Zhou Liqing's query "Taobao search is not working well, what should I do" shows over 410,000 related notes on Xiaohongshu, while "Taobao search shows identical clothes" has over 540,000 notes. These posts all point to the same fact: the user experience of Taobao search is declining. When users try to find desired products on Taobao, they often encounter monotonous search results; even after scrolling through dozens of pages, they only see three or four different styles. Many product titles, in pursuit of exposure, pile up lengthy descriptions and functions that actually differ greatly from the product itself, not only greatly interfering with users' effective searches but also exacerbating user fatigue and distrust. In the past, through Taobao search, users could quickly find desired products with simple keyword searches and filter by sales volume, positive reviews, and other ranking mechanisms. However, imperceptibly, those good product designs and user experiences have quietly dissipated. To improve search effectiveness, Taobao has introduced advertising services similar to search engines' "top positions" and "priority display." These services allow merchants to gain higher display and sales opportunities in keyword searches, usually appearing in more prominent positions in front of consumers. While this ad-oriented search mechanism provides merchants with more exposure opportunities to a certain extent, it also makes search results increasingly utilitarian and homogenized. From a higher perspective, the search issue may be a legacy of Alibaba's internal strategic vacillation. For a long time, Alibaba's internal understanding of who the "customer" is in the corporate core value "Customer First" seems to have been wavering. In the second half of 2020, then-CEO Zhang Yong responded to employees' questions about who the customer is: "The discussion of 'buyer first' or 'seller first' is meaningless. Neither is appropriate; we need to look at specific situations." This year, Joe Tsai said in an interview video: "Our customers are the people who use our app to shop." This statement provides an answer to the question of who Alibaba's customers are. From this point, it should become an important guide for Taobao to optimize its search experience, meaning Taobao should focus on consumers and improve its search experience accordingly. Currently, the actual situation is not satisfactory. This may reflect deviations in the execution layer during the top-down transmission of decisions within Taobao. Falling into a search "ghost wall" "Taobao can't find treasures, but there is definitely a ghost wall." Wang Miao has been busy on Taobao for nearly a month preparing for winter. Initially, he planned to buy a thick coat to withstand the cold, so every day after work he would open Taobao and type the keyword "thick coat" in the search bar. After a period of searching, Wang Miao found that the coats he saw every day were the same few: "Fanluo FanoStudios, Huangmaimai Hmaimai, BUD Bade. I almost see the products from 'familiar' stores every day, and scrolling down doesn't reveal anything new." It felt like being trapped in a loop; no matter how he changed keywords, he couldn't avoid those familiar "faces." Wang Miao also noticed that when searching for "thick coat," a large portion of the displayed products came from stores he frequently visited or had followed. For Wang Miao, who wanted to find more new options, this was somewhat helpless. "I want to see products from more different styles and stores, but it always recommends those I'm already familiar with, making the choices feel narrower," Wang Miao said with some resignation. On one hand, encountering the same products during searches; on the other hand, the styles and types of products found are severely homogenized. Sun Tong shared his experience searching for "cotton-padded jacket." Once, by chance, he was attracted by a hooded "Zuoshandiao" cotton-padded jacket with a fur collar on Taobao's homepage and clicked in to view details. But when he exited that product page and continued refreshing, he found that the screen kept showing "Zuoshandiao" jackets from different stores, almost identical in style, with only slight price differences. "I originally wanted to see more styles of cotton-padded jackets, but all I saw were 'Zuoshandiao,' feeling like I had walked into a dead end and couldn't find a new direction." Even more absurd than the "ghost wall" is the mismatched search function. When Wang Miao typed "short down jacket" in the search box, the results were mixed with many products completely unrelated to "short down jacket," such as lambswool coats and leather-fur coats, and these product titles did not contain the search keywords either. Multiple search functions fail simultaneously Currently, searching for stores on Taobao also encounters difficulties. Sun Tong mentioned that once he saw a store shared by a fashion blogger on Xiaohongshu, and the style matched his preferences, so he went to Taobao to search. However, when he entered the store name from memory, he couldn't find it. "I thought the store had closed, but it seemed unlikely; how could such a popular store suddenly disappear?" So Sun Tong had to privately message the blogger, who copied the store name to him, and only then did he successfully find the store. "Especially stores with English names are even harder to search. I found that even subtle differences like capitalization and spaces can determine whether you can find the store. This is very inconvenient for consumers." Moreover, the "order search" function after purchases on Taobao has also received many complaints. Shang Wen complained about his experience: "The title clearly contains the keyword I searched, but I can't find the order." Shang Wen bought a pair of shorts this summer, and his friend wanted to buy the same pair, so Shang Wen confidently went to his order history to search. "The shorts are denim, so I searched 'jeans' in my order history, but I couldn't find the order at all, even though I definitely bought them. I then searched 'pants,' not daring to add the character 'zi' (子), and only then did I find it." Shang Wen said this situation is not accidental; now when searching order history, he only dares to use the most basic and simple keywords, such as "skirt" or "pants," for fear that a slight mistake might cause him to miss the order he wants to find. Compared to Shang Wen, Zhang Lan's order search situation is the opposite. The year before last, Zhang Lan bought two bags from the same store on Taobao. Last week, on a whim, he wanted to check if the store had new products, so he searched "bag" in his order history. However, after repeatedly searching, he couldn't find the orders for those two bags. It wasn't until he searched "tote bag" based on the bag type that he finally found the relevant orders, and then found the other order through the store name. The problems Zhang Lan and Shang Wen encountered were different in form, but they both faced a common dilemma: the keyword was clearly in the product title, yet they couldn't find the corresponding order. Many netizens also reported that the order history list is not arranged in normal chronological order but is irregular. When browsing orders, old orders from long ago often suddenly appear among recent ones. Some consumers consulted customer service about order search issues, but customer service often mechanically repeated "Sorry for the poor service experience" without providing practical solutions. Where is the problem? Taobao search is not an isolated case; the search functions of major e-commerce platforms seem to be in trouble. A netizen bluntly said: "1688's search is even more outrageous; if the search term is slightly detailed, you can't find anything." On Xiaohongshu, there are also over 40,000 notes about "Pinduoduo search keywords unrelated to results," and there are even ridiculous situations like "searching for hats brings up skirts and food." Taking Taobao as an analysis, from a technical perspective, for homepage search, an employee responsible for search and recommendation segmentation revealed that the core intelligent engine Ha3 and the core search ranking BERT are both operating normally. Taking the homogenization issue as an example, in technical implementation, the system inevitably performs deduplication, which is a basic operation to ensure diversity in search results. But in reality, consumers still frequently encounter duplicate search results, "which is highly likely due to deviations in search strategy." The search and recommendation team provides technical capabilities, but the final presentation of search results is determined by algorithms and operational strategies. Similarly, in order search, situations like searching "pants" but only showing some orders are highly unlikely to be caused by technical faults in the underlying inverted index; theoretically, the search is fully capable of presenting all related "pants" orders. The above-mentioned person said the problem is likely in the operational strategy, such as during the word segmentation stage, where the full name like "jeans" might be used for segmentation, thereby narrowing the search scope. Looking back at the development of Taobao search, the early Taobao search engine won user favor with its precision. A veteran user who has used Taobao for 17 years recalled: "The earliest Taobao search engine could accurately find all products under the corresponding keyword. For example, when you type 'blue square collar puff sleeve shirt,' it could really find all shirts matching that description on the Taobao platform, and sort them precisely by sales volume, positive reviews, and other dimensions." Times have changed; today's Taobao search is no longer purely function-oriented. According to Taobao merchants, Taobao's data analysis concluded that at least a dozen touches with buyers are needed to achieve the best conversion effect. Under this logic, the proportion of advertising in the entire recommendation system is increasing. E-commerce practitioner Jiang Xia told the author: "As users' time is dispersed across many platforms, e-commerce platforms increasingly focus on efficient traffic utilization. Therefore, platforms tend to recommend products that are more likely to convert, and if the product also has a high average order value, it's naturally a better choice." Under this premise, merchant behavior is also influenced by platform strategies. "When placing ads, merchants also focus on increasing bids for customers who recently added to cart or favorited, to improve conversion probability," Jiang Xia further pointed out: "To ensure traffic and conversion, merchants gradually tend to homogenize in product development, preferring to copy bestsellers or make minor innovations based on bestsellers, as this has a relatively high success rate. Those originally differentiated products, especially those that cannot effectively improve conversion rates, are difficult to obtain sufficient traffic, making it hard for merchants to recover development costs, so fewer and fewer merchants are willing to try." Merchants are all making "bestseller-like" products, which undoubtedly exacerbates the homogenization of products on the platform and further worsens the user search experience. As a result, users may be trapped in the strategic framework set by the platform and merchants, with increasingly limited product variety and a discounted shopping experience.
【New Order · Symbiosis】 The 10th China FMCG Innovation Conference Time: March 17-19, 2025 Location: Chengdu, China