WhaTech Weekly 11/08/20
WhaTech is back: what can we learn from TikTok success, Ant Group IPO delayed, Prop 22, Social Networks during US elections
Good morning, I changed the format of this newsletter by starting with a specific topic I wanted to dig into, followed by a few news from the week. I like to learn and to share what I know so I hope you enjoy it!
What can we learn from TikTok success?
I recently listened to the A16Z’s podcast TikTok & Beyond: The Algorithm Question, The Future of Product (published on 09/20), and I thought it was interesting to share with you the main insights and to explain what makes TikTok so successful.
How is TikTok different from other social networks?
First, let’s acknowledge that TikTok’s main purpose is not to connect people but to entertain them, and should then be called an entertainment network rather than a social network. Yet, it still makes sense to compare TikTok with social media such as Facebook and Instagram, in addition to other entertainment networks like Youtube.
TikTok differentiates from traditional networks in 2 ways:
TikTok isn’t based on a social graph. Like traditional social networks using a “social graph”, TikTok can also show you the content of users you follow, but this tab is secondary and only gets a fraction of the traffic the main tab (FYP, '“For You Page”) gets. The primary content you see is not based on a social graph but is based on your activity: TikTok shows you the content you like based on how you watched videos (loop-watched, liked, shared, didn’t watch the whole video, …). The big difference with TikTok is also in its algorithm automatically finding what content to show without even asking about your interests or without taking a look at your friends’ history.
Everyone can go viral with TikTok as the algorithm will put videos forward depending on the content of the video and not on how many followers the user has. TikTok actually shares each video to a small test audience and decides whether to push it or not depending on how the test audience reacts. As a result, TikTok adapts and pushes you the right content even when your tastes change over time.
TikTok Competitive Advantage
TikTok’s main competitive advantage is not its algorithm which is based on traditional thinking and could be copied, but it is the huge amount of training data allowing its machine learning algorithms to always push better personalized-content to each one of its users (the content that fits the most what each users want to see). While there is no such public training data, TikTok can rely on 6 years of data generated by its users (starting in 2014 with Music.ly, bought by ByteDance in 2017 to become TikTok).
And the magic of TikTok is not just having training data, but it’s having clean training data, allowing to better analyze it, to get better feedback to train machine learning algorithms, and to finally push the best content each user want to see:
TikTok relies on a huge operations team having people tagging each video to identify the content in it. It’s now very difficult for competitors such as Instagram to catch up, especially as visual AI is not ready yet to identify what videos are about.
TikTok benefits from an algorithm-friendly design: as users can only see 1 video at a time, the algorithm can perfectly understand whether the user watched the video or not (unlike Twitter where users see several tweets on the same screen, making it much more difficult to know what tweet the user reads).
Though I’m not sure I consider it as a real competitive advantage, another great feature of TikTok is its video tools such as editing functions, filters, or camera tools. It made video editing easy and fast allowing anyone to create high production film effects with just a smartphone, which was inconceivable before.
TikTok benefits from creative network effects
One of the most interesting parts of this podcast was about network effects: Eugene Wei, talked about “network effects in terms of creativity” as each user makes the rest of the community more creative. He explained that TikTok solved the “blank page problem”, as everybody can (and is encouraged to) borrow someone else’s idea, copy and remix it.
Traditional network effects are huge sources of competitive advantage, do you think that this creative network effects could also be one?
Here is the podcast if you’re interested in listening to the full discussion:
How do Silicon Valley companies thrive? An example with tech companies and AB5 law.
At UC Berkeley, my amazing teacher in innovation always used to tell us that Silicon Valley companies weren’t afraid of breaking laws and rules, because they could threaten governments to withdraw from the region, leaving all their contractors unemployed. That’s exactly what happened over the past few months with Uber and its peers:
In January 2020, the law AB5 took effect in California and classified anyone doing freelance work as an employee, with the main goal to give better work conditions to gig workers (from Uber, Lyft, Instcart, and others).
In August 2020, Uber and Lyft threatened to shut down their operations in California, which would leave approximately 500,000 drivers unemployed.
As a result, Californians were given the choice to vote in November for Prop 22, which aims at making drivers and delivery workers exempt from the law AB5. After tech companies like Uber, Lyft, Doordash, or Instcart spent $200m and spammed us with dozens of emails & notifications to vote for Prop22, the vote finally happened this week and Prop 22 passed.
AB5 is therefore a complete bust as Tech companies are not impacted while other freelances such as an independent musician can’t find work anymore. It’s also a failure because gig workers don’t get better work conditions (no healthcare benefits, no sick leave, …). That’s why, many of us think that another status adapted to new business models should be created in addition to the standard worker vs. freelance model. As Nicolas Collin puts it, it’s Time for a new social contract for the gig economy.
Ant IPO delayed
Ant Group was expected to raise $37bn at a valuation of $310bn by listing on Hong Kong and Shanghai stock exchanges, making it the largest IPO ever. However, the Shanghai stock exchange suspended the stock listing on Tuesday as Ant does “not [to] meet the listing conditions or disclosure requirements”, and Hong Kong followed by also delaying the IPO. While the exact reasons for the IPO delay are not clear, this could be linked to a conference a few days ago during which Alibaba & Ant co-founder Jack Ma criticized the Chinese financial system. If you want to dig deeper into this story, I recommend these two articles, on why China halted the IPO and what’s next for Ant.
The hard task of social media during the US elections
2016 US Presidential elections shed light on how bad the impact of social networks could be. That’s why social media recently implemented a few changes to limit their (negative) impact on US elections by for example flagging posts with misleading content and banning or reducing political ads.
But the specificities of the 2020 elections brought a new challenge to social networks. Due to COVID and the push to vote, postal voting was very popular this year. As a result, it took days to count all votes and social networks had the hard task to prevent any misleading information from spreading, especially any post that prematurely declares the victory of a candidate. To prevent this, Facebook & Instagram displayed warning messages stating votes are still being counted, Facebook & Twitter attached warning labels to misleading posts, and WhatsApp & FB Messenger limited message forwarding.
Thank you for reading me :)) Thanks to JoJo for editing!
Bonus: checkout my comprehensive guide to understand SPACs and their 2020 surge.