Feb 2018 Youth Small Group Update

Raw text

Transcribed text as recorded, including headers and formatting fragments.

Case 4:22-md-03047-YGR Document 2648-7 Filed 01/20/26 Page 1 of 5 AMENDED Exhibit 41 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT Case No.: 4:22-md-03047-YGR MDL No. 3047 In Re: Social Media Adolescent Addiction/Personal Injury Products Liability Litigation Case 4:22-md-03047-YGR Document 2648-7 Filed 01/20/26 Page 2 of 5 EXHIBIT Feb 2018 Youth Small Group Update We're organized around 3 pillars: Kids, Tweens, and Teens. Kids. Our core bet here is Messenger Kids. Our H1 Goal is to get to a retentive product of at least 25% WAP and then scale. Early data indicate we have relatively high retention but hasn't flattened yet; qualitative feedback from parents and kids is positive. Current WAP@Day 30 is 40%, although we may have early adopter bias. We're currently testing a more representative segment through a QP test. Growth is not currently our focus, but MAP is 60k and DAP is 21k.. Tweens. Our H1 goal is to 1) Define the primary product for Tweens on Facebook; 2) Define the migration strategy for Messenger Kids onto Messenger. Once defined, execute in H2. Due to increased scrutiny of whether Facebook is good for Youth (more on this below), it is possible we may decide to postpone execution on allowing Tweens on Facebook. Teens. Our H1 goal is to drive Teen MAP incrementally by +2% by end of H1 in our focus countries (countries where we are in or at risk of decline). Our goal by end of 2018 is to be- gin recovering and growing Teen MAP in these countries. We've never moved Teen MAP up, but we feel good about the bets this half. Broadly, our work is organized into 1) removing barriers for Teens in our existing product (e.g. Fake Account Detection, Zero Rating, Private Mode), and 2) building differentiated product experiences that give Teens a reason to use Facebook (e.g. High School Communities, Small Groups). We have a goal to scope out 1-2 more bets in the latter category by end of H1, as this is the area where we feel like we are light. --Fake Account Detection (SEV1). We discovered a longstanding issue (since at least mid-2017) where the Facebook Fake Account classifier is inadvertently checkpointing legitimate Teen ac- counts (the precision of this classifier is estimated to be in the 30-40% range). Exact impact quantification is difficult (part of the problem is that our "ground truth" labeling of legitimate vs. non-legitimate accounts is suspect), but we estimate this could be affecting 4M to 20M Teen MAP annually worldwide (which would be in the 1-5% range of global Teen MAP), be- fore network effects. We're working with CI to resolve this issue quickly. --Zero Rating. Our goal this half is to understand engagement we can add via zero rating and the associated cost/policy trade offs, and then ship this where it makes sense (ideally in at least a subset of our focus countries). The Internet.org holdout shows 7% Teen DAP lift, and a test where we made posting videos and photos free showed a 6% Teen sharing lift in Mex- ico and 20% in Honduras. We launched a more rigorous initial test to 0.5% of Facebook Operator Solutions users on 2/2 to understand the impact of offering our current maximum free product to teens compared the standard text-only free product. We're also assessing po- tential new products (e.g. offering video) and carrier/country opportunities. A high pri under- stand effort is underway on Snap's zero-rating approach. --Private Mode. Private Mode takes the principles from why teens use Finstas as an audience solution and applies them to Facebook: smaller audiences, plausible deniability, and private #7 Gelduert 5/8/25 DENTIAL (COMPETITOR) ΜΕΤΑ3047MDL-020-00104557 PENGAD 800-631-6989 Case 4:22-md-03047-YGR Document 2648-7 Filed 01/20/26 Page 3 of 5 accounts. Teens will be able to create a new profile using a nickname and color for the app, select up to 40 friends to share with, and their posts will auto-archive to only-me after 30 days. We call it a "mode" because it is entirely removed from the normal Facebook expe- rience: it has its own Feed, Notifications, Profile, and People surfaces. We're testing this in 3 US high schools in mid-Q1 to evaluate product-market fit; assuming positive results, we'll move to country tests in Q2. --High School Communities. In H2, we learned that increasing high school profile completion drives engagement and DAP, as it makes it easier for Teens to message each other and join peer-heavy groups. We're investing in aggressively driving high school profile completeness this half coupled with better recommendations (expected 0.5% Teen MAP/1% DAP in H1, with a 5% DAP lift over 12-18 months). Additionally, we're spinning up a product bet around providing a dedicated space for Teens to 1) connect and learn more about the friends and classmates they go to school with (profile/browse); 2) share and communicate only with the peers they go to school with, in a predictable, dedicated space (school feed, school groups). In a lot of ways this brings back the experience of college- and high-school Facebook circa 2006. In H1, we'd like to validate that this approach drives significant incremental engage- ment in a few US high schools so that we can scale in H2. --Small Groups/Lightweight Groups: This work is now being driven by Groups and Messenger. The earlier data showing a ~5% increase in top-line posting from Lightweight Groups turned out to be incorrect (due to a methodological error around how we measure network effects in paired countries), but we're still optimistic about this direction. GDPR The General Data Protection Regulation (GDPR), a new set of regulations which take effect in Europe in May 2018 requires that we raise the minimum age to use Facebook to either 14, 15 or 16 (depending on the European country). Raising the age or forcing all teens to get parental consent to use Facebook would have very negative impacts to teen growth in Eu- rope. Working with a cross-functional team we were able to lock in a creative solution, which allows us to comply with GDPR without impact to teen growth. Our solution, a stripped down version of Facebook, will offer teens in Europe a version of Facebook which disables certain sensitive features (e.g. 3rd party ads, certain profile fields like political views, religion). This version will be GDPR compliant and will enable us to keep all current 13 to 16 year-old teens and future teens on FB without forcing them to acquire parental consent or them hav- ing to lie about their age. "Facebook is bad for youth" narrative This is becoming a bigger thing externally, and we expect this to continue (e.g. on 2/7 Common Sense Media hosted an event in DC titled "Truth About Tech: How Tech Has Kids Hooked" with Chelsea Clinton, Roger McNamee, Tristan Harris, Senator Mark Warner, and more critics of FB). Specific concerns include: over-use contributing to lack of sleep (which then amplifies other negative health outcomes); promoting unhealthy body image; cyberbul- lying; anxiety and depression; generalized worse well being by self-reported measures from using our products. HIGHLY CONFIDENTIAL (COMPETITOR) META3047MDL-020-00104558 Case 4:22-md-03047-YGR Document 2648-7 Filed 01/20/26 Page 4 of 5 We don't think the recent CCFC campaign to shut down Messenger Kids has merit. They con- flate "social media" with "private messaging," which are quite different. Messenger Kids differ- entiates by giving parents more control over their children's messaging than competing prod- ucts. We feel good about the substance and press narrative around Messenger Kids and our product for U10s; the work we did leading up to launch to enlist allies was very helpful. We do need a better understanding around ground truth for Teens on Facebook. The best internal data we have around how Facebook usage affects well being (not specific to Youth) shows two things: 1) People who use Facebook more have lower well-being than people who use Facebook less (correlative; non-causal); 2) longitudinally, the effect reverses, and some Facebook interactions are positively associated with well being over time. It is important to caveat that this data is still very early and needs further vetting. We are working on a more focused version of this type of understand work for Youth, and on the product side have spun on a Youth Empathy team specifically focused on this. Team. We recently hired in the former head of research from Mattel, to lead our Youth research team. He has 25 years of experience researching kids and teens. Our key remaining open IC7+/D1 roles are PM for Messenger Kids (need full-time lead either for Messenger Kids or Tweens) and PM for Growth (need full-time lead for Zero Rating). DISCUSSION TOPICS "Facebook is bad for youth" narrative. Are we doing enough here? Are there lessons we can learn from our experience with Integrity over the last 6-12 months? Should we be front- loading more Understand work and staffing up leadership accordingly? Also, the concerns that Facebook is bad for you, while particularly sensitive for Youth, are by no means limited to this segment. David Ginsberg is working on framing up what would be needed to deliver Understand work as fast as possible and associated staffing tradeoffs. Key learning/challenge. Creating space for 0-1 FB Blue bets. For 0→1 bets like Private Mode, it's critical we're able to focus time and energy on getting to product-market fit and learning/pivoting quickly. Right now, we're spending too much time clearing approval barri- ers designed to reconcile interactions at scale. As an example: for Private Mode, we needed to escalate to 3 VPs to unblock a promotional notification for 300-500 teens in our test schools (~150 Teens affected per VP on the email chain). We're working on solving this by creating more space for product-market fit tests at a small scale (similar to how QP doesn't require approvals below a certain number of people affected), with the understanding that if/when we do find fit, there may be a longer process for reconciliation before broad launch. Key learning/challenge. Data and labeling: investing in "ground truth" understand work. We'll do a full postmortem of the Fake Account SEV1, but one clear issue is that we did not invest enough in validating our foundational "ground truth" labeling, which meant that 1) we didn't have a good understanding of what was actually going on; 2) everything else (e.g. the Fake Account classifiers) that was trained on this data was incorrect. HIGHLY CONFIDENTIAL (COMPETITOR) META3047MDL-020-00104559 Case 4:22-md-03047-YGR Document 2648-7 Filed 01/20/26 APPENDIX Messenger Kids Retention Retention Dash: https://fburl.com/mkearlyretention MK Cohort Retemion Rate Curve By Wesk زیز ! 8 75 Page 5 of 5 10 25 30 days since inst active day 2018-01-29 seketюн..ро 2017-12-25 retention pct 2017-12-11 retention. pct 2018-01-22 retention pet 2018-01-08 setention pot 2017-12-04 retention. pct 2018-01-01 retention pot 2018-01-15 retention.pct 2017-12-18 retention.pca High School Communities Concepts Lowell High School Coranumity 5000000000 * On-Boarding (School Graph) we 1006550-0000 Nock ops tés০০ ১৯০০০০১ NON διχά XXXX * : Trending News and Connections (Recommendations) Classmates Directory (School Community Vertical) HIGHLY CONFIDENTIAL (COMPETITOR) ΜΕΤΑ3047MDL-020-00104560