15% off membership for Easter! Learn more. Close

Determine the right number of photos to show in Facebook's newsfeed.

Asked at Meta (Facebook)
1.4k views
Answers (3)
crownAccess expert answers by becoming a member

You'll get access to over 3,000 product manager interview questions and answers

badge Platinum PM

Context

First off let's just get on the same page by establishing a common understanding of just what the Facebook Newsfeed is and how users use it. Facebook's Newsfeed is the first thing users see when they log into Facebook on both mobile and desktop. It is a continuous scrollable feed of what has algorithmically been determined to be the best most relevant and interesting content from around that user's network. It shows a variety of content like photos, videos, Facebook groups, people you may know, and ads in the form of posts. Users may also interact with these posts by liking, commenting, or reacting.

 

To better frame this question and understand the role Facebook Newsfeed plays in the bigger picture, it's good to examine it in the context of Facebook's mission statement which is to bring the world closer together and give people the power to build community. Newsfeed helps further this mission statement by making it easy to consume content and stay up to date on the lives of those you are close to. Similarly it makes it easy for your friends to consume the content you share.

 

Goal

At it's core the Facebook Newsfeed's purpose is to primarily drive engagement and by extension retention. The content in the Newsfeed is captivating thus motivating users to keep scrolling and to check back later for more new content. It is addicting, have you ever closed out a FB tab only to absent mindedly instantly re-open it again in a new tab?

 

A few metrics we could use to monitor Newsfeed's ability to engage and retain users are:

1. Average number of posts viewed per session

2. Average newsfeed session duration

3. % of users who had 3 or more sessions last week where they viewed at least 10 posts who also had 3 or more similar sessions this week

 

Trade Offs

Like most features, showing more photos in the Newsfeed doesn't have an entirely positive or negative impact. Let's go ahead and brainstorm what some of the potential tradeoffs could be for increasing the amount of photos shown in the Newsfeed:

 

Pros

-Photos are easier to consume than other posts -> Increased number of posts viewed

-More overall interactions with photos will encourage users to upload and share more photos

----Larger training set for our visual ML models

 

Cons

-Run out relevant photos to show -> Interaction rate and engagement goes down

-Less airtime with other post types like videos will lead to less interaction with them and declined usage over time for those post types as that feedback loop is dampered

-Users who don't post photos will receive less interactions and their engagement will drop

 

Near Term

If we were short on resources or needed to come up with an answer ASAP, I would propose we perform an A/B/C test where each testing option is applied to 1% of Facebook's users. This % is small but Facebook has a large number of users so each sample size should be sufficiently large. Based upon which bucket they are in these users would either see a photos in the Newsfeed every 5, 10, or 15 posts. We would then monitor and compare the following metrics to help use decide which frequency is best:

1. Average number of posts viewed per session

2. Average newsfeed session duration

3. % of users who had 3 or more sessions last week where they viewed at least 10 posts who also had 3 or more similar sessions this week

 

Longer Term

Ultimately, I would like to take a more nuanced approach where we decide whether or not to show a photo in the users Newsfeed based upon a model incorporating several factors instead of a totalitarian approach where a photo is showed every x posts no matter what. Several of the factors that we could consider:

 

1. Users affinity for photos - some users may like or interact with photos more than others

2. Relevancy of photo subject to the user - A user who is a musician would be more interested in a photo of a guitar than someone who is not musically inclined

3. Closeness of the use to the poster / tagged users - I would be more interested in a photo of my sister than someone I had a group project with in college 10 years ago and haven't spoken to since

4. Amount of photos the user has already viewed - Do they have any fatigue related to having already been shown a lot of photos?

5. Relevancy of competing posts - I think this is a large factor. The question is less about should we show a photo next, but more so about what should we show next period? The relevancy of other potential posts to the user would play a big role here. If their best friend just made a post that is probably more important than whatever photo is up next.

 

It would take a lot of time and we would need to work closely with the data science team to quantify and analyze how  these different variables affect engagement on photos in the Newsfeed but for such a critical core feature of Facebook I think it would be worth the effort.

 

Summary

Facebook's Newsfeed helps users consume relevant content from around their network. In order to figure out the optimal number of photos to show in the feed to users we could perform an A/B test with varying levels of photo frequency and then measure their impact on key engagement and retention metrics. Ultimately, we could work closely with our data science team to to build out a model for deciding whether or not we should show a photo in the user's Newsfeed based upon a variety of relevant factors.

 
Access expert answers by becoming a member
0 likes   |  
Get unlimited access for $12/month
Get access to 2,346 pm interview questions and answers to give yourself a strong edge against other candidates that are interviewing for the same position
Get access to over 238 hours of video material containing an interview prep course, recorded mock interviews by expert PMs, group practice sessions, and QAs with expert PMs
Boost your confidence in PM interviews by attending peer to peer mock interview practices, group practices, and QA sessions with expert PMs
badge Platinum PM
This is also an interesting one. Let me get confirmation on understanding of this quetsion first.

1) You want for us to determine what's the correct # of photos to post automatically on News Feed? (correct).

2) And you don't mean for me to come up with a speecific number but more so a methodology that we can apply to determine that number. (correct).

And there has to be a reason why we are interested in this. So let me first get an undersatnding of why we might be interested in this. Per my thinking, FB News Feed's posts are mostly intended to directly align with FB's mission of making people across the globe feel more connected together by means of people respoding to pictures posted (likes, reactions, comments, etc.). Basically, this drives business impact for FB since FB is essentially digital advertising company. So FB's goal is to see what # of pic shown on NF causes people to interact more with FB NF. (fair).

So then really we are looking at what # of posts casuses higher engagement and then as a secondary metrics leads to retaining a user for frequent visit. (correct).

Yes, so those would be my north star metric. Essentially, I would like to see an increase in # of likes, reactions, comments (engagement) to determine our success criteria. (correct).

So thinking about this a bit, let me see how we could possibly measure engagement.

1) % of users commenting or engaging when a post with photos is posted by FB

2) % of shares of a post with photos

3) atleast 10 min spent by about 90 percentile of users as a result of a post with photos. I say 10 here since in my personal experience, it's easy to spend 10 unproductive time getting consumed in various posts by users. This time shouldnt be chosen arbitarily though and there should be some reaserach conducted on this by using our data scientists.

There maybe more ways for us to come to a conclusion on this but if I had to hang my hat on one of these then I would consider #1 to be the key one. While sharing is imp since that dictates how many more will "likely" (just because I share with a friend doesn't mean he/she will view it) they are to see it, #1 is the actual and direct action of someone scrolling through and maybe clicking on a photo post. This means I get a good coverage of CTR and CPI.

I would say though that while this isn't part of our question, we should be concerned about evaluating emotions / sentiment of users (use reactions for this) since if people are psening more time but reactions lead for us to say that people are mostly feeling sad or depressed or frustrated or angry on many of the photo posts, then that's not going to be a good thing long run for FB.

In terms of choosing that 1 number that you asked for, my sense is that we should look for an inflection point that shows conversion from Engagement to Retention of a user viewing photo posts. So ther might be some measures like after a certain # of photo phosts the user definitely spends atlest 10 min and continues that behavior for longer period of time.

Overall, I thiknk we should consider our main intention of why we would want to have photo posts before we can get to the optimal number to be posted. This is important since it helped us come to an answer of what to measure (in this case Engagement primarily).
Access expert answers by becoming a member
0 likes   |  
Get unlimited access for $12/month
Get access to 2,346 pm interview questions and answers to give yourself a strong edge against other candidates that are interviewing for the same position
Get access to over 238 hours of video material containing an interview prep course, recorded mock interviews by expert PMs, group practice sessions, and QAs with expert PMs
Boost your confidence in PM interviews by attending peer to peer mock interview practices, group practices, and QA sessions with expert PMs

First of all I'd like to understand how this feature works. My assumption is that users see posts with photos when their friends create albums and share them. This is not a new feature (at least I've seen that 8 years ago already), so it will be helpful to know why we are looking to change the amount of photos now. 

We also need to take into consideration segments, for example is this for regular user or for some specific user type, as for example users require accessibility features. Also what is the platform?

Let's assume the following:

  1. We are doing this because we haven't touched this feature for a long time and believe there is an opportunity for potential improvement.
  2. Our target group is regular user
  3. We target both mobile and web platforms
  4. The current amount of photos is 10

I'd take a minute and think about how to approach this question:

  • Go through the userflow and define how improvement looks like from a business perspective
  • Design an experiment
  • Summarize and go over possible scenarios after experiment

 

Business goal and userflow definition

For facebook business goal is drive engagement, whatever is good for any feature is good for facebook as we can show more ads. So engagement rate will be the main metric I will be looking at here. 

 

Considering the user flow: User lands on the newsfeed -> they see new post with photos -> they either skip or engage

I will define a few data points we should have for the flow above:

  • A: # post views
  • B: # of shares / likes / reposts / clicks on thumbnail / comments
  • Engagement rate for this feature: B/A * 100%
 
Going from here we need to design an experiment to determine the right amount of photos, considering we are targeting multiple platforms we need to take that into consideration, as user behavior is different across different platforms. 
 
To find a good number of photos I'll create a couple of multivariative experiments (one for each platform) and considering our current amount of photos is 10, I will breakdown a certain amount of users into cohorts and assign different amount of photos for each cohort. This would have been an issue in any other company as multivariative test will require a lot of data to find out statistical significance, however that should not be a problem for facebook. 
 
Experiment
Reach: 10% of newsfeed visitors
Cohorts: 10 per platform
Experiment: Each cohort is being assigned specific amount of photos (we can start from 5) in photo-type posts
Measurement: Engagement rate
Length: 1 week or until statistical significance is reached for platform
 
We would compare engagement rate metric to the baseline engagement rate we have for the feature at current implementation and select best variant for platform once it reached statistical significance. 
 
Considerations:
  • If we see that engagement rate increases with amount of photos we will need to expand amount of cohorts to find out sweet spot, however also ensure this feature does not cannibalize other items in the newsfeed. To do so we can compare baseline engagement rate per user in the newsfeed with same rate for cohort in test. 
  • In case we can't reach statistical significance we will need to expand the Reach or reduce amount of options, or increase time.
 
 

 

 

 

Access expert answers by becoming a member
0 likes   |  
1 Feedback
badge Platinum PM

Things you did well

  • Clarifying questions: good job asking questions. I especially like how you asked about accessibility.
  • Approach: I like how you layed out your approach before diving into the answers.
  • Goal and user flow: great job identifying the goal and describing user flow.
  • Experiment: I think you nailed it to experiment the feature before deciding. 
I really like this answer. Great job!
0
Get unlimited access for $12/month
Get access to 2,346 pm interview questions and answers to give yourself a strong edge against other candidates that are interviewing for the same position
Get access to over 238 hours of video material containing an interview prep course, recorded mock interviews by expert PMs, group practice sessions, and QAs with expert PMs
Boost your confidence in PM interviews by attending peer to peer mock interview practices, group practices, and QA sessions with expert PMs
Get unlimited access for $12/month
Get access to 2,346 pm interview questions and answers to give yourself a strong edge against other candidates that are interviewing for the same position
Get access to over 238 hours of video material containing an interview prep course, recorded mock interviews by expert PMs, group practice sessions, and QAs with expert PMs
Boost your confidence in PM interviews by attending peer to peer mock interview practices, group practices, and QA sessions with expert PMs