Alimama's Wanxiang Jingdian AI tool helped Lin's Home to precisely target gaming enthusiasts with their fabric beds by identifying user search intent and behavioral signals and understanding product characteristics to enable cross-scenario matching.

Author: Cynthia

Source: GeekPark

In the summer of 2026, Lin's Home wants to sell a fabric bed.

Then, Alibaba's Wanxiang Pointing identified a group of esports players and recommended targeting users who had searched for terms like “esports room setup,” “gaming room renovation,” “RGB LED strips,” and “anime-themed rooms.”

For Lin’s Home, a leading brand born on the internet and long reliant on data-driven operations, this is a baffling decision.

In the past, Lin's Home Brand has accumulated over 55 million followers across all platforms, tens of thousands of SKUs for sale, and more than a thousand offline stores, earning recognition as an industry-leading creator of best-selling products.

Under traditional matching logic, the most natural consumers for a bed would come from those undergoing home renovation, newly married couples, people moving to a new home, or those who have recently purchased home appliances or furnishings. As a result, the advertising system relies on predefined tags to match audiences, and merchants instinctively select related keywords such as renovation, furniture, and flooring—this has become second nature to the team. Consequently, a young person searching for电竞房布置 (e-sports room setup) hardly seems like the ideal consumer a bed should prioritize targeting.

But common sense alone cannot drive growth in a saturated market.

But Wanxiang Dianjing doesn’t see just the label “e-sports”—it sees the entire set of consumption intentions behind it: it recognizes that these young people have a strong willingness to spend in order to enhance their gaming experience—they buy ergonomic chairs, 4K monitors, RGB ambient lighting, and transform their rooms into more comfortable, immersive spaces that reflect their personal taste.

Following this consumption chain, for PC gamers, a comfortable bed to relax on is part of the entire gaming setup experience; for mobile gamers, the bed is not just a place to rest—playing while leaning on a soft fabric bed is itself a more comfortable gaming experience. In other words, they are consuming a lifestyle.

Once this layer was understood, the team immediately decided: invest!

The final results also exceeded many people's expectations. New users accounted for over 80% of clicks and a remarkable 83% of transaction value. Beyond precisely targeting the new user segment, what surprised the team most was that no home furnishing brand had previously identified this blue ocean—resulting in the team’s add-to-cart costs being less than one-third of those for traditional home renovation audiences.

Thus, a soft fabric bed became associated with esports. Following this, this approach was even extended to nightstands, helping Lin’s Home discover a new target audience interested in easily accessible sleep wellness products.

A brand-new system has shifted advertising from a zero-sum game to an AI-driven era of growth, making it possible to discover new audiences and create new growth opportunities.

01Why is it becoming increasingly difficult for traditional advertising targeting to find new growth?

Digital advertising has not only begun to rely on algorithms in the era of large models.

For nearly 20 years, "collaborative filtering" (similarity matching) based on traditional machine learning has been widely adopted in advertising systems. In this system, the merchant's "purchase" and the platform's "matching" are two separate processes. The advertising system assigns labels to users based on their historical behaviors, such as what they have viewed or purchased. Merchants then buy audience segments or keywords, and the platform maps consumer demand to these labels for matching.

This approach is highly effective in incremental markets. When a brand reaches 100,000 consumers, there are often hundreds of thousands or even millions of similar individuals who have yet to be reached. As long as the similar audience continues to expand, growth can be sustained.

But the longer an industry has been operating, the more aligned its participants become in understanding high-value consumers. Furniture brands target people renovating their homes, beauty brands compete for high-spending women, baby and parenting brands focus on the family life cycle, and sportswear brands all understand what running, cycling, hiking, and camping signify.

As a result, it brings only increasingly expensive traffic costs and declining conversion efficiency after players saturate their investments.

In addition, it inherently has a limitation. In a system requiring real-time computation of massive data, the combinations of users, products, and ads are nearly infinite, making it impossible for the platform to calculate every relationship at the highest cost. If an esports player never enters the candidate pool for “bed,” no matter how accurate the subsequent CTR model is, it will never have the chance to score him. For this reason, historical data naturally rewards relationships that have already been proven.

But consumers themselves are not a stable label.

Someone who was renovating two years ago may no longer have furniture needs today; a consumer who typically engages in sports might suddenly enter a completely different purchasing state due to a trip, a concert, or a move. Recommendations based on such historical tags often result in a significant amount of outdated traffic.

Therefore, more important than who this person is is what problem they are currently solving. This is precisely the new capability that AI-driven "synchronous matching" brings to advertising systems: by recognizing users' search intent and behavioral signals, along with a deep understanding of product characteristics, AI can help merchants reach people beyond their original targeting labels and touchpoints they haven’t actively pursued.

Thus, a new incremental blue ocean market is gradually coming into view for everyone.

02 Person: From who he is to what he needs now

In the past, buying a suitable product was a technically uncertain endeavor for both buyers and sellers.

For merchants, they know that after someone buys a gaming graphics card, they are more likely to purchase a monitor; after someone searches for a tent, they are more likely to buy outdoor gear. But extending this to broader scenarios: Does someone who has set up a gaming room at home need a bed? What kind of stool does a refined young woman need before applying makeup and heading out each day? These are questions far beyond the scope.

Merchants naturally tend to categorize products and SKUs based on existing scenarios. But consumers don’t live by SKUs. They first move, travel, get married, work out, attend concerts, set up gaming rooms, or apply makeup—and only then generate a series of specific product needs. Before making a purchase, they must first translate their life needs into product language. For example, buying a phone for an elderly person requires breaking down “loud sound, large text, simple operation” into a set of product specifications; first-time campers must understand what a tent, tarp, and sleeping bag are before they can navigate the platform’s category system.

With Wanxiang Pointing, merchants can now not only identify who their users are, but also clearly understand what they are doing, in what context, and what they need. For example, Yisong, a leader in the ergonomic chair category, reached this audience of beauty-conscious women through this approach.

At the outset, Wanxiang Dianjing, grounded in its underlying world knowledge system, first helped Yisong identify the user's intent on the demand side, recognizing that ordering a nail table and makeup vanity actually signaled a need for a potential long-term grooming space.

Next, the large language model infers the user's consumption motivations, preferences, and purchase stage, encoding the user behavior sequence into a structured representation of consumption intent—such as identifying this as a female user who values beauty and seeks a high quality of life.

Then, through the semantic alignment layer, transition from label matching to semantic reasoning, mapping and aligning the feature vectors with the intent vectors in a high-dimensional semantic space, while using causal reasoning to surpass traditional correlation logic and understand the user’s deeper motivations for purchasing—such as the fact that ordinary makeup stools are hard and lack back support, so they need a comfortable chair that is also visually appealing for photos, and must be rotatable, adjustable in height, and滑动 to enable flexible switching between makeup, work, and study.

Finally, on the supply side, Wanxiang Pointillism leverages product understanding to uncover functionality, uses large language models to deconstruct product selling points into multidimensional structured expressions, and builds a fine-grained product attribute graph covering categories, functions, scenarios, and user groups—enabling deeper semantic matching between supply and demand. For example, a woman who spends time daily on her appearance may, after purchasing a vanity, also need a chair. An ergonomic chair, beyond simply being labeled as “office,” simultaneously meets multiple lifestyle needs such as long-sitting support, flexible mobility, and versatile space compatibility.

With this deeper insight into user needs, the advertising system has shifted from audience expansion to intent expansion, and its asynchronous targeting matching has been transformed into scenario-based real-time targeting matching.

But understanding people alone is not enough.

If an ergonomic chair is forever defined only as office furniture, it will still struggle to truly connect the chair with seemingly unrelated scenarios like dressing tables or nail tables. If a bed is forever understood only as bedroom furniture, no matter how well the system understands esports players, it will still struggle to appear correctly in an esports enthusiast’s gaming room.

AI targeting also requires reinterpreting the other end:

Cargo.

03 Asset: An item that can belong to multiple markets

Past platforms managed supply through categories, attributes, and SKUs, which is a prerequisite for the efficient operation of commercial systems.

A bed is first classified as a "furniture" item in the system, with attributes such as size, material, color, and style, and then it enters the market based on its type.

But the way consumers use products never fully conforms to the original product definitions set by companies: Coca-Cola was originally sold as medicine, microwave ovens originated from radar technology, and sneakers evolved from specialized equipment to everyday apparel across different eras.

The history of business is never short of stories where users redefine products. However, in the past, identifying problems often relied on lengthy market observation, research, or chance events.

Now, with Wanxiang Pointing, AI can deeply understand the subtle connections between this product capability and user needs.

Lin's Home selling fabric beds to esports enthusiasts, and Yisong selling ergonomic chairs to beauty enthusiasts, are just one example each.

More representative is how Wanxiang Dianjing helped Helena sell products to skiing and outdoor enthusiasts.

In May this year, Wanxiang Pointing identified a group of potential customers for Helena who surf and dive in summer and ski in winter. When the team saw the results, their first reaction was: “Could the AI have gone off track?” After all, these individuals’ primary spending patterns were almost entirely focused on sports equipment.

But AI keenly captured their need for sun protection and certainty. Among this group, 27.35% prioritize quality, and 40.25% place particular emphasis on materials. When purchasing snowwear, diving, and outdoor gear, they meticulously compare materials, performance, and reliable results—and the Black Bandage itself stems from a product lineage that emphasizes "certainty." In 2008, Helena Rubinstein partnered with Swiss Laclinic to develop the Re-Plasty series, inspired by post-surgical care; Age Recovery Night later evolved into the "Black Bandage," which Laclinic still positions today within the context of post-surgical recovery, active ingredients, and clinical care.

Therefore, it is likely that someone who consistently prioritizes performance, materials, and reliability in athletic gear will also be more willing to pay for clinically validated results, proven efficacy, and product texture when making skincare decisions in the thousands-of-yuan range.

On the final operational results, after the official launch, this batch of outdoor high-value audiences had a 2 percentage point higher new user rate compared to other groups, along with higher engagement and conversion rates.

Similar "translations" also occur with other more everyday products. For example, using Wanxiang Pointing,蕉下 sold sun-protective masks to fitness and cycling audiences seeking "breathable, non-slip, and photogenic" features, resulting in an 11% increase in new customer transaction share and a 39% increase in new visitors within two months.

And when people shift from fixed tags to dynamic intentions, and products shift from fixed categories to solutions to problems, the last thing to change is how people and products find each other.

04 Scene: AI-driven, redefining the way supply and demand are connected

When AI targeting can simultaneously understand "what a person is doing" and "what a product can solve," the previously distinct boundaries between search, advertising, and recommendation begin to blur. Because at their core, they are all addressing the same thing: at a specific moment, determining what demand a user is forming and which products should be involved in fulfilling that demand.

This change also readjusts the division of responsibilities between merchants and the system.

Previously, when creating a new advertising campaign, merchants typically had to first enter keywords, target audience, region, and time periods—essentially telling the system: “Who are my customers?” In the new system, Wanxiang Pointing allows AI to first analyze product highlights, potential demands, and possible consumer scenarios, after which merchants decide whether to adopt or modify the suggestions.

Meanwhile, subsequent creative generation and ad execution will follow the same logic to complete supporting actions. Leveraging existing user insights, Wanxiang Pointing will not display the original bed advertisement targeted at bedroom consumers directly to esports players. Instead, it extends the funnel further into creative and bidding: first understanding business goals and consumer intent, then identifying potential audiences, generating differentiated creatives with AI, and finally using intelligent bidding to determine which traffic opportunities to compete for.

Ultimately, transform the diverse needs of countless user segments into tailored advertising strategies and communication techniques, driving higher new customer conversion and incremental value. According to official disclosures, after its launch during the 618 campaign and full-scale rollout by the end of July, Wanxiang Dianjing has been used by over 150,000 merchants across keyword and audience targeting scenarios; within specific periods, new customer click-through rates exceeded 91%, and new customer transaction rates surpassed 76%.

Most importantly, although Wanxiang Dianjing is powered by an entirely new set of capabilities, merchants do not need to learn any new tools—it seamlessly integrates into existing promotional scenarios, from audience targeting and keyword advertising to brand-wide coverage and content marketing.

As Liu Bo (Jialuo), Senior Vice President of Alibaba Group and current President of the Tmall Division, President of Alimama, and Head of Taobao Live and Content Division, put it: “(The platform strives to) extend its strengths further, delve deeper into challenging areas, reduce merchants’ costs, and bring greater certainty to their operations.” At the same time, “There is still significant potential for the algorithm team to unlock—enabling better products and better merchants to connect with higher-quality consumers; conversely, high-quality consumers can also discover products that are more personally matched for them on the platform.”

Of course, in the long term, once this platform’s capabilities are fully opened, the information gap it creates will not persist indefinitely. Today, only one company sees the rest needs of esports players, but soon, all furniture companies may see the same group of people. The question then shifts from “whether it can be discovered” to “what to do after discovery.”

Similarly, a new scenario emerges: some companies may doubt, some may add a new advertising plan, some may redesign their creatives and product offerings, while others continue researching consumers' living spaces and real needs, ultimately adjusting product dimensions, features, or even new product directions.

During this process, crowdsourcing, real-time bidding, and certain aspects of creativity and consumer insights can also be platformized.

However, product selection, inventory risk, supply chain adjustments, and resource allocation still must be handled by the companies themselves. AI will enable more companies to see new opportunities, but it won’t give them the same speed of response.