How would you decide to display ads on the Facebook Newsfeed? After every 100 posts or 25 posts? Evaluate tradeoffs.

Context

First off let's establishing a common definition of just what the Facebook Newsfeed is and how users use it. The Newsfeed is the first thing users are shown when they log into Facebook on both desktop and mobile. It is a continuous feed of what has been algorithmically determined to be the most relevant / interesting content for you from your network. It contains posts with text, photos, videos, friend suggestions, etc. Users can then interact with these posts by liking, commenting, reacting, etc. in addition to viewing them. 

To further frame this question, Facebook's mission statement is to bring the world closer together and give people the power to build community. Facebook's Newsfeed help people connect with each other by making it easy to for users to consume and interact wit content from the people they care about.

 

Goals

At its core the Newfeed is a huge driver of engagement and retention. The Newsfeed intentionally shows you the posts you are most likely to engage with and makes it easy for you to engage on a deeper level than just viewing by reacting or commenting on posts right there in the feed. Additionally, people have this positive experience in the Newsfeed and then come back later to see any new activity since they were last logged in.

When it comes to ads, Facebook is a very mature product with a ton of users and the company makes the majority of its revenue through showing ads to the users. The goal of display ads is to show users products they might be interested in but to ultimately make money for the company by compromising on user experience.

 

Tradeoffs

Most new features or product changes are never purely negative or positive and display ads are no exception. In terms of the potential impacts both positive and negative that display ads have on the Newsfeed:

Pros:

-Generates revenue for Facebook

-Users may discover new products they are interested in

-Advertisers are finding new customers to connect with

-Increased amount of ad inventory available

-Facebook is able to fill ad requests faster

 

Cons:

-Decreased UX, the Newsfeed would be less engaging with more ads -> lower retention

-More ad impressions would lead to a lower ad CTR

-Users would begin seeing less and less relevant ads as we run out of inventory or the same ads as time goes on

-If ad relevancy drops advertisers would get less results from their campaigns and be less likely to advertise again in the future

 

Execution Plan

After walking through the above tradeoffs, I don't think the optimal approach would be to universally apply  a set ad frequency across the board. Realistically speaking there are a bunch of different factors we need to consider when deciding whether or not to show a user an ad. 

 

Short Term

That is going to be a little bit more intensive and complex to optimize, if we needed a near term solution while we figured out the complex solution we could perform a/b testing where we varied the frequency display ads showed up. 

 

Facebook has a large number of users so I think applying each of these ad frequences to 1% of users each would give us enough data for our test: 5, 15, 25. I would want to analyze the effect of these different ad frequencies over the course of a month to measure the impact on retention and get a better sense of how our ad inventory quality declines over time. In order to measure the impact of these different ad frequencies on the product we would want to monitor and compare the following metrics:

1. Average number of posts viewed per Newsfeed session

2. Average duration of a Newsfeed session

3. Average number of interactions per Newsfeed session (Likes, comments, reactions)

4. Ad CTR

5. Average Number of Display Ad impressions per session

6. Percentage of users who viewed at least 10 posts in a session in week 1 who also viewed at least 10 posts in a session in week 4

 

By analyzing the above metrics, we can find the approximate sweet spot in terms of ad frequency for mildy inconveniencing users / maintaining engagement while helping our advertisers reach customers and generating revenue for ourselves.

 

Long Term

As mentioned above, there are a bunch of different factors we should consider when deciding to show an ad to a user. Some of those factors could include:

1. Relevancy of ad to the user - My grandmother probably wouldn't appreciate BMX camp as much as the middle schooler next door

2. Format of the ad - 300x250, 728x90, or video? Some ads are less intrusive than others

3. User's Spare Income Level - That new graduate with a six figure job is going to have more disposable income for purchasing things we show them than the 

4. Ad Fullfillment Rate - If we promised the advertiser we would deliver X impressions in Y time period we may need to more aggressively show certain ads

5. CPM/CPC Rate - Not all ads are equal, if we have an opportunity to show an ad that has a CPM or CPC rate 10x the average we should make more of an effort to show it

6. Number of ads previously shown to the user - The users annoyance would increase with the number of ads shown

 

We could work closely with the data science team to quantify the impact some of the above variables have and then create a model to decide whether or not we should show an ad.

 

Summary

The Newsfeed primarily serves as a way of helping users engage with the Facebook platform by showing them a feed of relevant content they are likely to enjoy and interact with. Ads, while not users' favorite thing, are a necessary evil that helps Facebook make money. In order to figure out what ad display frequency is least disruptive to the users, helps advertisers makes sales, and generates us the most revenue we will perform an A/B test with several different display ad frequencies. Beyond that we will work with the data science team to develop a model based off of several variables listed above to determine whether or not we should show an ad to a user.

hmmm...ok so facebook newsfeed is the main page of FB and the primary product. Correct? (yes)

Ok and I assume that we aren't trying to decide a number here for a specific type of user but rather all customers of FB newsfeed would have the same number after which they will see an ad. Correct? (yes for the intention of this case).

ok and also, I did mention customers but FB is a 2 sided marketplace and I assume here we are talking about consumers and not the creators. Correct? (yes you can assume so).

Perfect! So I think it's a safe assumption here that we want to decide this because there is something called Ad fatigue for lack of better words or deteriorated user expeirence that could result from too many ads. (correct). So we really need to be able to measure after what min# of posts an ad results in and maintains certain level of interest from our user. Correct? (absolutely).

So I think how I want to go about this is by thinking about what my user typically does on NF, what FB cares about from a business impact / objetive standpoint and then using the user journey to define some KPIs we could measure. Is that a fine approach ? (yes).

ok from a business standpint considering FB is a mature business, customer acquistion and adoption of FB NF isn't relevant but what is relevent is maintaining user engagement and retaining that user. But really engagement drives retention so I think we should evaluate what is that min # of posts that there should be on NF before displaying an ad without impacting user engagement resulting from ad fatigue. Is that in aligment with your thoughts? (yes it is).

So we should measure:

1) % of users login in at least thrice a week (I assume that most users of FB are young adults who tend to visit FB frequently during the week. 3 is an arbitrary number if we had data i would be able to better estimate this #).

Ok before we go further, I am thinking of question again and I realize we need to evalute if ads will be posted after either 100 or 25 posts. So this makes me think this is more of an A/B testing case as well. I know I started measuring but I think I should take a pause here and talk about A/B test setup. I think we should have a random sample of same users, part of whom would be a control group (no changes for them), another part would have ads placed after 100 posts and third group would have ads placed after 25 posts. Then i think we should meausre these KPIs to see if ads after 25 posts is impacting engagement significantly or not. If there is only a slight deterioration I would still refrain from posting after 25 posts. In tehory, user experience should never be impacted since that has a longer term affect on the brand.

2) % users who spend atleast 5 min scrolling through posts / week. (I assume we hvae a way to measure scrolls vs looking at post and writting comments or reading comments - passive idle time. Scrolling through means FB is engaging me and # of ads posted isn't bothered me.)

These are the only mesures I culd think of since other peices of engagement such as commenting, sharing, liknig, etc, aren't relevant here. I think if I had to hang my hat on that one metric it be #2 since #1 just talks about user loggign in atleast 3 times a week but even if a user logs in it's no good if my user isn't spending as much time scrolling through or i see deterioration in amount of time user is spending on scrolls. Less scrolls means I haven't kept my user interested enough (in a sense engaged) and that will lead me to loose the user (no retention) and a negative business impact (branding, loosing revenue, etc.). + we user may be visiting FB often but we can't track that effectively because FB session was never ended or never logged out. So I think #2 helps us get to the core to analyze in what csae (control group, after 25 posts, or after 100 post) user spends at least 5 min scrolling and we see an increasing pattern in time spent on scrolls.

Overall, I think this is a problem that best evaluated by conducting an A/B + measuring success to see what # of posts before an ad keeps the user most engaged.