Below is how Facebook describes Rival Peak and its trailer:
Rival Peak is an experimental competition reality show, featuring artificially intelligent contestants whose fate is controlled by YOU, the viewing audience. You can interact with the real-time experience through an instant game right on Facebook.
Do you think the experience presented in the trailer above interesting?
The interactive experience will be launched later today so apparently I get my impression from the trailer. I know it’s totally subjective but at the moment I don’t think the experience is interesting or engaging enough to retain a level of audience / players that Facebook expects.
While I don’t watch reality shows and seldom play role-playing games, my first questions for this interactive experience are:
What’s the core experience for a reality show?
What’s the core gameplay for a role-playing game?
This experiment blends observing element in reality show (watching someone interacting with others) with immersive element in role-playing game (playing the role of someone) and adds some UGC element (audience decides some parts of the show / game) to create a new category / genre. This new experience feels confusing to me. I hope Facebook finds ways to make the experience enjoyable otherwise this will be just… an experiment where you blend two things together hoping to enjoy both but it turns out you get non.
Nobody cares about the cloud computing or AI stuff in cloud gaming. They are like new capabilities that people expect to have so that the gameplay will be more enjoyable or engaging…
In terms of IP, I think it’s not going to be long lasting. Again, very subjective, maybe it’s due to art style… Developers might further expand the IP during the show / game but I see limited potential overall.
Since it’s a reality show and a game played real time, live ops might be the only place that this experiment can have optionality. At glance, the setup is not very favorable but strong live ops that have quick iteration and creative operation / coordination might be able to turn things around.
It’s difficult to say an experiment is a flop (well it’s an experiment!) and the fact that the show / game is being distributed on a 2B+ social graph guarantees it will have some impressive metrics like peak concurrent viewer / players… That said, given my impression above, I expect retention will be weak even though Facebook might design various loops in the experience to maintain engagement.
People will talk about 2B+ build-in social graph cannot guarantee the success of a confusing experience and that the cloud gaming in this iteration is not what they expect (they’re not surprised or amazed by the experience).
I expect there will be other iterations on the tech but no season 2 for Rival Peak. Maybe I will change my mind after trying the experience.
I just skimmed the F1 of Yatsen Holding Limited (YSG.US). The stock is expected to list on NYSE on November 19th, 2020. I used to summarize S1/F1 but now I notice there are many participants doing similar stuff and even large part of broker’s PDIE deck is just a summary of F1 in PPT. Thus, I would just skip the duplicate work and only focus on what still stay in my mind. For underlying stories depicting what’s actually happening, check local news outlets or blogs.
* * *
[November 11th, 2020 ]
Yatsen is a player with four brands (including a French brand acquired last month) in China beauty market currently focusing on color cosmetics and skincare category and Gen-Z & Millennium demographics.
Yatsen utilizes direct-to-consumer (DTC) business model supported by well developed e-commerce platforms, social & content platforms and ODM/OEM & packaging supply chain infrastructure in China. Yatsen is a pioneer in key opinion leader (KOL) marketing and maintains a KOL network which the company refers to as direct-to-KOL. DTC business model makes Yatsen know more about its customers’ behavior and preference thanks to data collection and analysis. Data can help the company from product development to supply chain management.
The company adopts an omni-channel strategy to acquire and retain consumers. While the majority sales come from DTC channels (online and offline), Yatsen still sells to e-commerce distributors such as JD.com and VIP.com (that’s it’s omni-channel). Not surprisingly, there’s “margin” difference between channels on the face value but I’m not sure exact contribution margin and ROIC between each channel.
Thesis– Growth Runway
Yatsen looks like a typical consumer company that successfully leverages internet trend to disrupt the industry. This is evidenced by Yatsen’s impressive growth since its inception in 2016. The company ranked No. 5 beauty company in China in terms of color cosmetics retail sales value in 2019.
Since Yatsen is a consumer product company in China, the first thing is to check macro drivers to make sure it has enough runway. Some “typical” macro drivers for China’s beauty market listed below:
Consumption upgrade will make per capital spending go up
Increased demand from lower-tier cities
Large Gen-Z and Millennials population
Domestic brands taking share from international brands
Growth strategies for a typical consumer product company:
Growth strategies for a more ambitious consumer product company:
Acquisition and then incubation
Only companies that have brand value can talk about pricing power. I believe it will take sometime for some of Yatsen’s brands to acquire that market position.
The rests such as DTC, omni-channel strategy, KOL, supply chain ecosystem, data analysis… are just means to stay relevant in this high growth market. I view these more like sources of weaknesses and threats rather than strengths and opportunities, let alone competitive advantage.
Here’s how I think about this bet: huge growth opportunity and everything depends on whether management’s execution can capture the opportunity, engage consumer, fend off competition, maintain bargaining power over suppliers in particular those KOLs… China consumption bet is relatively robust but hyper growth, fast IPO bet is riskier. Last hyper growth consumer product company comes in my mind is Luckin Coffee so this might depress the sentiment. However, I believe there are still many participants happy to bet on macro story and high growth.
Investors that have indicated interest in subscription: Hillhouse, Tiger Global, and Tencent.
Yatsen’s disclosure isn’t enough for me to build a meaningful model but still need one to have a sense where it trades.
💡 This note shows some Free Fire user metrics disclosed in Sea’s filings and earnings calls. Combined with financial data in Free Fire – Preparation Revenue Estimation, one monetization metric, average adjusted revenue per daily active user, is also calculated. I fill the note with my murmur.
Since 2018, Sea has disclosed various Free Fire user metrics to help public investors gauge the traction of Free Fire. The disclosure is not very consistent as some metrics were disclosed at milestones only and some were missed (wonder why s-s analysts didn’t ask for peak 2Q19 and 3Q19 DAU numbers. Maybe they rely more on timely data from third party analytics. The reason why numbers were not disclosed in those two quarters is that Free Fire might experience drawdown in peak DAU).
The company doesn’t disclose the number of cumulative downloads which is fine for me. Number of cumulative downloads is noisy and only goes up (vanity metric) while it’s useful to gauge initial performance at launch or on-going game operation / product iteration. Since this metric is top of the funnel, if a new game flops on this metric, it would take lots of efforts to turnaround. We can find the metric used in a time-aligned comparison among games having similar genre, leveraging the same IP or even being under the same developer, publisher or IP stakeholder (basically, any cumulative metrics can be analyzed this way). For example, few days later, the first month performance of Pokemon Unite might be compared to other MOBA, other mobile games from Timi (developer), Tencent (publisher), The Pokemon Company (IP holder) or even Nintendo (IP stakeholder).
Once the game is downloaded and installed on the device, users need to open the game, register and start playing (activation). The game engages user in different ways (engagement) in addition to pure gameplay after an user is activated and before she/he is churned (retention).
Depending on the nature of the product, metrics for user base estimation are different. A marketplace might choose MAU (Monthly Active User), a productivity tool might choose WAU (Weekly Active User), and a messaging app or a mobile game might choose DAU (Daily Active User).
Sometimes user base in a shorter timeframe divided by user base in a longer timeframe such as DAU/MAU can be used a proxy for engagement. If MAU is very close to total active user base, then the ratio means that how many active user uses the product everyday.
Every company has their own way to calculate MAU/WAU/DAU but these metrics generally have average in their calculation because the timeframe of interest for the specific metric is generally longer than unit timeframe in the metric. We might be interested in a DAU on a specific day but it’s more likely we look for DAU of last month or last week. For example, the number series of active user for last seven days looks like: [25, 16 , 13, 20, 13 ,3 ,1]. DAU for last seven days is 13 (average of the number series) and the peak DAU is 25 (max of the number series). The difference between average and max of the number series is significant because the pattern in number series indicates the product might have specific behavior. The number series might come from an office productivity tool so it’s more often used on weekdays rather than weekend. One way to avoid this pattern is to use metrics with longer timeframe. The WAU for this product in this seven days might be 27 (the reason why it might be larger than peak DAU is we’re counting the total unique user during the week). There two observation from the exercise above: 1) peak DAU is generally larger than DAU; 2) the difference between two might be significant especially when there’s an event that drives product usage on specific day. Peak DAU looks better but it’s more aggressive and might be misleading. Maintaining a growing peak DAU is difficult than maintaining a growing DAU even though the overall user base is expanding and the engagement is increasing because peak DAU might be event-driven. What would you do if you find your peak DAU doesn’t grow but you know the underlying is still healthy. Probably you stop disclosing and wait for the overall user base catching up. Sea took a break for two quarters from 2Q19 to 3Q19. This is not a problem for investors as long as they can get third party data.
Sometimes, once the lifetime number of downloads/installed/registered users have reached certain level, it will become more difficult to interpret due to the nature of app user lifecycle (acquisition→activation→churn→reactivation if any→… how do you define those being churned and how do you measure them if you get them reactivated?), on-going update/iteration of software and on-going game operation. The metrics derived from these cumulative numbers get noisier and noisier as time goes by.
Free Fire User Metrics
Free Fire reached 100M MAU and 40M peak DAU in 4Q18. It had 80M peak DAU in 1Q20. If the engagement relationship hasn’t changed a lot, Free Fire might have ~200M MAU in 1Q20 (probably not reached yet because the company might have reported this milestone if it had reached)
The company has been reporting peak DAU since 2Q18 (from call transcript 3Q18). The reporting was suspended in 2Q19 and 3Q19 but resumed in 4Q19.
Combining Adjusted Revenue and User Metric
The company generally doesn’t disclose number of paying users for specific game.
In April 2020:
Free Fire had record high monthly paying user which more than doubled yoy (since the peak DAU only grew 60% yoy, the conversion probably improved)
In India, monthly paying users/monthly active user > 10%
We use peak DAU to calculate a monetization metric, Average Adjusted Revenue per Peak DAU (AARpPKDAU).
Fire Fire Adj. Rev. has been driven by expanding user base, improving conversion (paying ratio), and increasing monetization. As long as you build, grow and sustain a large user base, you always can find/try various monetization methods. While Free Fire focuses more on the monetization now, keep growing and sustaining a healthy user base is one of the keys to turning a game into a platform (key words: engagement, community, ecosystem…). Sea’s current goal for Free Fire is to make it a social platform so the company might try more monetization initiatives to bring in new revenue streams.
The monetization metric – AARpPKDAU – is not accurate but still provides a snapshot on monetization trajectory. The metric is driven various factors so a jump might not be good and a decline might not be bad. For example, a jump in Average Revenue per User (ARpU) might come from more aggressive initiatives which might sacrifice overall user experience and churn light users who might be converted into a paying users. A decline in ARpU might be due to new market penetration that expands the user base but hasn’t brought in revenue yet. If there’s a jump in user base proxy along with a decline a ARpU, I tend to interpret it more positively.
💡 This note tries to understand the traction of Free Fire. Generally, I would start with metrics which are more top of funnel such as downloads, number of active users, conversion, number of paying users or even engagement metrics. However, I was triggered by a Sensor Tower blog post the other day which mentioned Garena Free Fire revenue in 1H2020. I was confused by my back of the envelop growth rate calculation because I only had some rough ideas on how well Garena Free Fire performed financially before putting together this post. The confusion probably came from that there’re few financial metrics available in the market – revenue from third party analytics, Non-GAAP adjusted revenue from Sea’s Digital Entertainment, and GAAP revenue from Sea’s Digital Entertainment. It seems I encountered some difficulties when collecting those data to understand Garena Fire Fire financial performance. This post shows the traction of price-driving metric – adjusted revenue and digs out the history of a parameter which is key to determine Garena Free Fire adjusted revenue but not being disclosed anymore.
As of the end of October 2019, Free Fire had recorded a total cumulative adjusted revenue of over US$1 billion since launch.
6-K Reports Third Quarter 2019 Results
Estimation: Free Fire’s Contribution to Adjusted Revenue in Sea’s Digital Entertainment (DE)
Launched on Dec 4, 2017
Accumulated $1B adjusted revenue (adj. rev.) by Oct 31, 2019 → roughly 23 months since launch or average $43.5M per month
1st Attempt to Estimate Garena Free Fire Adjusted Revenue Contribution
As of Oct 31, 2018, Garena Free Fire’s lifetime adj rev. contribution to Sea’s Digital Entertainment since its launch was estimated to be ~60% (see estimation in Window Matching and Accumulated Adj. Rev. in the Window toggle lists below).
Numbers from third party analytics are more timely and detailed but we need to check the relationship between the reported numbers from analytics firms and numbers from the company. Sea discloses two top-line metrics: 1) GAAP revenue 2) Non-GAAP adjusted revenue which takes the changes in deferred revenue in the period into account. Sea believes that the Non-GAAP measure can better represent the performance of the business.
If there’s wide difference and inconsistent between numbers from analytics firm and the company, the third party number might be less informative while it’s still an estimate to gauge the underlying performance. The difference and inconsistency might come from measures adopted, coverage difference, my mistake in merging/inferencing data from different sources. For example, I calculate Garena Free Fire 2Q19 revenue in Senor Tower by deducting cumulative revenue 1Q18-2Q19 by cumulative revenue 1Q18-1Q19.
From the charts below, it seems the discrepancy between Senor Tower and Garena Free Fire adj. rev. expands in tandem with discrepancy between GAAP and Non-GAAP numbers.
Sensor Tower’s 1H20 number implies a 11.5% yoy growth which is difficult to reconcile with Garena Free Fire adjusted revenue trajectory as the gap between GAAP/Non-GAAP expands when game with large user base stated to ramp up monetization.
A quick guesstimate on Garena Free Fire adj. rev in 2Q20 by observing the pattern is $375M (+$30M from 1Q20 which is based the assumption of 67.5% contribution) or around 41% yoy growth which implies yoy growth around 50% in first half of 2020. I believe adj revenue has more impact on price. While third part analytic is more timely, it’s too difficult for me to interpret.
Stick with Price-driving Metric, Focus on Long Term, and Believe Platform Brings in Optionality
So for this game, we’ll not look at any quarter-on-quarter or month-to-month or even year-on-year short-term performance, but we are really looking to build it into a long-term IP franchise.