With the popularization of AI tools, how to attract and retain users has become a major problem for AI startups. However, the high cost of customer acquisition and fierce market competition have put pressure on many AI entrepreneurs. This article will explore the dilemma of AI startups in the battle for traffic, and how they are finding new ways in the ever-changing market environment.
Sam Altman, founder of OpenAI, believes that the AI era will create a new kind of startup, where one person can become a billion-dollar unicorn.
Now, this era has not really arrived, and AI entrepreneurs are first blocked by the high cost of streaming.
In the words of AI entrepreneur Zhao Tian, "In 2023, everyone is talking about technology, products, and scenarios, and in 2024, there will be two things: growth and making money." ”
Zhao Tian originally worked in a large domestic factory, and started his business last year, and his products are related to AI image generation. The early stage of entrepreneurship went smoothly, and he successfully completed the A round of financing. Since October last year, he has done some exposure-type promotion on Bilibili, Xiaohongshu, and Douyin, and the conversion effect is considerable, and the CPM (Cost Per Mille, cost per thousand impressions) is about five yuan.
"I can't say that our products are doing well, but in 2023, users are curious about AI tools and are willing to try them, and as long as they are exposed to the right people, there is a high probability that they will be converted." Zhao Tian said.
However, with the high-profile entry of the dark side of the moon, the threshold of this game, which was originally small and big, suddenly became unattainable. The products of unicorn companies and Internet giants have absorbed most of the traffic and attention, and pulled the price of streaming to an unprecedented price, and ordinary entrepreneurs have either gritted their teeth to keep up or left the market early.
They cast out of anxiety, which in turn fuels anxiety. No one knows whether each plan and confirmation in the ad auction system will allow them to live longer or be closer to death.
01|"I haven't seen such a generous owner since Pinduoduo"
Kimi is the most out-of-the-circle AI product after the Wonderful Duck Camera, and unlike most short-lived "Internet celebrity products", Kimi has always maintained a high popularity after its release.
In March this year, Kimi, a subsidiary of the Dark Side of the Moon, surpassed Baidu Wenxin Yiyan in terms of monthly visits, becoming the top of the ChatBot track in terms of monthly active users and visits. According to the data of the AI product list aicpb.com August, Kimi's monthly visits reached 23.36 million, still ahead of Wenxin Yiyan's 19.16 million, while Minimax, which is also a large-scale model star startup company with the dark side of the moon, has only 513,900 visits to its conch AI.
Naturally, Kimi's ability to cast streams and user fission has also become the object of research by various AI companies.
It is almost universally recognized in the industry that the dark side of the moon is the most aggressive large-scale model startup in advertising. According to AppGrowning's data, Kimi's advertising was 15 million yuan in March this year, and nearly 50 million yuan in July and August, with a total investment of 140 million yuan in six months.
Screenshot of the official website of the dark side of the moon
Station B has become the "home field" of the dark side of the moon. A UP owner told "On the Mountain" that he can only receive two business orders a month, and the business orders for two consecutive months have given the dark side of the moon. As far as he knew, not only the science and technology area, but also the workplace, knowledge and other fields were also waiting for the gold coins scattered on the dark side of the moon, "I haven't seen such a generous gold owner since Pinduoduo." ”
In addition to content collaborations, Kimi also frequently delivers direct in-feed ads. Some media quoted a number of practitioners as saying that Kimi's CPA (Cost Per Action) offer at Station B was as high as about 30 yuan, and at the beginning of this year, the price was about 10 yuan.
The dark side of the moon is ferocious, and other companies are quickly following suit. On station B, Doubao purchased a competing entry, and searching for Kimi will instead show the delivery of Doubao; On Xiaohongshu, many people are teaching you to use Tencent ingots to plan your travel itinerary; Podcasts, which were previously seen as niche channels, are increasingly being heard of by AI tools.
The advertising war continues from online to offline. Beijing's subway stations already have advertisements for Zhipu Qingyan; In Lujiazui, Shanghai, where white-collar workers are infested, Minimax began to advertise in elevator halls; Like Feishu, Doubao has also begun to tell everyone how to "use AI to drive business growth" at the airport.
Companies are also taking the initiative to embrace these AI "financiers". At the second quarter of 2024 financial report meeting, Chen Rui, CEO of Bilibili, will take the initiative to mention that Bilibili has high-quality artificial intelligence content; Xiaohongshu has also increased the operation of artificial intelligence content this year, and some ordinary operational activities will receive direct platform support from the company's product and technology executives; Focus Media also began to ramp up sales to AI companies, eventually convincing some unicorn companies to place elevator ads.
Hundreds of schools are flowing, and the flow is also rising. According to Similarweb statistics, the total number of visits to the AI Five Dragons (Zhipu AI, MiniMax, Baichuan Intelligence, Zero One Everything, and Dark Side of the Moon) products skyrocketed by 963% in 6 months.
02|It's one thing to know the pattern, it's another thing to copy the next Kimi
Zhou is in charge of product operations at a two-dimensional related AI startup. In the past, her job was mainly involved in product planning and design, expanding partners and channels, etc., but this year, her job is almost the same: doing launch and driving growth.
"Almost every meeting, we have to start with a talk about Kimi's recent placements, and then talk about our own investment growth." Zhou Xian said, "Growth has become the most important thing for the company. ”
Zhou Xian said that the growth law is not difficult to summarize, to put it bluntly, it is a large number of delivery, in the simplest and rude way, quickly occupy the minds of users, and then achieve download and conversion.
It's one thing to know the pattern, though, and quite another to copy the next Kimi.
Kimi's streaming model is CPA (Cost Per Action), that is, when a user completes the corresponding action - often registers or downloads, then the dark side of the moon needs to pay a fee. This is a streaming strategy that can directly drive growth, simple and crude.
Streaming offers are similar to auctions, if others bid too high, then those with low bids will get less traffic. Large Internet companies and large-scale model unicorns with abundant funds will raise the cost of investment, and it is difficult for ordinary entrepreneur companies to follow up.
Zhao Tian said that there is no way for an entrepreneur like him to afford a CPA cost of 30 yuan. He calculated that now the early AI companies take money, often 3 million to 5 million, "We don't say anything else, the money is all taken out, advertising, up to 200,000 users, and this is not guaranteed to be retained." ”
As for Douyin, it has also become an "AI forbidden place". Zhou Xian said that except for ByteDance's own bean bags or AI products with investment partnerships, traffic will be restricted, either the content will be "violated", or the streaming consumption will be slow - which means that your content is not likely to be watched.
Doubao's advertisement on Douyin
The cost of streaming has increased, the novelty of AI has weakened, and the customer acquisition cost of AI companies has also increased. Some practitioners say that the total cost of Kimi's customer acquisition is now about 15 to 20 yuan. According to Sina Technology's previous estimates, Kimi's daily customer acquisition cost reached 200,000 yuan. This means that not only will the dark side of the moon have to pay more and more expensive growth expenses, but other AI companies will also have to get involved in this advertising auction and be forced to increase the cost of streaming, which has discouraged many entrepreneurs.
03|An unsustainable streaming war
The war has entered the white heat, and the unicorns and Internet giants who are the first to launch the charge have also begun to be anxious about the increasingly high cost of customer acquisition.
Kimi is the only product on the dark side of the moon, and founder Yang Zhilin has also said on several occasions that they do not have a next product at the moment, and they want to gradually turn Kimi into a "super app".
In the era of mobile Internet, making a "super app" is the ultimate dream of many entrepreneurs, and eventually WeChat, Douyin, Meituan and other applications that are all-encompassing in their respective fields and have the ability to expand to other fields were born, and the company has become a giant in the new era based on this.
To become a super app, there are two basic criteria, the first is that the number of users must be large enough and the use of it frequently enough, and the second is that these users will not easily migrate to other platforms, that is, build a high enough moat.
On the surface, it seems that the problem of the number of users to be solved by the stream. However, the improvement of user indicators such as DAU cannot fully enhance the competitiveness of the platform.
One of the product managers of the dark side of the moon, "Pine Goose", immediately posted on social platforms that high DAU will not necessarily crush other competitors, "AI provides the value of intelligence, and does not directly provide the value of information exchange." In most current product designs, there is no multilateral network effect, and the growth of DAU will not bring exponential value to the product. ”
However, while the growth in headcount may not bring scale to Kimi, it can help Kimi get data and feedback in a more natural environment to improve the product experience.
Yang Zhilin positioned Kimi as a "super assistant", he said: "I think in the era of AI, there is a high probability that a super application will be an assistant, and I think the need for intelligence is a very universal need, but today the ability is still in its infancy." ”
An important step in training to become a "super assistant" is to have enough data. Yang believes that data should not be a fixed data set, but a "variable", and that "how to use data, or get feedback from users, will become more and more important things here." ”
In this way, acquiring enough users is also a way to build a data moat. However, for a commercial company, ROI is measured in any strategy. According to AICPB, in the last three months, users have used Kimi for an average of four days, with an average usage time of no more than two minutes. It's hard to say how much data feedback can be brought about by such interactions.
Companies have also begun to reduce the cost of investment. 36Kr reported that Kimi's high marketing costs have made some shareholders feel unsustainable. An employee in the dark side of the moon said that in the fourth quarter of this year, the company will try to reduce the cost of streaming and increase the delivery of new channels such as video accounts, Weibo, and Zhihu.
On these platforms, Kimi's related materials are more "down-to-earth", with titles like "You can leave work early with Kimi", "Use Kimi to get Excel in ten seconds", "Use Kimi to become a self-media blogger with a monthly income of 100,000", etc., some of these materials come from material service providers, and many of them are produced by Kimi's market growth department, and some employees will also appear in person.
It's hard to say how many truly valuable users these creatives will attract. When users are drawn in by these exaggerated headlines, they actually use them only to find that the AI can't make them work once and for all. Many owners of small and medium-sized enterprises who pay attention to cutting-edge technology have experienced it, and finally found that manual labor is more convenient and cheaper than AI.
Some investors predict that the money burning war may end in half a year, when unicorn companies may not have much money to continue to invest, and large companies also need to assess ROI, so that everyone will gradually become more rational and pay more attention to commercialization and those users who are willing to pay.
04|Find other ways out
Zhao Tian said that he was now in a dilemma. If you don't continue to invest in the flow, it will be difficult to continue to grow, and it will be difficult to get the next round of financing; But if you continue to invest in the stream, you may only run out of money faster.
Zhu Xiaohu, founder of GSR Ventures, has always been skeptical about AI's C-end products, and is more resolutely opposed to "burning money to buy users", in an interview, he said: "It is very difficult to burn money to buy users today. ”
Starting in 2024, various investment institutions have tightened their investment in AI startups. An investor told "Mountain" that those products with more clear paid conversions, especially B-end products, may still get some money, and C-end products are almost impossible to get more than A round of financing at this stage. Zhu Xiaohu is even more outspoken, believing that the only way out for the current large-scale model unicorn is to sell it to large companies.
Entrepreneurs like Zhao Tian are not unaware of the cruelty of the situation, nor are they unwilling to look for entrepreneurial opportunities with more payment prospects. However, when they devote themselves to AI entrepreneurship, both entrepreneurs and capital are more patient, willing to wait for the market to mature, willing to make a small but beautiful product, and willing to tolerate temporary losses.
In just one year, when Zhao Tian went to talk about financing again, everyone told him that either he had enough users to verify the market viability, or he could immediately prove his commercial prospects. Proving the prospect of commercialization means almost having to start a new project. In front of Zhao Tian, it can only be a way to get customers.
More large companies are also realizing that it is difficult to retain users brought by investment when it is difficult to open the gap between technology and products. Doubao has the advantage of Douyin, which can "occupy the user's mind" with a lower cost, and now their focus has shifted to sales rather than marketing, and they have begun to customize technical solutions for B-end customers.
An Alibaba insider told The Mountain that they feel that AI should be more like a cloud service, so it should be targeted at "producers", rather than just an entertainment application.
Zhao Tian said that in the current environment, it is too difficult to make a "small but beautiful" AI application. He plans to focus on overseas marketing in the future, where marketing is more cost-effective, the cost of customer acquisition on Google and Twitter is only about half that of China, and people are more accepting of a small product, rather than having to have a high number of users to get financing.
Panic hangs over everyone about missing out on AI. A lot of people don't know why they do it, but if they don't, they may not be able to stay at the table. Perhaps both entrepreneurs and investors need to be more patient, after all, AI is still a child, and the technology, products, talents, and markets needed to grow take time, and it is difficult to force maturity.
"In fact, AI sells services, so as long as you serve some people well, you can survive. The cost of finding this part of the population at home is too high, and it is much better abroad. Zhao Tian said, "There is no one there who is urging you to reach a hundredfold scale in a short period of time." ”
(Zhao Tian and Zhou Xian are pseudonyms.) ）
Author|Qi Ming Editor|Xue Xingxing This article is written by Everyone is a Product Manager Author【山上Hillvue】, WeChat public account: [山上], original/authorized Published on Everyone is a product manager, without permission, it is forbidden to reprint.
Image from Unsplash, based on the CC0 license.