download it is an age-old problem for app businesses. Tracking attribution is key to tapping into your best user acquisition (UA) sources, but with advertising spread across multiple channels and things like word-of-mouth hard to track, teams are often left in the dark.
‘How did you hear about us?’ (HDYHAU) questions seem to solve this problem, but are they really the solution?
Why use HDYHAU?
Many subscription apps face growing data discrepancies when it comes to tracking attribution. This is largely because iOS remains the primary source of revenue — but while these users typically have higher conversion rate (CVR) and stickiness, iOS offers limited attribution tracking. Structured frameworks for modelling attribution on iOS can help, but that’s only part of the equation.
To address these gaps, teams have opted for a simple way to measure the impact of ads: directly asking users “How did you hear about us?”. This question, commonly shown during onboarding or purchase flows, allows users to self-report their acquisition source. The answer options typically include:
- TikTok
- YouTube
- Friend
- Other
(There are some apps that randomize the order of the options to ensure answers from users are intentional, but in my experience, this doesn’t tend to impact the distribution of the answers.)
For early-stage teams running a small number of social channels (often Meta or TikTok) HDYHAU data has been a gamechanger; helping surface under-attribution in probabilistic systems and align performance with blended results. In these cases, self-reported data often appears more consistent than network or mobile measurement partner (MMP) reporting. However, as apps scale and expand into additional channels, HDYHAU responses become increasingly unreliable as an attribution source (more on this later).
Using HDYHAU for social ad attribution
The HDYHAU question fixes a common issue that all advertisers have with iOS: it gives the possibility of comparing like-to-like in a landscape where every attribution methodology works in a different way. If you use tools like Amplitude or Mixpanel, you can have graphics like this:
As you can see, this graphic is simply showing the answers for the HDYHAU question on a weekly basis, so if you take the campaign data, the MMP data, and compare them against this first-party data, you have a very simplistic way of seeing if your campaigns are underattributing or not.
According to my experience, when you use probabilistic methodologies like ADC for TikTok or AEM for Meta, you will likely track less conversions in your campaigns than have actually happened — leading you to make wrong paid UA decisions.
However, with HDYHAU, you quickly solve that problem by comparing both sources, so you can see if certain attribution channels are over- or under-attributed. That looks like this:
Since you (understandably) expect the user-reported data to be most accurate, marketers using HDYHAU as an attribution source typically place more importance on this first-party data than the MMP or ad network data. In most cases, they see greater impact on the blended results when they enable campaigns (like in the screenshot above, from a current project of mine).
Example: how HDYHAU impacts revenue
In my example above, the average difference between HDYHAU and the installs