Quantitative data analysis
We are used to quantitative analysis. We make statements like “Engagement for April cohort for iPhone 6S was 5½ minutes per user.” Is that good? Yes. No. Maybe?
To drive engagement, for example, an app publisher may want to first identify users with low engagement and take necessary action. Engagement for an app that serves as a an interface to a search engine, would be very different from a creative app, for example, an app that lets you edit and trim videos. You would expect people to be in and out of the search app but visit it many times during the duration of a day. On the other hand, people would spend considerable amount of time in a creative app but launch it a lot fewer times during a week.
Qualitative data explorations
Most marketeers do not want to be bothered by quantitative engagement (8¾ seconds of engagement may be great on one app and very low on another). They are mostly interested in folks with different levels of engagement, for example low, medium and high, so they can take appropriate actions.
Data scientists on the other hand care about centers of each dimension for each cluster and how it changes over time. We will discuss how Pyze helps data scientists in another post, at another time. This post focuses on bringing the power to marketers.
For Customer Satisfaction attitudinal data, net promoter score was calculated as a difference of percent of people who gave a score of 9 or higher (promoters), and percent of people who gave a score of 6 or lower (detractors) on a 10 point scale, to the likelihood to recommend question.
Mobile popularized the two point scale: thumbs up and thumbs down. Mobile- and e- commerce brought the 5 – star rating scale for products, services and apps
A typical good ol’ 5-level Likert scale works best for qualitative intelligence for certain types of clustering algorithms where predictability in clustering levels is far more important. For engagement, the qualitative scale could be:
- Lowest Engagement
- Low Engagement
- Medium Engagement
- High Engagement
- Highest Engagement
Actionable Qualitative Intelligence
Lets say an app has 12 million active users. Marketers can explore Low and Lowest engagement across all users of the app, combine that with users who have low revenue and are of high attrition risk, and do something instantly with this “target set” of users, like send them a promotion.
The selected path in Intelligence Explorer represents the above venn diagram. “Low+Lowest Engagement, Low revenue and High attrition risk”
Instant Explorations and Automation
Instant explorations allows marketers to maintain a competitive edge. They can make 100’s of different explorations in a few minutes till they find a targetable user set that lines up with their strategy, using Intelligence Explorer. Once they are satisfied with the results they can put this strategy on autopilot using Growth Automation™, where a set of intelligence bots work to keep your user base automatically growing.