Shulou(Shulou.com)06/02 Report--

"Top flow" is generally a real-life product with absolute popularity created by entertainment companies and platforms. In the past, the "top class" were all popular entertainment students who were better than "singing and dancing RAP". However, this year's "top" jianghu, but killed an entertainment outsider-e-commerce anchor Li Jiaqi.

The 1992 boy used the words "OMG" and "Buy it" to make countless fans "lose their money". The record spread around the world is that he sold out 1500 lipsticks in PK,5 minutes with Jack Ma and brought more than 100 million yuan with goods on Singles Day.

Li Jiaqi is not alone. Live streaming with goods as a new phenomenal consumption channel, in all the short video content ecology, there are such "first brother and first sister". Weiya of some treasure and Simba of Kuaishou have also created eye-catching e-commerce sales records.

But the scenery favors only a few people. Every day, more than 70% of the carrier broadcasters are unattended, talking to themselves in the sparsely populated studio to the mobile phone camera day after day. Even Mei ONE, which successfully created "Li Jiaqi", failed to replicate the myth and finally chose to give up the positioning of the MCN organization and concentrate on creating the IP of "Li Jiaqi".

Xiao Hong depends on praise, while Da Hong depends on her fate. In the world with cargo broadcasts, red is a kind of metaphysics. It may be a "black box" called the recommendation algorithm that controls their fate.

The flow black hole behind the scenery

This year is a year in which e-commerce live streaming is completely out of the circle. Many people have seen the huge opportunities for the industry Nuggets through the popularity of front-line anchors such as Li Jiaqi and Weiya. Live broadcast with fire this sales model, so that for a long time in digital advertising "ghosts hit the wall" brands see a real transformation. Along with online traffic is the adrenaline of retailers.

It is said that on the day of Singles' Day, Taobao alone had more than 100,000 live broadcast rooms. Even in normal times, there are still a lot of people explaining products on the platform at three o'clock in the morning.

However, a platform has only one "first". In e-commerce live streaming, the curve at the top of the pyramid is unusually steep.

While "Li Jiaqi" is out of the circle, the living environment of more middle-waist anchors is still not optimistic.

Survey data show that nearly 40% of the "economy with goods" practitioners work an average of 8-12 hours a day, and 17.8% of anchors have a longest continuous live broadcast of more than 10 hours.

Even Li Jiaqi, Weiya and others, even if there is a head anchor's platform guaranteed traffic. It is also necessary to maintain a high-intensity broadcast. Li Jiaqi broadcasts for more than three hours a day on Singles Day, and Weiya broadcasts for six hours in a row. After the broadcast, she has to breathe oxygen to review the day.

In contrast to the high-intensity work is the income of the underlying anchors. In general, their income structure is composed of basic salary and commission. If the anchor's ability to carry goods is not good, the salary will hardly look good. Carrying goods needs to be transformed, and the huge flow has become the premise of everything. But how do you get traffic? Has become one of the most uncertain things in the industry.

The only parameter that can bring about a change is the broadcast time. New e-commerce anchors will be told that they must broadcast live broadcasts continuously and cannot stop broadcasting easily. It is common to work more than 10 hours a day.

Under the intensity of "007" work, many people are considering changing careers after less than half a year in the profession. It is often the rural youth who are not highly educated and good-looking, it will be easier to stick to it.

Another key parameter summed up by operators in the industry is transformation. If the traffic is sent to the broadcast room, but can not be retained within an hour, the viewing time or purchase proportion is not enough, the recommendation system will not hesitate to divert.

A joke is popular in the industry-the scariest day of studio operation is the day when your data are bad, and the second is the day when you have good data but don't know why.

In a sense, e-commerce anchors are not facing the spotlight, but a cruel Colosseum. The machine algorithm has become the supreme being in charge of life and death, and even the "first brother and the first sister" have to wait for it, making all kinds of treasures and colorful clothes on the platform day and night.

Copy the N technical questions of Li Jiaqi

Recommendation system originates from the problem of information overload in the network era. When users are faced with a large amount of information, it is difficult to get the information that is really useful to them, so through user modeling, personalized calculation is carried out according to their individual needs and interests, and then information services with thousands of people are provided. Created a lot of successful commercial products.

In most cases, the recommendation system will be used to upgrade the user experience. But in the field of content, the recommendation system is the creator's life, and the "algorithmic survival" of e-commerce anchors is more complicated.

First of all, the content of the short video is understood. As a direct "product" of e-commerce output, "content" is not as easy to identify as picture and text information, so it can be matched according to relevance or user preferences. Especially for some long-tailed anchors in the vertical field, the various attributes and characteristics of the studio are often very generalized, and all of them are based on real-time operation, and the relevance between the content and potential consumers can only be recommended. The precipitation effect of customers getting traffic is also very random.

In addition, if you want to control the various variable parameters of the algorithm, it is difficult to do any coquettish operation on the live recommendation system. The information platform can also be used as a title party to cheat clicks, but e-commerce live broadcast platforms often have accumulated a large number of user characteristic data and models, and recommendation systems are often designed according to the hybrid mechanism. the demography, classical collaborative filtering, product relevance, matrix decomposition and other different recommendation algorithms are combined according to a certain weight, and the specific weight will be tested repeatedly. Unified packaging to the deep neural network this huge "black box" inside.

During the cold start-up period between LVB, is it easier to recommend users and VJs with high similarity, or is it more likely to trigger platform traffic when watching Gao more effectively in history? It is not surprising that the overall ranking optimization of complex multi-models has become "metaphysics".

Of course, platform operators also have certain artificial recommendation rights, such as bringing out traffic highlands such as home page focus and hot bits to cooperate with some live studios, or recommending some high-quality potential VJs with long tails to balance the overall ecology.

But this inevitably touches the "class solidification" of the whole e-commerce live broadcast. The core significance of live broadcast + e-commerce lies in breaking through the bottleneck of online customer acquisition, opening up the situation through real-time sales, and the limited time window is more suitable for impulse consumption. Li Jiaqi sold men's skin care products on the street, while Li Xiang failed miserably in selling mink coats, indicating that rigid demand and low prices are the killer mace of head anchors, and only head flow can convince manufacturers that they are willing to trade profits for sales. Mid-waist and long-tail anchors who do not have the bargaining power are more difficult to hand over beautiful prices and conversion homework, so they can only accompany them for a long time.

This is why, whether algorithmic or manual, even if they know a lot of truth, they are still unable to copy the second "Li Jiaqi".

It's not AI, maybe it's people.

An interesting phenomenon is that when countless anchors try to make the algorithm fall in love with them and reach the pinnacle of their lives with the help of intelligent machines, Li Jiaqi and his team want to turn him into a machine.

Not long ago, USTC used AI to synthesize a virtual anchor of Li Jiaqi. After training, the "AI Li Jiaqi" began to sell eye drops and instant noodles.

AI said, "all girls buy it," will there be a magical effect of carrying goods? The answer is NO.

"unlike a real person" and "don't want to buy without passion" is the most common feedback. In a sense, outside the "routine" of team professional operation, something that cannot be copied can explain "how Li Jiaqi became Li Jiaqi."

First of all, there is a distinct sense of smell for digital culture. Unlike physical sales, which can be guided in multiple dimensions, user feedback in the studio is a code made up of "0" and "1". We should not only have humanized and flexible chat and communication, but also have the digital literacy to adjust the state according to the data of the studio, and grasp the operation law of the machine, which are the "dual wisdom" that individual anchors need to have.

In an interview, Li Jiaqi revealed that he would be particularly excited to see the data change and think about how he should behave. On one occasion, he asked his assistant to help him sell the goods, and he was still there to explain the goods after he had finished shooting. It can be said that the sensitivity to data is testing e-commerce anchors like never before.

Secondly, e-commerce anchors need to output empathy and trust. Different from show anchors, e-commerce anchors are burdened with commodity transformation indicators and endorse "decision reference" with personal brands as consumers, which requires a strong sense of IP trust. For example, Li Jiaqi often tries dozens of lipsticks within 30 minutes from her early beauty makeup to today's universal products, and selectively recommends them. The people of "good Life selection" and "Devil" stand up like this. Including Weiya, Lieer Baby, Li Xiang and others, they will mention family and family members in the live broadcast, share the details of life with fans, and deepen emotional connection and trust. This kind of fan influence cannot be infused quickly through systems and algorithms.

From this point of view, to rely on human charm to achieve a connected live broadcast, even if it cannot rely on the algorithmic incarnation of "Li Jiaqi 2.0", at least it is an iron rice bowl that will never be taken away by AI.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.