We have received your feedback.Thank you for taking your time in view of your busy schedule.
Why does this "anti-humanity" product get another $230 million in funding?
How does a content app avoid being restricted by algorithms?
Some information flow products have relied on machine learning to understand users' preferences and recommend content that they are interested in, which has led to great commercial success.
"Algorithm recommendation" has now become a standard in the industry. It seems that only by constantly "pleasing" users can we have a chance to retain them.
Now, there is a product that is not over-personalized but recommends the "most suitable" or "most desirable" content to users, which is also popular among users.
It is SmartNews, a Tokyo-headquartered news aggregation website, and app that's grown in popularity despite hefty competition from built-in aggregators like Apple News. And it has been the number one news app in Japan for multiple years running.
In addition, SmartNews has successfully expanded to the United States. According to AppAnnie data, as of July 2021, in the U.S. market, SmartNews achieved the highest monthly user time spent among all news apps.
In mid-September, it raised $230 million in its Series F round of funding. This brings the company's total capital raised to date to more than $400 million and secures its status as a "double unicorn" with a valuation of $2 billion, the highest for a standalone news app.
How does this "anti-personalization" and even a bit "anti-humanity" product make it?
How to use algorithms to fight the information cocoon room?
SmartNews and Toutiao were both launched in 2012.
Compared to the multi-functional and platform-oriented design of Toutiao, SmartNews' product page is a bit simple, it just classifies news content and then pushes it in the form of information flow.
The content that SmartNews pushes to users can be roughly divided into two categories:
One is "information that all users should know". This part will appear in the "Top News" column of the product homepage, including current affairs, social hot spots, and so on. All users will receive the same information.
The other category is "limited personalized content". Users can subscribe to the information that they are interested in. To some extent, this allows users to have a personalized experience, but the content that users can choose is "limited."
Because SmartNews only cooperates with authoritative media, publishing houses, and other institutions, and has strict controls on content sources. In the official introduction, SmartNews says: "We cooperate with thousands of media, use algorithms to analyze articles, and only 0.01% of the content are pushed to users."
This is where SmartNews differs from other personalized content products - the algorithm is different.
An information aggregation product that advocates "a thousand people with a thousand faces" uses algorithms to analyze the interests of a large number of users, tag much content, and then match the two to achieve personalized recommendations.
SmartNews uses algorithms to analyze articles and selects 0.01% of the content to find the set of greatest common divisors among the information that users "should know". In this process, many analysis variables are removed and the algorithm is subtracted.
Therefore, SmartNews does not force users to log in with social accounts, nor does it rely too much on third-party data to analyze user preferences. It is precisely to prevent users from receiving very personalized or even narrow information based on their own preferences and habits.
The firm also differs from other personalized content products in terms of how it sets KPIs for its algorithms.
SmartNews CSO Ren Yi told GeekPark (ID: geekpark): "Many companies naturally set click-through-related KPIs for their algorithms, which are short-term conversion rates."
The result is that the algorithm will recommend a lot of similar content to users over a period of time. Slowly, one's interests and preferences will become narrower and narrower.
In Ren's view, the main purpose of users is to broaden their knowledge and gain new thinking. Therefore, SmartNews sets longer-term KPIs for the algorithm, and the content pushed may not have high click-through rates in the short term, but in the long term, it helps users discover new knowledge.
Take the 2016 U.S. election as an example. The general algorithmic mechanism would recommend users reports with similar political stances, but SmartNews has added a "political balance algorithm" to the product so that users can also see content that is different from their inherent stance.
As a result, SmartNews is more like a "personalized discovery" product than a personalized recommendation.
"We want to bring users back to themselves and lead them to discover new things. The core of SmartNews is about personalized discovery, which is different from personalized recommendations", says Jeannie Yang, its senior vice president of product.
One of SmartNews' goals is to surface news that its users might not seek out on their own but might be interested in. So, the firm thinks it is important to free readers from the filter bubbles.
Where is the next stop for information aggregation products?
In order to break the "information cocoon room", SmartNews has also made a series of attempts in product interaction.
During the American presidential election, SmartNews launched a feature called "News From All Sides". First, it will track the election coverage in its entirety. Then, it will sort the stories by political stance and present them to users.
When users open a political story, they will find that the opinions of all parties are arranged side by side. In this way, users can see reports from different angles for the same current affairs news.
Ren said, "We don't want to turn ourselves into a referee, telling users what is good and right. Instead, we take a step back and show the user a panoramic view of the information. In this way, they can have a more general understanding."
In addition, for large public events, SmartNews will urgently open a corresponding Tap column to gather the most real-time credible reports.
Last year, the firm took only 52 hours to launch the epidemic news bar during the outbreak. Now, users not only have the latest outbreak information but can also easily find out where to get vaccinated. In addition, SmartNews has launched a hurricane tracker for weather disasters such as hurricanes.
Compared to general news reports, this information is more granular and is an aggregation of information specific to a particular area, or even a particular event. It is understood that SmartNews will pay more attention to users' health and safety issues, launching features such as tracking wildfire disasters and crime reports.
In this regard, Ren explained that the idea of SmartNews is to provide users with quality information. One of the criteria for "quality" is whether users can make a judgment and take a decision after receiving the information.
Therefore, SmartNews is not striving to cover the entire content category but to achieve the best experience in the key areas.
An industry insider said that in July this year, ByteDance was rumored to be developing a high-end version of Toutiao, perhaps in the same vein. According to reports, different from Toutiao, which covers information contained in various industries, the high-end version only focuses on relatively limited areas, including high-quality content such as business, culture, finance, and history.
Behind the verticalization of information, a larger trend is the servicification of information. This is an experience that SmartNews has tried to add to its existing product format.
In the Japanese version of SmartNews, users can search for discount coupons for a number of stores. This promotional information was first concentrated in a dedicated tap column, and now it is optimized to display based on geographic location. When users arrive at a certain place, they can open SmartNews and see what's on sale nearby and decide whether to visit the store.
Different from the aggregated apps that focus on content consumption, SmartNews is more like a tool that provides information services.
"If you just look at the window period, you may not be able to make meaningful products"
The establishment of SmartNews and Toutiao in 2012 was the beginning that mobile Internet companies use algorithms to transform news information distribution methods and reshape the production mechanism of the media industry.
In Ren's view, there are four generations of information flow products. The first generation of information flow is Yahoo, which uses directories to categorize websites. The second generation is the keyword search, which is represented by Google and Baidu. The third generation in the form of Feed was invented by Facebook and Twitter, based on the friend chain for recommendations, so that users can keep swiping.
Ren added that Toutiao and SmartNews are both considered the fourth generation. Although they are still in the form of Feed, the recommended content is based on more information outside the friend chain.
Although it started out with Toutiao, SmartNews did not have a hot product to form a series of product matrices, as ByteDance did, nor making more attempts in overseas markets.
SmartNews has a mature product model, systematic recommendation logic, and algorithm technology, but it is very cautious about exploring overseas. Although the firm has launched its products in more than 150 countries, the only markets SmartNews has really focused on are Japan and the United States.
Ren said: "news content is closely tied to the local community. According to SmartNews' product ideas, we must first maximize the collection of local high-quality media content to form a complete PGC content ecology, and then install these content into SmartNews products."
"In India, people think that there are 10% English-speaking users in India, and these users contribute more than 60% of the advertising revenue. Five years ago, ten years ago, that was the case. But now the middle class in India is rising, and this user group may have 13 or 14 languages. If you're going to do an Indian version, you are going to challenge those languages."
"There are also different places with different reading habits. Some people may be used to reading from the left, some from the right, so should the interface change? High-quality information access may be different for users in each place."
"When we judge that we have the ability to do well in this country's market, we will enter a little more resolutely." Ren added, "Although there may still be many window periods in other overseas markets, you may not be able to make meaningful products if you just focus on the window period."