What A/B tests do you suggest to make it easier to find friends on Facebook?

ok so by finding friends we mean the engine be able to automatically recommend you friends you may want to be freinds with or the feature where you type a person's name because you want to be friends with that person and then finding that person easily in the search results? (the later).

ok, so basically we want to decrease the time it takes for people to find new friends so that it would help increase the # of friends. (correct). And to my understanding, increasing friend circle on fb is directly proportional to higher engagement , better user experience and higher business impact. (you can say so, yes).

So I want to approach this by thinking about how one may find friend on fb. Then I can think of few pain points as a result of present user journey and how to potentially resolve those by implementing minor quick tweaks. Does that work? (sure)

1)I navigate to fb 2)in the search bar I type firends name and I hit search 3)bunch of people with that name comes up 4)I have to scroll through mayn papges and i click on a profile i think might be that person 5)So I now read that profile hoping that that person has public profile or some information about thier profile so I can decipher 6)If it doesnt seem like the same person I am looking for then I scroll through more profiles and repeat steps 4 to 6. 7)ultimately I request friendship and await for approval from the other person.

Here I see the pain point is finding the right person from long list of results. We should prob improve that user journey. I often feel if only I could type the city or even the workplace to narrow down the search. So couple things that come to mind:

1) Allow for more fields to type information such as city of the person, hometown, occupation, work place, marital status, etc. to filter down the results.

2) Currenty the serach results are in list mode. This is painful to go through since sometimes we are looking to match a picture becase we met that person at a festival. It's rare I think that we are looking to become friends with someone we don't even know what he/she looks like. And listing results in a long list as is currently is not a great user experience. What if we were to list them as thumbnails in 1x1 in tiles. WE could fit more profile pictures in a page and it be easier# and faster for me to skim through. Wehn you point at thumbnail the title expands a bit and you can then go next / previous on various pictures to try and identify if that's the right person you want to be freinds with.

Third idea is around implementing AI logic but that would be more intensive and it's likely FB already does that. Given that, let me evaluate which one I may go with and share my thoughts on how we may be able to conduct A/B test. (sure).

I think we should go with #1 because with #2 I would still have to skim through 10-20 different pictures. But filing information helps narrow that 10-20 to prob 5 or maybe even 1. So it will improve efficiency and accuracy for our user. So assuming we implement this, we can test this out by having a control group whereby their experience hasnt changed and another group whose expeirence has changed to this enhanced capability. Both the user groups would be equal type of demographics so that we can compare apples to apples. I would prob run this test for about 2 weeks; enough data in 2 weeks.

What I would measure is:

1) Avg. time taken per user in finding a specific person to be-freind

2)% of new friends on FB (hypothesis is that less time taken in finding a person to be friends with will help increase # of friends which is good for user expereice and business).

One thing that this doesn't solve for what if I don't know much about that person. That is opossibly because it;s likely I never talked about this type of stuff when I met him/her at the party or festival. Or what if I am trying to find this girl someone mentioned but none of us know much about her. So yes this solution doesn't resolve those needs. I can certainly think through that if we want to solve that as well. (no it's fine. This is good. Thank you much).

I always believe that to come up with apt and reasonable tests and decide on how to improvise the situation and achieve a goal, One should first of all understand the product and the objective.

So, I would like to establish the background info related to the product and the goal.

Facebook is a Social Media product owned by Facebook Inc. The goal of the product is help bring the world together and help its users form meaningful connections. Adding a user as a friend is a way of connecting with them and through this connection, one can interact with their post, text and call them. Thus Adding a person as a friend is the basic step that the firm wants its users to take so that the product can move towards the goal and achieve it. Facebook allows its users to find friends in 3 different ways:

1. Search for a user
2. Friend Requests that one gets
3. Finding Friends from the recommendations.

Of this, I would like to focus on the 3rd touch point as that is where Facebook can continuously improve and help users find new people more and more easily depending upon the information shared and their past interactions.

So, Next I would like to establish the goal that we have for the test. The Goal is to make it easier to find new people and add them as their friend.

With the above clarification, I would want to list a few assumptions and as well as clarify a few things before brainstorming a few tests that are  possible. 

1.   I would like to focus particularly on Mobile Devices as people are using phones more and more these days. Okay. Its totally fine to focus on the same.
2. Are we focussing on any particular user type or demogrpahy? None as of  now. However, if you want you can do so.

1. Phrasing the "People you may know" as "Find your future Connection"
   1. Hypothesis: changing phrases as per the trend
   2. Experiment: 
      1. Control  Group: A group of people finding their new friends from the section phrased as "People you may know:
      2. Experiment Group: A group similar to Control group finds the phrasing as "FInd your Future Connection".
   3. Success Metrics:
      1. # of Friends added WoW from "People you  may know"
      2. Total number of friend requests sent from each  group
   4. Trade off
      1. The trend might change very soon. So changing everytime will affect the user's comfortability
2. Listing the commonness between suggested friend and the user
   1. Hypothesis: Showing the common factors like same previous school, work will improve people becoming friends
   2. Experiment: 
      1. Control  Group: A group of people finding their new friends from the section "People you may know" with same info, i.e, Profile Pic, Dp,Mutual friends Count 
      2. Experiment Group: A group similar to Control group finds information such as Common School/College or Prev workplace along with already shown information
   3. Success Metrics:
      1. # of Friends added WoW from "People you may know"
      2. Total number of friend requests sent from each  group
   4. Trade off
      1. The screen realestate occupied increases and might make the user feel uncomfortable
3. Moving the placement of the People You may know from middle to the top.
   1. Hypothesis: Seeing the suggestion might increase in the CTR of adding new  friends
   2. Experiment: 
      1. Control  Group: A group of people finding their new friends from the section "People you may know" At same place-Somewhere in the middle.
      2. Experiment Group: A group similar to Control group finding their new friends from the section "People you may know" at the top.
   3. Success Metrics:
      1. # of Friends added WoW from "People you may know"
      2. Total number of friend requests sent from each  group.
      3. CTR of More Button
   4. Trade off
      1. The user might get disturbed by the  placement as  they would want to see some post, instead would find some suggestion

So, from the above prioritization, I w ould like to run the 2nd idea as my A/B test. 

I'd take the following approach to this Facebook AB testing question.

Would you please let me know if you think the AB tests I suggested are good tests for an AB environment? I understand that certain ones feel a little bigger than others (ex. suggested friend v. language tweak). 

____

1. CLARIFY: 
   1. Do these tests need to be channel specific? Assume no.
   2. Is there a specific user we must focus tests on? Assume no.
   3. Why is Facebook doing these tests now? Assuming they want to continue to build connections as per their company mission.
   4. Are there any limitations for testing, ex. budget / tech? Assume no.
2. DESCRIBE FACEBOOK / FRIENDING: Facebook is a social media platform that allows users to create profiles and "friend" people - i.e. build connections. You can comment / like posts to enhance connections. Other features of Facebook enhance the sense of community - ex. creating events, wishing happy birthday, Facebook Watch etc. Facebook allows users to find friends 3 ways:
   1. Search: User types name into search bar
   2. Find Friends: User sees recommended friends based on contact information, Instagram friends, etc.
   3. Friend Requests: User receives friend requests from others.
3. GOAL: The goal for Facebook is to get users to add more friends / build more connections. For the purpose of AB testing, I'm going to focus on fleshing out the "Find Friends" recommendation aspect of friending to make it easier for the user to find suggested friends. 
4. USER PROFILES: Based on the groups below, I will focus on the Moderate User, as there should be sufficient data to surface recommandations for friends and they are not as active at engaging in the platform to send friend requests. 
   1. New User: Does not have many friends. Has not engaged in platform too much. Not much information on user.
   2. Moderate User: Has established set of friends but not actively friending many people. 
   3. Power User: Actively friends people all the time. Has a large number of friends. 
5. AB TESTING IDEAS: 
   1. Find Friends Widget: Move the "Who you may know" / "Find Friends" widget to home page. 
      1. Hypothesis: Users who see widget to top of home page v. within the friend section (where it is currently) will add more friends.
   2. Notify to Friend: Notify users "Hey, you haven't added a new friend in a while. Why don't you send a friend request?"
      1. Hypothesis: Prompting the user to send a friend request will get them to friend someone. 
   3. Similar Interests: Notify users that Facebook sees they have similar interests (based on likes) with XYZ people and that they should friend them.
      1. Hypothesis: Surfacing potential friends with similar interests will increases friend requests. 
   4. Similar Events: Notify users they have attended 3+ events with XYZ people and that they may know them. 
      1. Hypothesis: Notification will increase more friend requests.
   5. Tweak Language: Change the language from "Find Friends" to something like "Make a Connection" on feature.
      1. Hypothesis: Language change will increase number of friend requests. 
6. PRIORITIZING TESTS: 

   1. Test	Impact to User	Cost to Facebook
      Find Friends Widget	Medium	Medium
      Notify to Friend	Medium	Low
      Similar Interests	High	Medium
      Simliar Events	High	Medium
      Tweak Language	Low	Low

   2. Given the prioritization, I will focus on "Similar Events" AB test, as it has the highest impact to the user and the cost, while higher than some other tests, should not be too substantial. Similar Interests and Similar Events are similar, but I assume that a person is more likely to attend the same events as someone they know v. have similar interests (so suggested connections will more likely be friends).  
   3. NOTE: Given COVID, it's like that a user is not going to events as often as they were prior to the Pandemic. I'd ask the interviewer if COVID should be a factor. If it should be, I would have changed my test choice to "Similar Interests" and finished the response accordingly, as I believe it is still a viable test option. (Would need to clarify when this test would be run.) I'm running the "Similar Events" test assuming we are in the more "post Pandemic" phase where events are happening more BAU. 
7. DESCRIBE SIMILAR EVENTS TEST:
   1. Control Group / A: No change to current Facebook experience.
   2. Test Group / B: User will be notified if they have RSVP "yes" to 3+ events with another person. Friend prompt will appear. Prompt could say, "We've noticed you and X person have attended 3 events together. Do you know them?" with a "Add friend" button below.
   3. User Profile: Moderate User - Has established set of friends (ex. 300 - 700 friends) but not actively friending many people lately. Potentially has not added a friend in 6 months. Must have logged into Facebook within last month to show some sort of active engagement with platform. 
   4. Channel: We could try running on both desktop and mobile. If we had to choose one, I'd choose the channel where moderate users are more likely to log into Facebook. I guess that would be mobile but would need data to confirm. 
   5. Market: If limited to a certain market, I'd choose a market where there is sufficient numbers of actively logged in moderate users to make statistical significance. You could further narrow the location to a specific region, etc. I'd also need to ensure there are sufficient events happening in the market. (Per my note on COVID above, certain countries may have more events than others depending on the time this test is running. Noted assumption we are in more of a post Pandemic phase when the test is run.)
   6. Time Frame: Given that a user may not be going to events that often, potentially could run the test for 6 months to ensure the user has gone to at least 3 events. (Events should be relatively recent within the last year.)
8. SUCCESS METRICS:
   1. Number of users who add the suggested friend from Test Group v. number of users who added friends as they would normally
   2. Total number of friend requests sent in Test Group v. Control Group
   3. Number of friend requests / user in Test Group v. Control Group
      Comparing the metrics would allow us to see if the hypothesis was successful. If I had to choose 1 metric, I'd choose #3, as it most directly gets at the hypothesis. (It is similar to metric #1.)
9. TRADE OFFS:
   1. Users may feel Facebook is invading their privacy with enhanced ability to know their friends. 
   2. User may see random suggestions v. actual people they know. 3+ events may not be enough to ensure that a user knows someone. (Broader questions if Facebook wants to connect strangers who may have similar interests - i.e. help develop friendships or even romantic connections, which we know they've already ventured into with Facebook Dating, v. connect people who genuinely know each other already.)
   3. RSVP "Yes" may not mean the user has actually gone to the event. RSVP may not be enough to track event attendance. 
   4. It will take time to find users who have gone to the same events given that they are not happening everyday. 
10. SUMMARY: I'd run an AB test that notifies users whether they've RSVP "Yes" to 3+ events with another user and notify them they may know each other. I hypothesize that users who see this notification will make more friend requests than the control group with no change to the current Facebook experience.