In the field of early education hardware, Huohuotu was one of the earliest entrants. In 2010, Huohuotu took the lead in launching the first children's story machine, with sales exceeding 300,000 units that year.
In 2014, the number of players in early education has increased, and the form and function of early education machines have gradually evolved from screenless to screened, from the main entertainment companionship to the emphasis on entertainment early education.
With the development of artificial intelligence technology, early education robot forms began to appear in 2015, and the price gap is very large, as low as one or two hundred, as high as two or three thousand.
But after several years of development, the intelligence of early education robots has not achieved a big breakthrough: the speech dialogue is rigid, the recognition accuracy is not enough, and AI technology is limited to simple dialogue, on-demand and spoken language evaluation. High-quality content for children on robots is still lacking, and most intelligent hardware vendors do not have the ability to produce and operate content, and can only introduce third-party content. The problem of homogeneity is prominent.
In other words, compared with the original early education machine, the early education robot is not much different in nature.
The market response is also obvious. In the past two years, the growth rate of HuoHuatu has dropped to 10%. Last year, the smart companion robot, which reportedly achieved 1 billion yuan in revenue and volume, is now facing the bottleneck of growth.
Fortunately, the market for early education robots is still large enough. On the one hand, more and more parents are willing to invest in early education; on the other hand, early education robots are still better interactive forms of triggering content than mobile phones and tablets. What's more, when the field of early education robots is far from the emergence of industry leaders, no one has yet defined what the functions and forms of early education robots should be.
After removing the hustle and bustle of the concept of AI, whether it is early childhood education product companies or artificial intelligence technology companies, they have begun to calm down to polish their content ecology: who can make enough rich and personalized content, who can operate the content ecologically Well, then whoever is likely to seize the opportunity in the early education market.
Robots "not smart"
How can we count early teaching robots?
For entrepreneurs on this track, a basic consensus has been reached on its definition: it has the ability to interact with voice and can connect to the Internet. These two characteristics distinguish the early education robot from the early education machine.
At present, most of the early education robots on the market, the standard voice interaction functions are voice wake-up, simple question and answer, more complicated with on-demand, intelligent voice evaluation, and so on. However, most early-education robots are not satisfactory in terms of speech recognition rate, accuracy rate, and personalized performance of the experience. For example, children whose tongues are not clear enough, do not wake up the robot; when they ask questions other than the program setting, the robot will answer unanswered questions; intelligent on-demand is often not what you want.
The AI interactive experience is difficult to say in essence, because the voice technology of the entire industry is still in its infancy. Early education product companies generally do not have the ability to do AI on their own, relying on AI voice solutions from third-party Internet companies. Duan Haiyun, general manager of Niu Tingting believes that in terms of AI voice technology, Baidu, Xiaomi, Ali, and HKUST Xunfei have no substantial differences and no one has a clear advantage.
For early education product companies, the difference is in the degree of cooperation with artificial intelligence solution providers. The higher the degree of close cooperation and greater investment, the better the overall effect. In this regard, the smart partner thinks that he has a unique advantage. CEO Wang Fanfan said that the smart companion, voice company, and cloud company have built a private cloud platform in depth, and have their own children's speech recognition team, which can do better in children's speech interaction.
But in general, under the background of immature industry AI technology, it is difficult for product companies to erect barriers to AI interaction. Duan Haiyun believes that at this stage, AI functions are flashy. Once the desired effect is not achieved, the child will not use it again.
"Moving" is not easy
There is an open secret in the early education robot industry: to make children's educational robots, many companies have ready-made hardware models, which can be modified slightly based on the models, and can be sold after being branded. Need to change the wake word and APP, and development can be completed in about one or two months. Therefore, the problems of homogenization and poor quality of early education robots are prominent.
In the form of children's robots, the humanoid robots that can make a difference are the forefront. From 2015 to 2018, Ubisoft launched four Alpha series humanoid robots and Goku robots. In addition to being able to interact with voice, these humanoid robots also have limb movements, can dance and perform better.
To make the robot move, the core lies in the development of the steering gear. You must choose a steering gear for many years. Some core components are self-developed and processed. They are rich in resources and have strong financial strength.
But very few companies have this R & D capability. And even if the technology is more mature, it is difficult to reduce costs. Zhong Yong, head of UBI Education, said that the high cost is due to the immature supply chain, and the immature supply chain is essentially due to insufficient demand. Excellent selection of Alpha series robots, the price is 2999-9999 yuan.
Costs cannot be reduced, and products are difficult to sell. And, apart from feeling better about interaction, Alpha robots are not much different from ordinary early education robots in educational content. The Alpha series of robots produced by Youbi in the early years have average sales.
Bottleneck of sales growth
After the wind of AI has blown, the sales of early education robots are not as hot as the market has been. Before the fire, Alpha Egg Robots, Smart Companions, and Fire Rabbits have faced bottlenecks in sales growth in recent years.
The sales of classic story machines launched by HuoHuotu in 2010 exceeded 300,000 units in that year. In 2014, the sales of series story machines exceeded 1.5 million units, ranking first among children's smart toys in Tmall and Jingdong. In 2016, Li Xin, the founder of Huotu Rabbit, had expected product sales to exceed 2 million.
But in the past two years, the sales growth of HuoHuotu has slowed down. Xu Baolong told reporters that the annual growth rate of Huohuotu has dropped to about 10%. In Huohuotu's Tmall flagship store, the best-selling category is its story machine G6 series launched in the early years, with monthly sales of 11,000 units; and the recently launched early-education robot J7Pro, with monthly sales of only 390 units.
"In the past, we just made products and then sold them casually. Now we need to reform." "Sold casually" refers to relying more on word-of-mouth from users, without spending too much effort on marketing. Xu Baolong said that Huohuotu is currently promoting the standardization of marketing and operations. Specifically, each process of sales, including how well the dealers are capable, whether the coverage is scientific, and what stores sell what products, must be detailed. Currently, HuoHuotu's sales channels include mother and baby chain stores, e-commerce and provincial and municipal distributors.
In terms of channels, smart partners have always been considered to have advantages. According to Wang Bufan, Zhipan has three major sales channels: self-operated telemarketing, offline channels and WeChat. Among them, Wechat is the largest channel, and Smart Partner now has more than 100,000 small agents. In terms of offline channels, Wisdom Partners has its own counters in Wal-Mart supermarkets, Kids King, Xinhua Bookstore, and airport retail stores, mainly in second-tier cities. In the future, it will sink to third- and fourth-tier cities, and it is expected to enter more than 10,000 stores by the end of this year.
Although the channel advantage of Smartmate is well-known in the industry, many people hold different views on this selling method. Zhong Yong believes that the wide distribution of smart partners channels actually reflects the problems of low product repurchase rate and low retention, and can only increase users by constantly pulling new ones. "In such a strong channel, in fact, he still can't push it, and his performance is declining. After expanding the channel and eating out the demographic dividend, it will be gone." Zhong Yong said.
Grab content ecology
The technology and supply chain are immature, and the interactive properties of robots are difficult to break in the short term, and the polished content ecology naturally becomes an export.
In response, some early education product companies have already acted.
In 2010, HuoHuotu, an early-education intelligent hardware manufacturer, started production. Initially, the content team consisted of only two people. Back then, it was the age of the PC, with less content suitable for children. Some parents left insufficient content on the website, and Huohuotu slowly began to expand the content team.
Today, HuoHuotu has a 30-person content team that invests 67 million in content each year. It takes almost 6 months to produce 10 episodes of animation.
The richness of content is regarded by Xu Baolong as the core barrier that distinguishes Huohuo Rabbit from other early education robots. On the one hand, Huohuotu ’s self-developed content accounts for 30%, and has a good playback volume and word of mouth: “Huohuotu Illustrated Picture Book” audio has accumulated 13 million playbacks, and Himalaya and CCTV children ’s channels have purchased some Fire Rabbit self-developed content.
On the other hand, Huohuotu and Himalayan have reached a cooperation to introduce all children's content in Himalayan for free. Xu Baolong said, "The content of HuoHuo Rabbit is wider than all its peers."
Although the income contributed by content is small compared to cost, Xu Baolong believes that in the long run, content can build word of mouth, and "good children's content can be stored for ten years."
Smartmates also reload content. Wang Fanfan said, "I want to make today's headlines for children's content."
In the past two years, Smartmate has invested more than 80 million in content, and self-developed content accounted for 30%. In addition to increasing content richness and differentiation, Smartmates are also working to personalize content.
Wang Bufan believes that there are two advantages for smart partners to do this: On the one hand, there are 5 million app users and a large amount of user data. On the other hand, smart partners and some cloud companies and voice companies have built private cloud platforms instead of traditional end-to-cloud solutions, so they can achieve deeper operation of user data, such as personalized content recommendations and more accurate users. Portraits, product iterations based on user feedback, etc.
In addition, the smart partner content platform has also recently been adjusted to pay for the main content.
But for ordinary smart hardware vendors, it is difficult to make content. Whether it is to buy copyright or self-developed content, you need sufficient financial support. In addition, because of the lack of data accumulation, the marginal cost of making content for small companies is also high.
Positioned as a "children's audio content distributor", Niu listens, focusing on content operations, instead of self-developing the content, but doing a good job of organizing, filtering and recommending the content, so that the content can be distributed and used more effectively.
Niu Ting Ting, which focuses on children's smoked listening functions, has iterated over several versions, from only listening to listening, to increasing intensive listening and intensive reading. At the end of last year, it also began an attempt to check in the community.
Duan Haiyun said that the community punch-in function was communicated with the early education institutions. Niu Tingting has always wanted to cooperate with educational institutions. He has observed the needs of English institutions and found that most teachers will arrange ear grinding tasks after class, and they are generally distributed on the parent WeChat group. The problem with WeChat group assignments is that teachers can't get effective feedback, because children don't know if they actually listened.
Based on this, Niu Tingting launched the punch-in community, where teachers arranged the lessons in a rhythmic manner, and all lessons were recorded in a follow-up card. When the child is at home, the card can be heard by swiping the card on the listening device. At the same time, the listening record is generated. After listening, there are also Q & A interactions and game sessions.
In addition, Niu Tingting is also the first early education machine with online question answering function of the app, and the response speed of customer service can reach 30s.
Including excellent must choose, also began to attach importance to the construction of content ecology. However, Zhong Yong believes that the ecology of domestic children's content will take at least two or three years to mature. The preferred method is to no longer position humanoid robots as children's companions and do not do pan-education, but to customize the content for vertical scenes, such as serving autistic people and increasing the value of humanoid robots. .
How to raise the ceiling?
The content ecology must exert its strength, but at the same time, the shape and function of early education robots still have a lot of imagination.
Xu Baolong said that in the future, the products of Huohuotu will become more and more detailed, such as story machines, video machines and robots according to age. Robots don't just add "speak" to the story machine, they can also read picture books. From listening to audiovisual reading, the use of children's robots and age groups will be more abundant.
Picture book robot is also a form that many early education product companies are exploring. At that time, the competition will be whose robots can read more picture books, more accurate page recognition, better reading aloud, and more personalized service. It is still a contest of data precipitation and operational capabilities.
"When it comes to children's watches, you will think of small geniuses, but what about early education robots? No one has sat at that position," Duan Haiyun said. If you really make educational value, maximize the value of listening books and form your own complete closed loop, the market space will be very large. "