At the campaign’s launch in Shanghai on Thursday, Alibaba vice-president Zhang Kaifu, who also serves as head of applications for the company’s e-commerce businesses, said the firm’s use of Qwen’s large language models (LLMs) had led to a “significant uplift across a number of key metrics”. Qwen models are developed by Alibaba Cloud, the AI and cloud computing unit of Hangzhou-based Alibaba, owner of the South China Morning Post.
Some metrics cited at the launch include a 20 per cent improvement in relevant results for complex queries, a 12 per cent increase in the return on merchants’ advertising costs and a 10 per cent increase in click-through rate for certain recommendation scenarios. Click-through rate is a key performance indicator that measures the percentage of people who click on a specific link or advertisement online, compared with the total number of times it was shown.
Achieving such double-digit results was very rare for Taobao and Tmall – Alibaba’s main Chinese shopping platforms – because these massive sites record sales worth trillions of yuan, according to Zhang.
“That usually involves the introduction of new technologies or a fundamental paradigm shift,” Zhang told the Post in an interview on the sidelines of the Singles’ Day campaign’s launch. He attributed those results largely to the deployment of Qwen’s LLMs.
In this year’s promotions for Singles’ Day – the world’s largest shopping festival – Alibaba is carrying out the first large-scale deployment of generative AI on Taobao and Tmall, integrating LLMs into the two platforms’ core search and recommendation engines. Using AI to drive more consumers to Alibaba’s promotions is expected to be key to boosting sales, as this year’s annual retail extravaganza has been extended to a four-week event.
