You are the PM of Friends on Facebook. The % of friending has dropped by 5% over the week. What happened?

Clarify

• Total friend requests vs friend requests accepted – its the number of requests sent

• In the US

• On all platforms

• Gradually dropped throughout the week

Check for Internal factors:

• Any decrease in DAU as well?

• Bug? A/B test?

• Change in friend recommendation engine?

• System slowness?

Check for External Factors

• Competitor grabbing major market share?

• Bad PR or a ban?

User Journey

Users can add friends in different ways

• User add friends through friend suggestions

• New users can import their phone/email contacts to search friends

• User searches someone on facebook and add friend

Hypothesis:

Taking cues from user journey lets generate some hypothesis

1. Users not providing access to contacts due to privacy issues or change in platform settings or some email provider (gmail) blocking this feature

• Check with the relevant business teams if they have any info

• # of new users (signed up within 30 days) importing contacts (check for decrease)

2. Friend suggestions are not relevant or less notifications

   • # of friend suggestion notifications per user per week (Check for a drop– a drop might indicate less notification. Check for both app and email notifications)

   • % of friend suggestions accepted by users - This indicates the relevance of friend suggestions. Any drop will indicate that the recommendations arent relevant and need to improve

3. New privacy setting which stops other users from adding them as friend (or an extra step added)

   • Check if anything has changed

   • # of people searches per user per week (Should not be affected if hypothesis is true)

   • # of friends added after search per user per week (Any drop might indicate that users are trying to add friends but not being able to)

Segment:

Lets also try to segment user based on behavior and see if we can generate any hypothesis

Users can be divided into 

• new (signed within 7 days) and 

• regular users. 

Most of the users add majority of friends when they are new and then the rate slows down

Possible that new user arent adding that many friends. Possibly lot of fake accounts created in last week

• # of new users signed up per week (possibly wouldn't drop otherwise the issue is acquiring new users)

• # of friends added in first 7 days (a drop might indicate a change in overall new user population signing up, possibly rise in fake accounts)

• # of reported new users every week (A rise might indicate fake accounts)

Clarifying questions:

1. When you say % of friending has dropped by 5% over the week, what is the exact metric that has dropped by 5%

• Is it the % of engagement adding new friends
• Is it % of engagement accepting new friend requests
• Is it the DAU that went down
• Is it the Customer acquisition metrics

2.  Assuming the interviewer  says % of adding new friends went down:

     Is it happening on facebook across all regions or is it specific to certain region?

3. Assuming it is for all regions:

    We need to investigate internal and external metrics in the workflows of adding a friend.  Let us go through the workflows of adding a friend:

 #1.  User searches for a name or an email of a friend -> Finds the friend  ->  Clicks on the Add Friend button -> Friend request sent appears

#2.  Same as workflow 1, excepting that user clicks on the profile of the user -> Views the profile -> Clicks on Add Friend -> Friend request sent appears

#3.  User looks up mutual friends of a friend -> follows add friend workflows in either of the above #1 or #2 

#4.  User looks up suggested friends by facebook -> follows add friend workflows in either of the above #1 or #2 

Below are my hypotheses/possible issues:

1. If it is adding a friend:  

• Maybe there has been a new update to add friend that might have changed the UX of the add friend button which is not that intuitive for users to add a friend.
• Check if any platform issue like an API change has been accidentally pushed that might be causing this.  See if it is an intermittent issue happening under some condition.  

2. If it is Searching and finding a friend:

• Maybe there was a change in the search platform that was shipped which is making a friend search ineffective.
• Maybe the search is too slow and users are just navigating away before the results come up.  Look for the search platform's metrics like availability, usage metrics and other performance metrics.
• Any possible design change or placement of the search field that is not intuitive for the users to look for friends.

3.  If it is looking for mutual friends:

•  It is possible that the mutual friends is not showing up results for users who have not even set the priivacy to hide mutual friends.  Look for any breakages in this piece.
•  Same for search results while looking up mutual friends. Look for any breakages in search results as mentioned in the #2 hypothesis.

4.  See if suggested friends is working accurately.  Check for internal metrics that determine the effectiveness of suggested friends and see if it has gone down.  Maybe there was some change in the algorithm that is recommending friends the users are not interested in.

5. If it is looking up a profile of a friend, probably the profile loading needs to be looked into. Maybe the loading of the profile is slow.

The above steps can be taken by a PM if it is something to do with a breakage based on the above assumptions.

A few clarifying questions

• What does % friending mean here? Does this mean % of DAU who clicked the "Add Friend" button? (Assuming interviewer says "Yes").
• Could we assume that this decrease is isolated? (This is happening because of another incident ) For example, a decrease in new user signups might have had an impact on the use of the "Add Friend" feature, as new users often add friends at a much higher rate than returning users. Also, a decrease in DAU might result in a decrease in % Friending (Assuming that the interviewer says that this decrease is indeed isolated, and not due to the decrease in other parameters.)
• Could we also assume that this decrease is not a result of periodic trends? (Assuming that the interviewer says that this is indeed not due to periodic behavior)
• We also assume that this metric is shown on a dashboard that is updated periodically. (Interviewer agrees)
• Could we assume that this is for a particular region and not multiple regions? (Interviewer agrees)

• There might be an error in how the metrics have been calculated, leading to the wrong result.
• Changing the location/design of the "Add Friend" button might have caused a decrease in its use. The new design might have made it less intuitive for individuals to use.
• Introduction of a new feature on the website; this might have had a cannibalistic effect on "Add Friend" Button.
• Change in website design that might have decreased the load time for the "Add Friend" button feature
• Given that this metric is shown on a dashboard, there could be a change in data collection procedures. Has there been a change in the schema of the database that stores Facebook interaction data? Are we using the right variables to calculate this metric?