Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Metrics we should know - or not! but we are gonna learn it!

Dinorah Tovar
November 11, 2021

Metrics we should know - or not! but we are gonna learn it!

This is a story about maths!
In this talk, we are gonna focus completely on metrics, from AB testing to analytics using Firebase and the importance that we need to address metrics when we start working on a project!
We are gonna talk about Bayesian and predictions models to track our AB testings and also we are gonna talk about Conversion vs. Retention vs. Count perusers

Bring your calculators! because this is a story about Firebase and Charts!

Dinorah Tovar

November 11, 2021
Tweet

More Decks by Dinorah Tovar

Other Decks in Science

Transcript

  1. Metrics we should know - or not! but we are

    gonna learn it! Dinorah Tovar Google Developer Expert - Android @ddinorahtovar @ddinorahtovar
  2. Disclaimer

  3. A story about data

  4. @ddinorahtovar

  5. Data gives the answer to the correct questions @ddinorahtovar All

    hit on the same place Planes on the research Result Research
  6. Analytics

  7. Data vs Metrics @ddinorahtovar •They are not the same •Data

    is generated by metrics •You don’t pick your data, you pick your metrics Data Metrics •Metrics are what you measure •Consistent •Cheap •Quick to collect
  8. Analytics @ddinorahtovar •Analytics is the systematic computational analysis of data

    or statistics •Provides insight on app usage and user engagement. •Helps you understand how your users behave
  9. Firebase has what you ned and more @ddinorahtovar

  10. How does an analytic looks in Firebase @ddinorahtovar firebaseAnalytics.logEvent("SomeNameOfTheEvent") {

    // Extra parameters if you need more info param(FirebaseAnalytics.Param.ITEM_ID, id) param(FirebaseAnalytics.Param.ITEM_NAME, name) param(FirebaseAnalytics.Param.CONTENT_TYPE, "image") } •On android
  11. How does an analytic looks in Firebase @ddinorahtovar Analytics.logEvent(AnalyticsEventSelectContent, parameters:

    [ AnalyticsParameterItemID: "id-\(title!)", AnalyticsParameterItemName: title!, AnalyticsParameterContentType: "cont", ]) •On iOS
  12. Conversion

  13. First, an example @ddinorahtovar app-ca app-email app-account 🏆

  14. Conversion @ddinorahtovar •Conversion analysis is the process of analyzing data

    related to conversions •A conversion is defined as a specific, desirable action that’s taken by a user •Depends on the analytics selected
  15. Conversion @ddinorahtovar •Conversion rate can be calculated over a math

    func coR = totalConversions/totalInteractions coR = 50/1000 coR = 5%
  16. Conversion @ddinorahtovar Conversion Event app-account Second Event app-email First Event

    
 app-ca
  17. Conversion @ddinorahtovar

  18. Conversion @ddinorahtovar •Conversions are related to multiple events •We decide

    which is the event we are interested
  19. Conversion @ddinorahtovar •Conversions are directly related to the num of

    users and the number of clicks on the conversion •But it also can be unique
  20. Conversion @ddinorahtovar •Conversions are directly related to time dimensions

  21. Retention

  22. Retention @ddinorahtovar •Retention analysis (or survival analysis) is the process

    of analyzing user metrics to understand how and why customers churn •Retention analysis is key to gain insights on how to maintain a profitable customer base by improving retention and new user acquisition rates
  23. Retention @ddinorahtovar Time - Survival probability 0 0.25 0.5 0.75

    1 Time (months) 0 2 3 5 7 9 10 12 60% of probability of surviving beyond 5 months - 25% of probability of surviving beyond 11 months
  24. Retention @ddinorahtovar •Firebase - has this chart - you don’t

    have to do anything just plug the dependencies in your app and you are ready to go
  25. AB Testing

  26. But first - what is an prediction model @ddinorahtovar •Assume

    the probabilities for both data and hypotheses(para meters specifying the distribution of the data) Bayesian Frequentist •Assume the observed data is sampled from some distribution •Frequentist vs Bayesian
  27. Bayesian Models @ddinorahtovar •How does it looks More overlap =

    Less con fi dence Less overlap = More con fi dence
  28. Bayesian Models @ddinorahtovar •How does it looks

  29. Metrics we should know - or not! but we are

    gonna learn it! Dinorah Tovar Google Developer Expert - Android @ddinorahtovar @ddinorahtovar