Facebook friend requests have dropped 10%. How would you solve this problem?

problem solving, metrics

Candidate - Facebook is a social network that aims to connect people with each other and build a sense of community amongst its users. People use Facebook to connect with their friends/acquaintances, to discover events that are happening nearby, to join common interest groups and to play games.

If you’re “friends” with someone on Facebook, you will be able to view each other’s posts (if the profile does not have public access), send messages to each other and can see each other’s profiles. The more friends you have on Facebook, the more likely that you’ll spend more time on Facebook and keep using it everyday.

Now in order to solve the friend request usage drop problem, I would like to understand the following

1. Is this drop sudden or gradual? How long has this decrease been observed?

2. Is there a particular region where this is happening?

3. Is there a particular device or channel where this is observed - web, mobile - android, iOS, tablets, iPad, etc.?

Interviewer - This drop is sudden and was observed over the past week globally. There’s no correlation between this drop and the device type, it looks like there’s a problem across all channels.

C - Alright. To rule out factors that are not in Facebook’s control, did any of the following happen in the past week?

1. Did Facebook receive any bad publicity?

2. Is there a competitor that launched recently that’s making users use this competing product more?

3. Is there a significant global event that happened which is affecting user behavior ?

I - No, nothing. There’s been no influence of external factors.

C - Ok. To summarize, there’s been a 10% sudden drop in the past week for friend requests globally and this is not because of external factors. 

Now, I would like to get into the friend request journey

Sending friend requests

Let’s say Parvathi is a user on Facebook and she wants to send a friend request to someone. Let’s take a look at that journey in detail.

1. Parvathi feels a need to send a friend request to someone. This need could stem from any of the following factors

   1. this person is actually known to Parvathi IRL.

   2. She wants to get to know this person - wants to see their profile, send them a message, view and interact with their posts, etc.

   3. She sees suggestions of people she may know and decides to use those suggestions and sends requests.

   4. She is not seeing enough on her Facebook feed and sends out a flurry of adhoc friend requests.

2. Parvathi sends a request and waits for a response from the user.

3. If the user accepts, she is able to connect with her friend more - see their posts, profile and can message this user. Parvathi therefore has more to do on Facebook and will spend more time there.

Acting on received friend requests

Let’s take Parvathi again and understand how Parvathi receives friend requests and acts on them

1. Parvathi gets notified that she has a new friend request.

2. She clicks on the notification to see who it is.

3. She tries to understand from the notification if this is a known user or not. She uses the following information at a glance to understand this

   1. Profile picture

   2. Name

   3. Number and list of mutual friends

4. If Parvathi does not get enough information to make this decision, then she’ll click on the profile of the user who sent her the request and go in there to see who this person is and if she can learn more from this person’s bio and photos or anything on the profile.

5. Parvathi then decides to accept, reject or ignore this request.

 Now, I would like to understand if there’s a problem with sending requests, acting on requests or both? To do this, I would look at the following

•  Number of friend requests sent

• Number of people acting on friend requests (confirm, delete)

Has there been a change in these metrics?

I : Yes, The number of requests sent has increased. The number of people acting on them has decreased.

C: Hmm, strange. Since one user can send a request to another just once, there is no spam involved. Ok let’s try to understand why people are not acting on friend requests anymore.

• Was there a new feature that was released recently?

• Are users getting notified about these requests? To understand this, I would look at the notification service to see number of requests notified and understand if there’s a change. I would also ask the team to run some tests on the request receiving feature.

• If users are getting notified about these requests and they are not buggy, then I would check if the number of views on these requests has dropped.

I: No new functionality released recently. Let’s assume that notifications are working perfectly and let’s say that you notice that there is no change in people viewing the friends request tab. What next?

C: Uhm ok. So this means people know that they are receiving requests and are checking them out. They’re choosing to ignore them but the number of requests sent has increased. 

This implies that they are receiving requests from people who are not known to them/do not seem legit. If the sender is not known, I would assume from my experience as a Facebook user, that this is something that happens usually and would not cause such a drastic drop in the number of people acting on these requests.

This leads me to conclude that users are suddenly receiving requests from fake profiles/bots. To corroborate, I would look at the number of accounts created in the past week. If this has gone up by a lot, then this conclusion is most likely to be true (ruling out factors like campaigns, good press, etc). I would also look at the average number of mutual friends for every user that receives a friend request and if this has drastically reduced, this would further bolster my theory.

I: Bingo! Well done. Users have suddenly started receiving requests from fake accounts which is why they’ve started ignoring friend requests. How would you go about solving this problem?

C: I would first try to understand if we need to solve this problem and why. To do this, I would look at the following

• impact of fake accounts on the important metrics. Let’s assume DAU, time spent on the app, retention and ad revenue are the important metrics Facebook is looking at. If none of these metrics are affected by the existence of fake accounts, I would not prioritize solving this problem right away. 

• the number of fake accounts on Facebook today and the rate of increase of these types of accounts being created. I would work with engineering to understand how to identify fake accounts and understand how many are there and how fast they’re growing. 

If the important metrics are not impacted and there aren’t too many fake accounts, then I would not prioritize solving this problem immediately. 

However, if there are a lot of fake accounts getting created every day, then this would eventually affect the user experience leading to a drop in one or many of the important metrics. In this case, we could do a clean up of the existing fake accounts, understand how people are able to create these accounts and try and improve the account creation process. However, an important aspect to keep in mind here would be that cleaning up a high volume of fake accounts will result temporarily in a decrease in the MAU.