First, writing stories on Olympic game results required data, and Toutiao pulled it from three sources: [a] real time score updates from the Olympics organization, [b] images from an image-gathering-company it had recently acquired to find relevant visual media, and [c] monitoring live text commentary about the game.
Second, Toutiao had to figure out how to combine data from these three sources to ensure an internally consistent and relevant story.
As a result of these efforts they were able to publish news story of a game approximately 2 seconds after the event ended as they collected and processed the information as the game progressed.
In Toutiao's case news is the element with which they launched but today there is a lot more reason why users use the app (much beyond news). Local news, weather, information on agriculture, etc. are surfaced regularly which is why the app is quite popular nationally in China and not just in Beijing and Shanghai.
Toutiao is now focused on video as it is increasingly becoming more important (more supply of video content and more consumption as well). Outside of Chinese language centric markets they have an investment in India (which they have announced publicly). India is definitely a market that could potentially benefit from algorithm based content discovery (several different states/languages and a long tail of content that is not easily available to users).
Perhaps to your surprise, many people in China are okay with certain freedom tradeoffs vs. improvements in other areas of quality of life. Its not an issue that you can easily simplify the more you familiarize yourself with it.
1: If so many people are ok with it, then why not take a vote? If they are ok with it, they will vote for the current government. If they are not ok, then this argument is false and the current government should be voted out of office. Either way, a vote is good for China.
2: If the citizens are ok with it, then why does the government have to censor criticism and non-government ideas, run massive propaganda operations on their own citizens, jail dissenters, and oppress Tibet and Xinjiang? How do you know if the citizens are ok with it?
3: Parts of China that do have democracy and human rights, Taiwan and Hong Kong, seem to like those things very much - it's not a popular idea in those regions, or even an idea at all, to trade in democracy and human rights for the mainland's political system; no Taiwanese candidate makes that pitch to the voters! Also, people in Taiwan and Hong Kong are far more prosperous per capita than those on the mainland (I'm excluding Hong Kong from the "mainland" in this case.)
> certain freedom tradeoffs vs. improvements in other areas of quality of life.
It's not a tradeoff, it's just a needless sacrifice. All the wealthiest, most secure nations it the world have democracy and human rights. It's not a Western thing; that includes the wealthiest regions of China, Japan, South Korea, etc. What is the Communist dictatorship providing in this tradeoff?
> Its not an issue that you can easily simplify
'It's complicated' is not an answer, it's the avoidance of one.
Hmmm ... trade is important, but it's hard to miss the overhwhelming correlation between democracy and prosperity.
Name a '1st world' nation that isn't democratic. But plenty of poorer nations trade.
One theory is that democracy provides greater political stability and the rule of law, and those are essential to prosperity. People are not willing to risk their life savings on a business when the government or some other powerful figure can just take it away, or if political instability will undermine their economic opportunities.
Plenty of poorer nations have some kind of democracy as well, although in many cases it's debatable how free their elections really are.
I think the correlation between democracy and prosperity is indicative of causation in the opposite direction: citizens of prosperous countries want a say in matters that affect them, and they have the time on their hands to campaign for their causes. Many countries have only become democratic when their economies were already prospering, e.g. South Korea, which was a military dictatorship for much of its initial growth.
I also don't think that democracy is the best way to provide stability, since the election cycle tends to create uncertainty on relatively short timescales. How are you going to build a business if the electorate might decide in a few years that they want to make it illegal, or impose stricter regulations, or demand higher taxes? For maximum stability, an autocracy with a slow but predictable bureaucracy would be much better. Of course, if that means your slow but predictable death, it becomes the much worse option, but few people seem to be concerned about that in China.
> few people seem to be concerned about that in China
See my original post, above. The Chinese government clearly is concerned that many people in China are concerned about it.
> I also don't think that democracy is the best way to provide stability, since the election cycle tends to create uncertainty on relatively short timescales. How are you going to build a business if the electorate might decide in a few years that they want to make it illegal, or impose stricter regulations, or demand higher taxes?
That's not how democracies work; they don't elect kings and queens, who can arbitrarily change laws. They elect officials to a wide variety of roles, all responsible to the electorate, who must agree on changes through legal processes. In the U.S., a majority of the House, of the Senate (and sometimes 60% of the Senate), and the President must all agree on a new law. Also, the law must pass muster in the courts, and be legal under the Constitution. Also, voters get a powerful say in what the laws are. If you don't like a proposed or actual law, you can speak up, build support, and influence the outcome - it happens all the time. Finally, the federal government's power is limited to federal issues; states and localities override federal policies or create their own policies all the time. You can see the current U.S. president's lack of success in his legislative agenda as an example of how democracies work.
In an autocracy, one person can arbitrarily change whatever they want - they are not subject to other laws, a constitution, representative legislative bodies, electorates, or any of those other influences.
And the outcomes are clear: All the most stable countries are democracies, and even when they were much poorer than today (such as the U.S. in the 18th century).
The Chinese government is concerned that more people will become concerned about human rights issues, not that there currently are many who care. Otherwise, censoring undesired opinions on the internet wouldn't be enough.
In your explanation of the way the US political system provides stability, the important factor is not that bureaucrats are elected, but that their process is slow. It would be equally stable if they were appointed by birthright or any other undemocratic process.
That voters get input into the legislative process is of course a good thing for them in aggregate, but the progress it creates diminishes stability. You can't just keep running your business the same way when the legal environment keeps changing.
Worse, the number of people who get input means that the outcome will be very uncertain. Maybe the law fails to gather support, or it is passed but later ruled unconstitutional, or it is amended after the next election cycle. You have to account for all of these possibilities. Things would be much more stable if you knew far in advance what will happen.
You have convinced me that a single person being able to influence everything might not be very stable either, depending on how fast that individual changes their mind. I now think that a large body of decision makers, appointed for life by any process (be it democratic, autocratic or random) requiring a large quorum of agreement for any decision, would be the most stable system.
If you really want to argue that democracy causes stability, your last sentence should have been: "All democracies are stable countries ...". But I'm not sure you could really argue that, considering the number of democracies who went through very unstable periods in their history. (E.g. civil war in the US.)
Requiring large numbers of people to agree to a change makes it predictable - I've seen few surprises from Western governments; whatever one thinks of Trump, his behavior is what he advertised in the campaign.
But whatever your theories, the data is that democracies do have the most stable and predictable business and legal environments, and more so than China where instability and lack of rule of law is a complaint of businesses, especially foreign ones. Almost all the world's leading businesses are based in democracies; the vast majority of business is done in democracies. Those monarchies that ran the world for millennia didn't perform so well.
It's not just predictability either; corruption is a huge problem. In places with lifetime appointments, what happens to corrupt leaders? They can't be voted out, and lacking the rule of law and separation of powers, they will not be investigated and tried (unless a more powerful leader causes them to be). And why should we think that lifetime appointees will care at all about the voters? The lesson is, power corrupts, and absolute power corrupts absolutely.
Countries without democracies, in fact, stagnate (such as China until the encounter with the West in the 19th century and again under the Communists (compare w/ Taiwan and Hong Kong), the USSR, N. Korea (compare with S. Korea) and many, many more). Also, by not giving citizens a forceful voice, they ignore and overlook citizens' interests, leaving serious issues unaddressed. Even the well-intentioned cannot centrally plan politics any better than they can centrally plan economics - nobody knows enough to understand and address the needs of hundreds of millions.
Finally, one reason democracies have more prosperity is that they are more dynamic. In autocracies, economic success tends to depend much more on access to those in power. The government will never allow a SpaceX to challenge the favorites of the powerful, Boeing and NASA; or allow Microsoft and Compaq to take down IBM; the Internet to take down the telecomm system (at least not until it was proven to be a necessity in the West), etc. It's because the autocrats have no reason to allow a challenge to their own power and every reason to prevent it.
I'm afraid this debate was settled long ago, around 1776 in the U.S., and monarchy and autocracy clearly lost.
I agree that overly stable countries tend to stagnate when that stifles innovation; I certainly didn't intend to claim that stability was universally good. After all, progress is impossible without the instability it brings.
However, I find it interesting that you mention the examples of PRC vs. ROC and North vs. South Korea as autocracy vs. democracy. In fact, after World War II, all four of them were military dictatorships/one-party states, none of them particularly inclined towards democracy. The only difference was their alliance to either the USSR (PRC and NK) or the US (ROC and SK). This helped ROC and SK to develop their economies (still under autocratic rule) and likely contributed to their eventual transition to democracy (while keeping their economic growth).
That would seem to indicate that economic development leads to democracy, and not the other way around. I tried to make this point in my previous comments, but you did not address it. I'd like to know what you think about that.
> I agree that overly stable countries tend to stagnate when that stifles innovation
That's not what I said. I said autocracies stagnate; wealthy democracies are the most dynamic, innovative economies, in business, technology, science, culture and everything else. That's where the Internet, web, Facebook, the startup business model, financial engineering, bitcoin, globalization, open source, rock'n'roll, hip hop, electronic dance music, personal computers and smartphones, modern 'deep learning' AI, and almost any other innovation you can name comes from. It's where science, capitalism, democracy, universal human rights, and much more came from.
Engagement alone is not success but sustainable and improving engagement (I.e. strong and improving cohort retention) is a measure of success because it creates a high barrier of exit for you as a user to stop using the app. 74 mins daily is not easy to create and sustain especially when you don’t have a social graph (On Facebook you are spending time talking to friends, reading updates from friends and content) but on Toutiao - you are only consuming content.
Good point. I believe Toutiao tries to show different content ocassionally to solve the exact problem you are describing. There may be new areas or other hot trending topics especially locally that you may need to be aware of and they use these attributes to push content. But this is something they continue to work on as well.
Good question. They did not have many social features at the start - especially in year 1 (no followers, no friends, no chat). Over time they added like, dislike, retweet, comments functionality.What made them really stand out in the first year was the catchy title, delicate product design and pretty much the first personalized news aggregator with recommendations in 2012. They also iterated heavily to remain as one of the top apps in the app store during the first year. In the first month - users gravitated towards the app as it became the destination to read all top trending news /stories that were going viral that day (it solved a gap in the market, no one else was doing it). Over time they recommended stories to users and tracked data to ensure only relevant stories were being pushed which helped increase engagement.
I'm a little disappointed in this post given China's penchant to controlling news/media/content.
I see how this could easily be manipulated into an extremely powerful propaganda tool - especially hiding behind a guise of "algorithms" are doing all the work - and yet it is not mentioned at all in the post.
We have already seen how content systems like Facebook, Google, and Twitter can be manipulated via botnets, but when a central power manages this it can obviously get significantly worse.
Does this concern you as an investor who lives in America?
I am afraid I cannot do justice to this question by answering it in a few sentences in this forum. Toutiao is one of the first apps in China to enable creators to write content and find audience without having to go through other media outlets. So in a way they are transforming the way information is made available to people. If you would like to discuss more email me directly at anu@ycombinator.com
I agree it's tough to do justice in a few sentences in a forum, which is why I assumed it would be done in the very in depth and otherwise interesting article you wrote :)
In my opinion it's the absence of this topic from the article that makes YC look tone deaf once again by glossing over a serious repercussion that technology can have on society for the sake of self-promotion -- especially given current events regarding the manipulation of social networks by foreign entities and a very serious history of censorship and media control.
All media are controlled by the government in China, not just Toutian. It certainly can be manipulated easily for propaganda. However, it is now a common problem even for a company like the Facebook's fake news problem.
You talk here about creating an "addictive" product. People in the industry usually use the euphemism "engagement" rather than "addiction" but either way, do you think that this type of product will eventually be treated by society the way cigarettes are now?
As someone not familiar with Toutiao, could you describe how the Q&A function works from a user perspective?
I'm guessing there's a part of the app where the user goes for Q&As. If a user asks "How do I flush my car's cooling system?", does the question get sent to all users who like cars?
Once a question is submitted, the AI engine automatically recommends a list of relevant experts/potential answerers to the questioner to choose from. As a user you can choose who you want answers from. In addition the system also pushes the question to relevant users based on their interest graph.
The system then also sorts the recommended answers based on its quality.
It's worth noting that the panel is largely made up of companies whose products seem strongly communal by nature. Slack, Pinterest, Instagram, YikYak, Facebook, and Twitter are social platforms. AirBnBs are shared with fellow travelers, Ubers with other riders (and on the supply side, driving as a side job or renting out a property seem likely dinnertable conversation topics). Stitch Fix, Fanatics, Square are the three that stand out as having a more 1:1 interaction between the company and the customer, where the nature of the business is a product/service being exchanged directly for $$ and isn't likely to rope other humans into the process.
Given this, not sure how well this early-growth-team framework generalizes. The article is a "Growth Guide: How to Set Up, Staff and Scale a Growth Program", but seems to be a model built to explain a somewhat homogenous data set, and I want to question how it would fare if applied as a template for early growth teams generally.
100% agree with the following:
- "Startups that have seen amazing growth have developed teams and processes that are intentional, exceedingly metrics-driven, and thrive on experimentation."
- "Scientific approach to growth"
- The retention checklist, picking core metrics, etc.
- Experiment dashboard, analytics tools, peer review, user research
- That all of the advice is well-suited to companies whose growth is (1) primarily driven by the product itself, esp. by referral/virality, and (2) for whom retention is especially important. (1) and (2) are certainly not true of all startups, and maybe not even true of most, though I'd trust YC to have a better statistical view on this than I do.
Less sure these heuristics are broadly applicable:
- Growth PM = first growth hire at a company
Depends on how we're defining "growth", but if we say "the first growth hire is the first person whose primary mandates are acquisition, activation, revenue, retention, referral", then it could make sense for a first growth hire to focus on e.g. paid acquisition if that is the highest-leverage growth driver early on. This was my experience at Upstart (ML-driven lending platform), where my mandate as the first growth hire was to do whatever it took to grow. It turned out that priority #1 was to build out targeted paid acquisition campaigns. If some flavor of performance marketing, or partnerships/sales, or content or in-person events are the strongest growth tactics at the early stages of scale, does it not make sense to build a growth team starting with those early hires, and add eng/PM/design/data science from there? This seems like a case worth mentioning if it's not extremely rare.
Perhaps this is just a semantic thing, where the panel would respond that the aforementioned first growth hires would would be called something other than Growth (i.e. you might build a 1-2 person Marketing or Sales team to deploy ad spend or build partnerships, and the Growth team would sit separately with Product)? Seems odd, but... maybe. I'm especially curious what YC has to say about this from observed patterns across many different business models and early growth team structures at companies in the 10-50 employee range.
- "70% of experts mentioned that referrals were the top channel within the first year."
This stat seems to follow from the homogeneity I mentioned; seems worth calling out that this doesn't mean it's 70% likely that any randomly selected growth-stage startup should focus on referrals in its first year. While any great product will be talked about and benefit from referrals, it's a stretch to say this will be the top channel for most startups, and I worry that the guide recommends a team structure that lends itself to virality-first growth when this isn't a template that applies as widely as the article implies
- The circle chart Sean called out (where 70% of the growth teams sit in Product)
I particularly disagree with this snippet:
"Traditionally, a company’s marketing team has been responsible for driving user acquisition (and the associated budget), so this is sometimes a default department in which to house a growth team. Often this evolves from prior functions that have lived in the marketing department (like performance marketing and user acquisition). In these cases, the Head of Growth would report to the Head of Marketing. The general sentiment about this approach is that the line of reporting is a bit rooted in the past, and most growth experts cited this as the least-favorable option."
I don't think the line of reporting is necessarily "rooted in the past" such that the orange 70% in the pie chart are the modern, smarter ones and the rest are the dinosaurs. Instead, I suspect that the type of growth those companies experience are more deeply rooted in the product itself, so it's natural for Growth to sit within Product, and this may be true of many modern companies. But it's not a function of modernity, it's a function of the nature of the product itself.
This is all coming from someone who has built growth frameworks largely from first principles (rather than direct mentorship from someone broadly experienced) with only a few years of personal experience, so I'm spelling this out as a request for someone to please point to any blind spots in my reasoning.
We struggle to attract the 'Ideal Engineer' to apply, most of our recruitment pipeline is filled with mediocre candidates. Do you have any tips or examples on how to communicate and attract the 'Ideal Engineer'?
Apply on your site or getting candidates you are interested in to actually interview? If you want the 'ideal' engineer, you need to identify them and go after them.
Hello Anu! How would you recommend selling these metrics to people focused on things like app downloads or user signups? Essentially, how would you recommend advocating for adopting these metrics?
It all comes down to knowing how a company makes decisions. What's the thing that sells. At most companies it is revenue, so you basically run experiments that correlate to revenue strongly and then use that causation to adopt new metrics. People like metrics that lead to their higher order KPIs a lot.
Great question. This really has to come top down and starts with the CEO. If the CEO urges the team to focus on the right metrics and asks questions on weekly reports then over time this will be in the DNA of the entire organization. Else it is really hard to get the organization to adopt these metrics
Great post and overview, Anu. Do you have any specific learnings for B2B companies and companies that generally have lower volumes of traffic for experimentation?
If you have only a small sample size to experiment with then you need to run experiments that can cause dramatic lifts (not minor changes like moving a link, etc.) Because if you have a sample size, then you need to wait much longer to measure the significance of these results and as a startup, time is not on your side.
I think if you can't get statistical significance than the growth team model is probably too early. You mostly need to be in the "invest" stage and not the "test" stage. That's not to say you can't try to hack merely that measuring through multivariate tests is a waste of time and effort.
Yes. We plan to! There are so many nuances with B2B that we felt it deserved a separate post. Having said that products like Slack (though B2B) is a lot similar to how a consumer app would scale. So it really depends on the type of product you are building
There were quite a few local Chinese competitors to WeChat when it originally launched. It took them almost a year to surpass Xiaomi's MiTalk user numbers.
no, not really, when they launched QQ was big and I was just surprised why some people have different status to find out it's wechat, but anyway QQ was dominating markets, everything else including Xiaomi or MSN had just leftovers