How to deal with unhealthy product data

What is unhealthy data?

Compare the calculated indicator data with industry data (which can also be expected data or past data). If it is low, it means that the data is bad data.

Example 1: If the frequency of ad impressions per capita on that day is low

**Thinking: **

1. Find the missing link in the funnel
2. After finding the lost links, conduct comparative analysis by subdividing the dimensions.
3. Find possible reasons based on the results of page loss and comparison with segmentation dimensions.
4. Formulate a hypothesis

1. Funnel process:

1. Number of function triggers

• Number of active use functions
• Number of times the function has been used through guidance
• Number of triggers of guidance
• Number of clicks on the lead

2. Scanning
3. Scan results page

• Conversion rate (= scanning results/scanning)

4. Click Clean

• Click rate (= click clean/scan results)

5. Cleaning up
6. Cleaning results

• Conversion rate (= Cleaning/Cleaning results)

2. Dimension comparison

Find out the lost links and compare them by dimension. This can help find the cause.

Division of dimensions:

• National dimension: The culture of different countries may have different acceptance of the interface.
• Mobile phone model: Maybe high-end mobile phones don’t have many lags, so they don’t need to clean up and speed up.
• System version: Same as above
• New and old users: New and old users have different logical understanding and acceptance of functions and advertising triggers. It can be divided into: new users, retained users the next day, and all old users.
• Advertising slots: Different functions may have different needs for users
• Traffic channels: Different channels have different quality.

If retention is low

1. Analysis based on the funnel process from installation to use
2. Find the links in the funnel process where churn is high
3. Check the events generated by users who have uninstalled. If there are no events generated by the next link with high churn, it means that uninstallation may be caused by churn in this funnel.

1. Alternate possible reasons

-Demand

-Failed to meet user needs

• Users cannot perceive the effect
• The function is fake and it is still stuck after cleaning
  • The effect of cleaning and display is too weak and does not give users a strong perception

-Basic needs are not met

• Basic cleaning functions are not done
• The user does not have this requirement

-The redirected users do not need to clean up

-Experience

• Inconvenient to use
• The use process is cumbersome (the funnel can be used for evidence collection)
• The scanning time is too long (the page duration can be verified)
• Annoying users with useless information
• Push notifications are disturbing users. Users do not need prompts at all or prompts are inaccurate.
• Too many ads are triggered, which makes users sick
• Performance

-Product bugs that prevent normal use (evidence available)

• Application survival is low (evidence can be obtained)