Die nächste Aufgabe von Yuanfudao

No one can stay in one place forever, especially when the ground beneath their feet is already loosening and collapsing.

After the implementation of the "double reduction" policy, Yuanfudao, which once boasted hundreds of millions of users, gradually withdrew from the core of the tutoring industry. It has since developed a new framework through its educational hardware product, the Xiaoyuan Learning and Practice Machine.

In less than two years, the sales volume of the Xiaoyuan AI Learning Machine exceeded one million units, almost single - handedly opening up a niche in the learning and practice machine market. On Tmall's product page, there is a prominent slogan stating, "For every 10 learning and practice machines sold, 8 are from Xiaoyuan." In offline stores, the "Xiaoyuan Learning and Practice Machine" and the "Xiaoyuan" brand have started to replace the original "Yuanfudao" in customer - facing interactions.

Clayton Christensen wrote in The Innovator's Dilemma that the power of disruptive technology often does not lie in its strength. It enters the market's niche in an incomplete and immature form, attracts marginal users with certain characteristics, and ultimately erodes the existing market structure.

Yuanfudao's bet on the learning and practice machine is, to some extent, a replication of this path. When entering the educational hardware market, it did not engage in direct competition with iFlytek and Xueersi in the mainstream learning machine market. Instead, it deliberately avoided the intense competition and quietly entered through an emerging niche - the "learning and practice machine."

Strictly speaking, the learning and practice machine is not a "superior" educational hardware but a "sufficiently functional" and more specialized alternative. It does not attempt to challenge the more comprehensive and content - rich learning machines. Instead, it precisely targets the low - value but high - frequency and essential scenario of "doing exercises" and has established a strong substitute value in the "practice" function dimension.

Currently, the Xiaoyuan Learning and Practice Machine has built a certain level of user stickiness, especially excelling in specific scenarios such as "replacing parents in supervising homework" and "automatically organizing wrong questions." However, its limitations are also clear. The nature of disruptive innovation means that it will always face doubts in the mainstream market as a "semi - finished product."

If the previous achievements can be regarded as a phased validation of the product form, it remains uncertain whether Yuanfudao's next stage will be to achieve structured growth or be overdrawn by the potential energy of delayed validation.

The "Absorbed" Learning and Practice Machine

Specialization allows a product to establish a clear position in a niche market and, to some extent, avoid the intense competition in the mainstream market. However, it also inevitably bears more "specific" pressure from user scenarios.

On social media, many parents have reported a gap in their experience with the Xiaoyuan Learning and Practice Machine. "During the final review period, there were wrong questions every day, but many of them were not actually mistakes made by the children; it was the system itself that made errors. The scoring is very rigid. If a child makes a mistake in one blank of a fill - in - the - blank question, the whole question gets zero points." Some parents also bluntly said, "What's most annoying is not that the correct questions are marked wrong, but that the wrong answers mislead the children."

Answer errors are not unique to "Xiaoyuan." In a physical test conducted by Guangzi Planet on several learning machines, including those from Xueersi and BBK, there were also cases where the standard answers did not match the system's grading. It is not uncommon to see two "correct" options in a single - choice question.

However, users are often more forgiving of learning machines with rich functions and comprehensive content. With a large number of courses, extensive learning materials, and well - produced videos, users may only frown slightly when the grading system occasionally malfunctions.

The logic of the Xiaoyuan Learning and Practice Machine is quite different. It has placed almost all its value bets on the single scenario of "practice." The more specialized the scenario, the lower the user's tolerance for errors. Therefore, in the eyes of users, once an error occurs, it is not just "a problem with a certain function" but a fundamental "flaw" in the product.

On the other hand, a salesperson of learning machines told Guangzi Planet that the learning and practice machine is essentially just a "question - brushing tool," and the "learning and practice" function is gradually being integrated into the learning machine systems by mainstream manufacturers. iFlytek has a practice mode, and BBK has directly integrated the learning and practice machine system into its learning machine, allowing for one - click switching at the system level.

The underlying logic is that the demand and form of intelligent educational hardware are not yet fixed. What users buy at the moment is not crucial. What really matters is what the product in their hands will eventually evolve into.

If the integration of the learning and practice machine by the learning machine is a form of downward compatibility, then conversely, the learning and practice machine can also evolve upwards. Take Xueersi as an example. Although its learning and practice machine was initially positioned around the "learning and practice" function, with continuous OTA updates, its content has now highly overlapped with its own learning machine, except for the preschool module.

However, this evolutionary path is not easy for the Xiaoyuan Learning and Practice Machine.

"Among the mainstream models, we are the only ones using an e - ink screen. The screen does not emit light, and the visual experience is similar to that of a paper book, with excellent eye - protection effects." At the Xiaoyuan store, the salesperson repeatedly emphasized the "eye - protection" as the core selling point. To compensate for the lack of visual performance, in addition to the black - and - white e - ink screen, it has also launched a color e - ink version.

In actual tests, the Xiaoyuan e - ink screen performs well in terms of writing smoothness and response speed during the question - brushing scenario, and the experience is not sluggish. However, once entering the built - in courses or involving complex interactions, the shortcoming of the low refresh rate of the e - ink screen inevitably appears, bringing with it picture delays and a sense of stuttering.

At present, the refresh rate of most mainstream e - ink screen devices is maintained at around 30Hz, and some advanced products can reach 60Hz. According to a staff member at the Xiaoyuan store, the refresh rate of the e - ink screen used in their product is only 10Hz, almost at the level of a static reading device.

In the context of the increasingly rich content and multi - modal interactions of learning machines, there is a natural tension between the technical logic of the e - ink screen and the mainstream trend. This makes it difficult for the "Xiaoyuan Learning and Practice Machine" to "extend upwards" in its product form. Even if its content and functions intend to move closer to those of learning machines, the hardware itself has set an insurmountable obstacle.

Xiaoyuan clearly realizes this problem. As the boundary between the "learning and practice machine" and the "learning machine" becomes increasingly blurred, it has quickly launched a new - style learning machine focusing on in - depth content and advanced capabilities, trying to get rid of the product label of a "question - brushing tool." It is worth noting that Xiaoyuan, which once promoted eye - protection as its biggest selling point, finally chose an LCD screen when moving towards the content - and - AI - intensive market.

Disruptive innovation is like a knife that cuts open a specialized demand. However, whether it can truly reshape the market structure depends on whether it can establish systematic advantages after the initial cut, rather than being quickly covered and replaced by more complex and integrated products.

For Yuanfudao, the learning and practice machine was once a tool to break the situation. However, this once - new species is gradually being absorbed and integrated into the learning machine ecosystem and is losing its clear boundary as an "independent product form."

At this moment, Yuanfudao is standing on the edge of a knife. Whether it will take advantage of the situation to establish standards, expand the market, or stop at the initial dividends and be re - integrated by mainstream products will determine whether it will be the namer of the "learning and practice machine" era or be submerged in the vast current of learning machines.

A Wide Question Bank, but It Needs to Be Deep Enough

As a pioneer in the category, Yuanfudao has a natural psychological advantage in the learning and practice machine track and is naturally reluctant to fade into obscurity.

However, for a player that started with a popular tool and gradually penetrated into the online education market, the waters of the educational hardware industry are much deeper and murkier than expected.

Xiaoyuan is proud of its "question bank advantage," which is repeatedly emphasized at the sales end. According to the salesperson at the store, "Three - tenths of learning is theory, and seven - tenths is practice." 70% of a child's knowledge acquisition occurs during the practice process. Xiaoyuan not only has a wide - ranging question bank but also supplies question sources for many learning machine brands.

However, an educational technology practitioner told Guangzi Planet that the core value of a learning machine does not stop at the breadth of the question bank. Its essence lies in two aspects: on one hand, whether the content design, interface interaction, and positive feedback are sufficient to make children willing to "use" it; on the other hand, whether it can drive personalized learning based on data and make the product truly "useful."

There is a misunderstanding that "hardware penetration rate" is mistaken for "learning effect conversion rate." In fact, sales data can only prove market acceptance, not an endorsement of the learning effect. For every learning machine sold today, it may not necessarily mean a child's academic improvement. It is more likely to be an extremely anxious parent or a self - confirmation of "whether I am doing enough for my child."

In the aspect of personalized learning, it is not just about pure technical ability. It also depends on whether each player has access to a sufficient amount of high - quality learning situation data.

According to the practitioner, learning situation data can be mainly divided into two categories: one is micro - data at the individual level, which comes from students' operations such as question - searching records, course - booking behaviors, and AI interactions during the use of learning machines or products within the brand's ecosystem; the other is macro - data at the regional level, which is mainly obtained by players entering the school scenario as ToB technology service providers.

"For example, providing a complete set of intelligent grading systems and supporting hardware to schools to help teachers grade homework and generate learning situation analysis reports. This reduces the workload for schools, and for the company, it can obtain learning situation data to support the research and development of its AI models and teaching materials. If 'everyone stumbles at the same place' in a certain region, the wrong - question recommendation will be more targeted."

In other words, if learning effect is used as the yardstick, simply brushing questions is far from enough. The real competitiveness lies in the ability to achieve "precise matching." "Localization" is not only a key support for this ability but also an important dimension for learning machine players when promoting their products at present.

Take iFlytek as an example. Its offline stores prominently display cooperation cases with many local primary and secondary schools on standing signs, such as big - data - based precise teaching systems and smart classrooms, highlighting its advantages in localization capabilities and the integration with in - school teaching systems.

Although other players may not be as "high - profile" as iFlytek in terminal promotion, most of them have also incorporated local content into their products, but in a more restrained way.

Taking the Chengdu examination area as an example, according to our incomplete statistics, as of June 23, iFlytek has included 5498 sets of local test papers, and the number of newly added questions nationwide this month has reached 9393 sets; Xueersi has included 1916 sets of local test papers, with 4420 newly added questions nationwide this month.

In contrast, the "phase assessment" function of the Xiaoyuan Learning Machine only supports provincial - level division, lacking options for specific cities or regional examination areas. The time dimension only covers the past three or five years, without monthly breakdown.

It can be inferred that localization construction and the frequency of question bank updates may not be the top priorities in Xiaoyuan's product system. The so - called advantage of Yuanfudao's question bank may not be as strong as presented in the sales pitch.

Online Players, Making Up for Deficiencies Offline

For Yuanfudao, which has shifted to educational hardware, another difficult hurdle to cross is the sales channel.

With its online education background, Yuanfudao is different from players rooted in offline education, which have classrooms, teachers, and a "human touch." Once it enters the hardware market, it naturally needs three things: channels, scenarios, and human - to - human connections.

According to media reports, as of 2024, Yuanfudao had only about 300 offline stores, so it has formulated a "doubling expansion" plan this year. However, based on the incomplete statistics of Guangzi Planet, Yuanfudao still has room to catch up in terms of offline store layout.

Taking a certain area in Tianhe District, Guangzhou as an example, within a 10 - kilometer radius, there are a total of 6 stores and points for the Xiaoyuan Learning and Practice Machine. In contrast, there are 12 for Zuoyebang, 25 for Xueersi, and more than 50 for iFlytek.

Of course, each brand focuses on different cities. However, from a horizontal comparison across multiple cities, the offline coverage ability of the Xiaoyuan Learning and Practice Machine still lags behind that of established players such as Xueersi and iFlytek.

Building offline channels is a long - term and painstaking battle. However, the educational hardware market is currently in a high - speed and cut - throat competition. Yuanfudao clearly needs a faster way to break through. Therefore, Yuanfudao has chosen to invest heavily in advertising for the "Xiaoyuan Learning and Practice Machine" in elevators, a narrow, private, but high - frequency field in modern cities.

A relevant person from Focus Media told Guangzi Planet that its elevator media is divided into three categories: large - screen advertisements in the waiting area outside the elevator, static frame posters inside the elevator, and smart screens inside the elevator. The former is sold in building - based packages, while the latter two support refined placement by location. Taking the smart screen inside the elevator as an example, a 15 - second advertisement on a single screen is played about 300 times a day, and the placement cost for a week is about 25 yuan.

Yuanfudao has always been generous in this regard. The person revealed, "During the years of running tutoring courses, Yuanfudao's annual advertising investment was between 600 million and 800 million yuan. Although the budget has been reduced, the annual investment still remains above 100 million yuan."

It is worth noting that Yuanfudao's elevator advertisements do not just stop at brand exposure like some liquor companies. They come with a clear e - commerce entrance, which is more targeted. The core intention is obviously to convert the exposure into online traffic and promote immediate transactions.

Obviously, this is not just simple exposure. It is an attempt to force a gap through "elevator media" and reconstruct a "consumption field" that was originally carried by bookstores and mall stores.

This path is relatively easy and may fit the "quick - win" logic of the online education era. However, the educational hardware business is still a heavy - duty one. Currently, brands are competing fiercely by launching online models, and the price war is intensifying, continuously squeezing the profit margin. Perhaps only by reaching real users through offline scenarios can the unit price be increased and the brand premium be supported.

It can be seen that when shifting from online education to educational hardware, Yuanfudao still needs to make up for its deficiencies. However, from breaking out of the "learning and practice machine" framework and actively entering the mainstream learning machine market to increasing investment in offline channels, it can be seen that Yuanfudao