As a Product Manager on Facebook Newsfeed, how would you decide whether to show an advertisement or the "People You May Know" widget?
0 votes
in Product Design by (18 points) | 1.3k views

3 Answers

+3 votes
So lets understand the intent of Facebook Newsfeed. The newsfeed is the first page that user see after loging in into the account that shows a collection of posts that may interest the user. From the users perspective, she will want to know

1. activity on the posts she created

2. interesting updates from her friends - ranked in order of user proximity, trending posts, current affairs in social circles

3. Updates from grps\websites that she may follow

4. Potential contacts - ppl you know

5 Ads.

From: business standpoint: ads provides monetization, ppl you knw - helps activation and engagement

Given facebooks mission to build authentic communities the ppl you knw brings a better value than ads also provides for a better landscape for monetization, engadement, etc.

Some of the other considerations to factor in the algorith for page creation should be

1. User usage patterns: If the user doesnt logon freq and hasnt developed the stickiness, we should prioritize UUMK instead of Ads since that provides more value for the user to revisit the page

2 History ad freq: If a user logs in multiple times a day and have a pretty dense friends graph, we should prioritize ads

3. CTR and impressions: user personalized ads and events on these ads. Based on past behaviours, followed posts, demographics we can use AI to understand if the specific ad content is useful and relevant

4. Real time - Another dynamic variable is time of day and current events. For example, it might make sense to render ads for Personalized gifts around Mothers day or Valentines day, Ads during night time/liesure time browsing.

Given there are multiple variables that need to be considered by the algorithm to determine the content, I would design this to be a dynamic AI based rule engine optimized by different ranking and relevance engines. Also, the success of these decisions should be monitored by metrics. Another key metric we need to be mindful of, while rendering ads is cannibalization. We dont want to adversely impact the users experience and move away from the core value proposition
+2 votes

Note of caution: I landed this answer in an anti-climactic state. It felt natural and it was a good opportunity to employ some A/B testing. But would love to get some perspective on it.

Starting with the Newsfeed product goals

  1. Update user on their friends & family activities and posts
  2. Show what is trending in their social & demographic circle - news or a topic
  3. Ability to show contextual ads that users may find helpful in enhancing their life

Based on the goals I have mentioned above let's do a cost-benefit analysis of the two features:


Business GoalAdvertisementPeople You May Know (PYMK)
User Engagement & Retention  Yes (Primary)
Monetization Yes (Primary) Yes (Secondary)

If we look at "PYMK" as a lagging indicator or opportunity of goal #3 mentioned above, (that is better, target ads for the user depending on what their demographics, social topics, preferences are) then it is a great long-term or fail-safe feature. 

Since I want to understand how showing ads may help or deter users, we can proceed with an A/B test hypothesis:

  1. Hypothesis: user's who are actively engaged and already have social interactions on a regular basis are good candidates for targetted ads.
    1. To evaluate this test, we will probably want to measure the CTR on ads by this segment of user & some health metrics for a given period that shows if the engagement goes up with the introduction of the ads.
    2. Tradeoff: my concern with the engaged set of users seeing ads every time they log in will be that this will create fatigue and their interaction levels may decline.
  2. Hypothesis: User's who are passively active on facebook (check their friends feeds but not post much/anything for long periods) may find an incentive in helpful targeted ads as an informative platform to find cool/new things as it relates to them.
    1. To evaluate this test, we will need to measure the CTR on ads by this segment of users, keeping in mind that these users are not daily or even weekly active user and so the metrics will need to gauge over a longer time period. 
    2. % of WAU for the segment above will also be a helpful indicator of the engagement going up or further down with this feature.
    3. Tradeoff: I feel the biggest trade-off here is the user churn. So this has to be a calculated step. But the reward is big too. If the feature is positioned to this segment correctly, it may enhance user engagement and may give them something more to come back to the FB newsfeed for.

Summary: to capture the analysis above, it may be best to run an A/B test on a segment of users to decide if the algorithmic change of showing ads vs. showing PYMK widget is the right fit. If we go by the model above it may be a great opportunity to engage a passive segment of users with the caveat that the ads are informational and useful to them in addition to just CTR.


Ideas to curb the negative impact of the feature deployment: One last thing to add here is a kind of snooze button or a thumbs down that tells us from those segments of the user if they would like to view ads or if the add was useful. This will help us refine the ad content engine.


I hope you find this solution helpful!

by (45 points)
0 votes
Great answer. Only thing I had to add was potentially aligning with FB's overall goals, being more revenue or greater user engagement. The key strength behind products like FB is the network effects it is able to build upon by having users establish multiple connections. I would look at establishing common metrics between both cases such as CTR, revenue, and user engagement (number of friend requests). These metrics would have different weights based on the organizational goals. I would recommend A/B tests using one view with the ads and the other with the "people you may know" widget and try to make a data driven decision base don the impact on metrics.
Your name to display (optional):
Privacy: Your email address will only be used for sending these notifications.
To avoid this verification in future, please log in or register.
Your name to display (optional):
Privacy: Your email address will only be used for sending these notifications.
To avoid this verification in future, please log in or register.