Tenjin DataVault customers now have a new training that can view multi-click attribution patterns for user acquisition channels. Recently developed for customers who requested us to help investigate potential Mintegral attribution claims, this became available as our latest training. Our clients care about Mintegral because, in the last 12 months, Mintegral rose to be a Top 3 channel for hyper-casual games.
Before we get started discussing this new DataVault training, we note that analysis falls into two categories:
- Diagnostic: the ones that can provide ‘yes’ or ‘no’ answers to a question with a high certainty level. For example, Tenjin Dashboard checks pro-actively for signals that are diagnostic of fraud. One example is that any click that happens within 1 second of an install is blocked as Fraud because a human can’t install an app that quickly!
- Non-diagnostic: these exploratory data analyses are useful for understanding what is happening and forming theories and trendlines. However, these types of analyses don’t offer explanations why something is happening. An example of this could be ROAS by country. You may be able to theorize why ROAS went up or down in a specific country, but many things could explain the results you’d see from one of these analyses that need to be tested diagnostically outside of the analysis.
The new “Multi-click claim analysis” training allows the marketer to see what channel's attributed installs has more than one click. Installs can often have multiple clicks from competing channels. We think this gives a fascinating insight into where clicks and their trends come from, but, at least in Mintegral’s fraud investigations, this training isn’t diagnostic of fraud (or not fraud). Recently, Mintegral released an explanation of their technology, which points to why similar results in this analysis can appear outside of fraud. It is up to you, the developer, and your partner, to discuss what these insights mean.
In the ‘Multi-click claim analysis,’ we take a look at four things:
- The time between last click and click assist
- Share of attributions that has one-click vs. multi-clicks by network
- Attributions for a network that had previous clicks from its network
- Attributions for a network that had previous clicks from other networks
The analysis’s idea was to determine any abnormalities among how many attributed installs have multiple clicks and where those previous clicks come from.
This analysis can give some insights into how users interact with ads and be used for ad revenue optimization purposes (ex: how many ads did they see, clicked on, and from which ad network were the ads). But, the interpretation of the data is open and an excellent basis to start a discussion.
We pride ourselves in providing data and analyses to our customers on-demand and help create more transparency. When we help developers analyze fraud, we remain transparent that several of these cases have come back with multiple interpretations, so be cautious about what you hear and make sure to talk to your partners.
If you are a DataVault customer on an Advanced or Pro Plan, please reach out to email@example.com to receive Training # 427 Advanced - Multi-click claim analysis of your marketing channels.
Special thanks go to Mei-Han Weng for creating this Training.