Interviewer : Friend requests are down 10%, what would you do?

Me : Ok, before we start drilling down, I just want to clarify a couple of things here

1. When we say Friend requests are down – we are strictly looking at the supply side ( a user initating a request) and not the complete funnel – User initating and the other user accepting

I : Yes, that is correct

2. Ok, and when we say friend requests are down 10% – are we saying users initiating friend requests is down 10 % or the average friend requests that users receive is down 10%?

I : The former – users initating the friend requests.

Ok – thank you. The way I would like to go about answering this question is by asking drilling down deeper into what may be causing the problem, by looking at the data. Can I assume that you as the interviewer will be able to provide answers to my data question?

I : Says yes, can assume that.

Great, thank you.

First – lets eliminate the most obvious question. Is this 10% decrease sudden or gradual?

I : Sudden.

Ok – Is this geographic or global?

I: Specific to North America

Assumes no code changes have happened in friends.

I : Sure, that sounds good to me.

Ok, so its Sudden and Global, and nothing major changed on friend management side. Give me a couple of minutes to brainstorm here

Right, am ready. The first thing that comes to my mind is – what if this issue is not related to friends app – but facebook as a platform overall? So, first lets look at

Has there been an overall decrease in DAU/WAU/MAU wrt facebook overall?

I : No – nothing significant.

Ok that indicates that the users are still coming into the platform, yet somehow are not hooked into adding friends.

• Did any of the entry points change ( assuming current state) Are we able to search and add friends Are we showing suggested friends in the application under friends hyperlink Are we showing add friends in the main app between posts

Answer is no to all

• Is there any difference between native app vs browser usage metrics
• iOS vs Android DAU login metrics – No , but Android seems to be gradually dipping
• Any connectivity issues ( assuming 4g vs 5g rollout and the 5g rollout having issues for a specific provider)
• Customer segment change
• Time spent per customer segment – Yes, We see a dip in that. and Geographically consistent to North America, but not specific to one customer segment

OK – i would like to parking lot that along with friends algorithm issue above.

Now, going back to the problem We have a sudden 10% dip in North America in the act of a user Adding friend(s), and we have established users are coming into facebook and we are not seeing anything that is a major red flag that will point to an obvious issue here.

Ok – Lets look if we had any major launches with the fb app itself and say to North America specifically. No

• Ok – Could there be a hardware issue here – have we had longer retrieval times, is the news feed loading slower, are our servers running optimized? Yes
• What about the user interface – Nothing in the facebook app.

Ok – so far we have eliminated – change to add friends feature; changes in fb ecosystem overall, geographic, hardware and UI. That leaves us with competitors.

Is there anything major that was launched by any of our competitors?

No, but TikTok recently ran a major marketing campaign for Daddy Daughter dance for mental health during Covid times and it was a campaign specific to North America and aimed at 16- 24 year olds.

Ok – thats interesting. And when was this trending at peak – coincides with our date of drop.

Good – thats something – and going back to the parking lot, we did see a dip in time spent in Nor am as well – so we can assume that users are logging into fb, but not spending enugh time in the application, instead are jumping over to tik tok where all the activity is. We should be able to cross reference by seeing if we see an overall dip in key action metrics like comments, likes, videos watched etc.

Interviewer confirms that the data is corroborated.

Ok that brings me to a metric that gave us a false positive customer segment – The reason why 116-24 did not stand out was because of daddy daughter dances, both the segments were switching the 16-24 as well as the 45 – 60 to tik-tok to share and enjoy.

Mitigation :

• Run marketing campaigns to bring users back to platform and increase the time spent, but this may not still solve the problem – younger users are looking for something new, and different esp while being locked down
• This one is out there – Consider a new hook to add friends – people are currently yearning for human connection and something personal – so from events and groups, we can consider having a match up with like minded individuals and encourage them to do a virtual watch party or a gameify night ( idea being encouraging users to interact more thru various channels which ultimate should lead to new friends)
• Again, this one is out there – Assumes we have reached a saturation even though the dip is sudden. Folks were using fb because nothing else caught their attention, but now competitors are coming up with new things that is holding their attention more. So, I want to consider the strategy of doing a major overhaul to friends. Redefine add friends. This is infact what one might consider counter intuitive – but we may have reached a point where pretty much all your 1st and 2nd and 3rd circle are friends. So do we need to actually filter down the list – Show a People You May not know – and encourage users to remove folks that they just accepted to not be rude, but they really dont know them? While this may seem like its working directly against fb’s goal of connecting people together, is it really connected if you just added someone that you vaguely remember? Instead wouldn’t the users spend more time in the platform if the news feed that they have is from quality friends – whose life updates they actually care about? Isn’t that the founding principle of connection? Correspondingly permit a second layer of say ‘Barstool-Friends’ to allow for adding friends into this layer who are folks who you dont know personally, but more like pen pals ( folks that you feel connected to in groups, game casts, watch parties etc). These are akin to the strangers you meet at a large group party every now and then and you have terrific conversations with. Redefine metrics now for add friends now per these assumptions. Adjust for removing of friends, as well as the new ‘Barstool-Friends’

Launch Plan

• I will pick the last option, and will try to do some A/B scenarios on this and read how the newly defined metrics are doing against them. one example would be to actually only show remove friends to one group and show remove friends , and add barstool friends to other group. Hypotheses is that the second group should have further improved metrics than the first group and more user engagement
• Controlled Launch to may be just one user group ( Ages 25 – 35) – Millennials – more loyal to fb
• Monitor metrics atleast for a month to notice gradual increase in barstool friends

Broad Approach:

1. Understanding the question
2. Dig deeper to narrow down
3. Hypothesize for probable cause
4. Test the hypothesis
5. Propose solution

Understanding the question

• What do we mean by Friend Requests?

FB Users sending requests to other users through ‘Add Friend’ button. It doesn’t take into account whether the request was accepted or declined.

• When we say 10% down, what’s the time frame? Is it sudden or gradual?

Sudden could be because of a bug, downtime or some change in data capturing.

Gradual points to product reason like user behavior change.

Assuming- Gradual over a week

Digging deeper to narrow down the cause

External factors:

1. Time:
   1. A one-time thing or has a seasonal pattern? Say, yearly around the holiday season we see this drop as users might be less active on social media.
   2. Any pattern like significant drops at specific times during the day or day of the week?
   3. If it were sudden, what happened around the start of the drop? Some external or internal event that caused drop? [ignoring this as we established it to be gradual]
2. Region: Region-specific or happening globally? If it’s regional, we might have to check for recent news wrt geopolitics, regulations, epidemics, or regional competition. Or study the shift in user behavior in that region.
3. Platform: Any specific platform like web (any browser), android, iOS? If it’s platform-specific, it could be some new platform feature that introduced some bug, or users moving to the other platform because of some new model/improvement.
4. User Segment: Any specific user segment or demographic? There could be a new product/feature/activity that’s attracting this segment for networking/connecting
5. Industry & Competition:
   1. Is the trend specific to FB requests, or similar drop is seen in similar products like Snapchat, Twitter etc.? If industry-wide drop, that points to change in user behavior.
   2. Any new competition? May be a new competitor like Tinder.

Assuming : No seasonality, no pattern in drops, globally distributed across platforms, uniform across user segments and no new competition. No information on similar metric of other competitors.

Internal factors

6. Scope: Is the drop experienced across FB (overall traffic decreased or usage of other features affected) or only in the Friend Requests? If this experienced across FB, it could be a bigger issue that needs discussion with other teams.
7. Cannibalization: Any new feature or improvements launched that let users view/interact with other users without adding them as friends? It may not be a bad thing as long as it aligns with evolving user trends and meets FB’s vision and goals.
8. Reporting Issue: Is the data logged correctly? Any changes in data capturing?
9. Any recent change in feature that might affect the usage? Any specific entry point where we are seeing the drop? Eg, users can send friend requests from ‘Suggestions’ or visit the profile and send request from there.

Assuming: No reporting error and scope is limited to Friend Requests

Hypothesize for probable cause

10. It could be coz of cannibalization from other feature
11. Change in the feature introduced in last 1-2 weeks

Testing Hypothesis

12. Cannibalization: Check for corresponding increase in any other feature’s usage? Suppose Messenger introduced a feature of adding people on Messenger without adding as friends.
13. Change in Add Request feature: check if

• there’s drop at specific entry point for ‘Add Friend’ option (eg, suggestions section or from profile)
• have we changed any of these flows. Eg, moved/removed Suggestions section or changed the algorithm for Suggestions
• is it an overall decrease in the clicks for ‘Add Friend’?

Solution If cannibalization is the cause, we should look at larger picture and evaluate the pros and cons of pivoting users’ behavior. Does it align with FB’s goals and vision in long term?

If it’s due to change in Add Request feature, identify the point in the user flow where the drop is happening. We might need to roll back the change. Or if the change was made to decrease undesirable behavior like reducing spam requests or requests from fake profile, then this could be the intended result.

Clarifying question 1: Define Friend requests — no. of friend requests sent in the platform through the “Add Friend” button

Clarifying question 2: Time period for comparison – 10% – WoW or DoD or MoM? — WoW

Gathering context:

1. Is the decline progressive or a one-time event? –> progressive
   1. Because the decline is progressive, ruling out technical glitches, downtime, or any other reason impacting the feature uptime.
2. Is this decline global or regional? –> global
   1. Because the decline is global, ruling out any regional influencing factors. No change of preference in any region, no competing product in a particular region, no regulatory constraints
3. Is the decline on any specific platform, web (chrome, firefox, IE, safari), mobile (Android, IOS), mobile handset (MI, iPhone, Galaxy, etc) –> Decline in Mobile both (IOS and Android) – no statistically significant difference observed with handsets.
4. Is there a decline in any other feature usage in mobile –> No. The other metrics dipped but not by 10% – no correlation with any other usage metric found.
5. Is there a prominent user segment where we see the decline? – 19-24

Reframing the problem statement with the context:

6. Progressive decline in friend requests sent on mobile platforms globally and no correlation with any other product usage metric among 19-24 user segment

Hypothesis:

7. One of the sources for adding new friends is broken
8. People are not able to view friend requests and not accepting, discouraging users from sending friend requests
9. Concerned about adding bots/fake accounts that scrape information – not confident about the authenticity of the person
10. A new feature that negates the friend circle requirement to interact

Hypothesis testing:

11. Sources:
    1. The sources of finding and adding a new friend is through – “people you may know”, “Add friend” on profile – Has there been a change in algorithm recommending friends or ordering the people in people you may know? — Need to investigate.
    2. Lower/infrequent placement of “people you may know” on mobile
    3. UI issues with “people you may know”
    4. Broken “Add Friend” feature in mobile might have been detected within a week – unlikely cause
12. People not accepting friend requests:
    1. QA the friend request display and acceptance functionality on mobile
    2. Check the friend acceptance metrics for the decline at the same time period
13. Bots/Fake Accounts:
    1. Work with engineering to identify and crack down fake accounts, check if there has been a rise in fake accounts or news articles in the same time period
    2. Focus groups to identify inhibitions and work with engineering to resolve (FB verified or something that assures that the user is real)
14. New Feature: 1. Any interactive feature in FB ecosystem (all social products) that do not require them to be friends yet interact with each other in the same time period? 2. Has there been an increase in that particular feature usage in this user segment? – considering FB’s policies and vision this reason is unlikely

In summary after drilling down through the problem; the probable reasons for the decline of friend requests on mobile are algorithmic or UI changes in “people you may know” or concerns over fake accounts. The next step is to test these hypotheses to find the exact cause and fix it.