China's e-commerce market does not slow down for anyone. New AI tools drop every quarter, platforms keep changing their algorithms, and brands are under constant pressure to move fast. So when a French Tech event in Shanghai dedicated an entire evening to AI and e-commerce, the room filled up quickly — and the conversations did not disappoint.
But somewhere between the live demos and the panel discussions, one question kept surfacing that nobody had a clean answer to: are we actually measuring the right ROI when it comes to AI in China's e-commerce?
That question is what this article is about. Because there is a real difference between AI that looks good on a slide and AI that genuinely grows your business.
What the French Tech Shanghai Event Highlighted
The event brought together a mix of founders, marketers, and tech leads working across China's digital commerce ecosystem — French and international profiles, many of them scaling brands on Tmall, JD.com, and Douyin. The format was open, the level of conversation was high, and the energy in the room reflected just how fast things are moving here.
A few themes dominated the discussions:
- AI-powered personalization — dynamic product recommendations, personalized landing pages, and content tailored to individual browsing behavior across platforms
- Livestream commerce automation — AI tools optimizing scheduling, script generation, and real-time audience targeting on Douyin and Taobao Live
- Customer service bots — WeChat-integrated AI handling pre- and post-purchase queries at scale, with response times under one second
- Predictive inventory management — using AI to anticipate demand spikes around 618 and Double 11, reducing overstock and stockout risk
The tools being used were genuinely impressive. But the more I listened, the more I noticed something missing from most of the conversations: a serious discussion about whether any of this was actually translating into business results. Not technically — but commercially.
If you want to understand how AI is reshaping specific categories in China's digital commerce, the recent deep dive on E-Commerce & Livestream trends in China's cosmetics market published on this blog is worth reading alongside this piece.
The Problem: We Are Looking at the Wrong Numbers
Here is where things get uncomfortable. Most brands and agencies are tracking AI performance through metrics that feel good but do not tell the full story.
Click-through rates go up. Session duration increases. Chatbot response times drop. The slides look great. But none of that automatically means more revenue.
This is the classic trap of vanity metrics — numbers that are easy to report, hard to argue with, and disconnected from real business outcomes. Two examples came up at the event that stuck with me:
- A brand deployed an AI recommendation engine that lifted product page engagement by 40%. Impressive. Except their actual conversion rate barely moved, and average order value dropped because the algorithm kept pushing lower-priced items.
- Another company used AI to triple their livestream output. More content, more views, better reach numbers. But customer retention stayed flat because the content quality suffered and the brand message became inconsistent.
The lesson is simple: engagement is not revenue. Output is not impact.
There is also a short-term vs. long-term problem. In China's e-commerce market, the pressure to show results fast is real. Brands are running on quarterly targets, and AI investments get evaluated on the same cycle as a flash sale. That is the wrong framework. The real ROI of AI — especially in personalization and retention — often takes six to twelve months to show up in a way you can actually measure.
"The question is not whether AI is working. The question is whether you have built a measurement framework that can actually capture its value — including the value that does not show up on a dashboard next quarter."
What We Should Actually Be Measuring
So what does a smarter AI ROI framework look like in the context of digital marketing in China? Here are the KPIs worth prioritizing:
- Customer Lifetime Value (CLV) — If AI is genuinely improving personalization and experience, you should see it in repeat purchases and total spend over time. CLV is one of the clearest signals that your investment is compounding, not just creating a short-term spike.
- Retention and Repurchase Rate — Particularly relevant in China, where a competitor can steal your customer with one scroll on Douyin. Track your 30/60/90-day repurchase curves. AI that builds real loyalty shows up there.
- Conversion Quality, Not Just Volume — More traffic means nothing if the economics do not hold. Look at the full basket: conversion rate × average order value × margin. All three together, not just one.
- Cost to Serve — AI in customer service should reduce operational costs. Track cost per resolved ticket, reduction in live agent hours, and escalation rate. These are concrete and directly tied to profitability.
- Attribution Across Touchpoints — In China's fragmented ecosystem — WeChat, Douyin, Xiaohongshu, Tmall, JD.com — attribution is genuinely complex. Any AI tool claiming to drive sales needs to be evaluated within a proper multi-touch model, not just last-click.
For a broader look at how ROI compares across digital channels in 2026, this article on ROI par canal digital en 2026 gives useful context on where marketing spend generates the most value right now.
My Take: The Real Strategic Gap
What I observed at the French Tech event — and what I see more broadly in the China market — is a gap between operational AI and strategic AI.
Operational AI is about efficiency: automating repetitive tasks, reducing friction, speeding things up. It is useful and relatively easy to measure. Compare before and after, run the numbers, done.
Strategic AI is something else. It is about making better decisions — which customers to invest in, which products to prioritize, where to allocate budget across a fragmented media landscape. The ROI is real, but it takes longer to see and requires more sophisticated measurement infrastructure to capture.
Most brands in China right now are doing operational AI. Very few are doing strategic AI well. And almost none have built the data foundation — clean first-party data, an integrated customer data platform, consistent tagging across platforms — that strategic AI actually needs to function.
This is both a significant opportunity and a warning. Brands that invest in the right infrastructure now will be able to extract real, compound value from AI over the next two to three years. Brands that just layer AI tools onto messy data pipelines will keep producing impressive-looking metrics that do not move the business.
There is also a China-specific dimension to this conversation that cannot be ignored. The platforms here — Alibaba's ecosystem, Douyin's algorithm, JD's supply chain intelligence — all have their own native AI capabilities baked in. Foreign brands especially need to understand that maximizing AI ROI in China is not just about the external tools you bring in. It is about how well you integrate with the platforms' own intelligence, and how well you understand the data they do and do not share with you.
According to McKinsey's State of AI report, companies that tie AI investments to specific business KPIs from the start are significantly more likely to report measurable ROI. That finding is particularly relevant in a market like China, where the temptation to chase platform-native vanity metrics is constant.
A study from China Internet Watch also consistently shows that Chinese consumers engage with personalized content at high rates — but engagement alone does not correlate with purchase intent unless the underlying commercial experience is tight. That gap between attention and conversion is exactly where the ROI measurement problem lives.
The Bottom Line
The French Tech event in Shanghai was a good reminder that the AI conversation in China's e-commerce is not slowing down — and neither is the competition. But enthusiasm is not a strategy, and impressive demos do not pay for themselves.
If you are investing in AI for your e-commerce operations in China, the most important question to ask right now is not "which tool should we use?" It is: "do we have the measurement framework to know if it is actually working?"
Because the brands that win here will not be the ones with the most AI. They will be the ones who know what value looks like — and have built the systems to see it clearly.
What KPIs is your team currently using to evaluate AI ROI in China? I would genuinely like to hear how others are approaching this.
Key Takeaways
- AI adoption in China's e-commerce is mainstream — personalization, livestream tools, and chatbots are now table stakes
- Most brands are measuring AI ROI through vanity metrics (engagement, output volume) rather than real business impact
- Engagement does not equal revenue — CLV, retention, and conversion quality are far more meaningful KPIs
- The gap between operational AI (efficiency) and strategic AI (decision-making) is where most brands are losing value
- China's platform fragmentation makes proper attribution essential — last-click models are not enough
- Building a clean first-party data foundation is the prerequisite for any meaningful AI ROI in the long run
