A Danish kitchenware brand called me last spring with a simple complaint. They were spending 90,000 RMB a month on Baidu PPC, getting around 12,000 clicks, and closing 41 orders. Their agency kept asking for more budget. Nobody had ever asked what happened to the other 11,959 people.
That is the whole problem in one sentence. Foreign brands in China buy traffic. Very few of them ever go back for it.
I am Olivier Verot, founder of GMA. I have lived in Shanghai since 2012 and I have audited a few hundred Baidu and Tencent ad accounts belonging to foreign brands. Retargeting is the line item where I see the most money left on the table, every single time.
Why Re-Targeting Matters More in China Than Anywhere Else
China has an enormous internet market. According to the 57th CNNIC report published in February 2026, the country counted 1.125 billion internet users at the end of 2025, an 80.1% penetration rate, with 99.6% of them online via mobile. That is the pool. The cost of reaching it once is high and rising.
Baidu still handles the majority of desktop and mobile search intent for commercial queries, and roughly 40% of visitors that a well-run campaign converts are reached through paid advertising in China on a second or third touch, not the first.
What Is Re-Targeting and How Does It Work?
Retargeting, also called remarketing, is a form of online advertising that keeps your brand in front of bounced traffic after people leave your website. For most sites in China, only 2 to 4% of traffic converts on a first visit.
Retargeting is the tool that goes after the other 96 to 98%.

The Classic Mechanism: Pixel, Cookie, Ad Slot
Traditional retargeting is cookie-based. You place a small piece of code on your website, often called a pixel. It is invisible to visitors and does not affect page performance.
Every time a new visitor arrives, the code drops an anonymous browser cookie. Later, when that visitor browses other sites in the same ad network, the cookie identifies them and the ad platform knows when and where to serve your creative.
This means your ads only reach people who already visited you. It concentrates spend on qualified traffic, on people who already know your name and recently showed interest. That is why marketers who run it properly report a higher ROI than on most other channels.
That is the textbook version. In China in 2026, it only describes about a third of what actually works.
What Changed: PIPL, Dead Cookies, and Closed Apps
Three things broke the old model at the same time.
1. The law. China’s Personal Information Protection Law, in force since November 2021, treats behavioural ad targeting as a regulated activity. You need explicit, separate consent before using someone’s personal information for marketing. You need separate consent again before sharing that data with an advertising partner. If a user refuses targeted advertising, you must still deliver the core service, and you must give them a way to switch personalised ads off. Cross-border transfer of that data needs its own specific consent, and a CAC security assessment above 100,000 individuals. Fines reach 50 million RMB or 5% of prior-year revenue. The China Briefing guide to PIPL compliance lays out the thresholds. In practice this means one thing for a foreign brand: your consent banner and your privacy policy on the Chinese site are not paperwork, they are the legal basis of your whole retargeting setup.
2. The cookie stopped travelling. Chinese consumers live inside apps, not inside a browser. A cookie dropped on your WeChat-opened H5 page does not follow anyone into Douyin, into Xiaohongshu, or into Taobao. Each of these is a closed environment with its own identity graph. There is no open programmatic layer stitching them together.
3. First-party data became the only durable asset. Everything that still works is built on data you own and can upload yourself.

The Three Re-Targeting Systems That Actually Work in 2026
1. Private Domain: WeChat Account, Group, Mini-Program
Private domain traffic is retargeting without ad spend. You capture the visitor once, into a channel you own, then you can reach them again for free.
The mechanism is worth spelling out. A Baidu ad sends a user to a Chinese landing page. On that page you do not push a purchase, you push a QR code with a reason to scan: a size guide, a 60 RMB coupon, a sourcing PDF. The scan follows the user into your WeChat Official Account or, better, into a WeCom (企业微信) contact belonging to a real salesperson. From that moment the user has an OpenID in your system. You can push an article, send a one-to-one message, drop them into a product group, and pull them into a mini-program store where you see their browsing and cart behaviour directly.
The numbers behind this are not small. iResearch put China’s private domain e-commerce sector at 2.3 trillion RMB in 2024 and projected it past 3 trillion in 2025. WeChat Mini Shop GMV grew 1.92x year on year, with order volume up 2.25x. That is where repeat purchase now lives, and it explains why a WeChat mini shop for foreign brands is usually the second thing I ask a client to build, right after the Chinese landing page.

2. Owned Audiences on Baidu and Tencent Ads
Both platforms let you upload your own customer list and target it. This is the part most foreign brands never activate.
On Tencent Ads, the tool is the DMP. You upload a hashed list of phone numbers, OpenIDs, or IMEI identifiers and it builds a 人群包, an audience package. You can then do two things with it. First, retarget that exact list across WeChat Moments, Official Account banners, Mini-Program slots, and Channels video. Second, and more useful, ask the platform for a lookalike expansion built from the same seed. The API also lets you refresh those audiences automatically, so a list of yesterday’s cart abandoners can be live in your ad account this morning. WeChat search and Channels ad revenue both grew around 70% in 2025, which tells you where the inventory is going.
On Baidu, the equivalent runs through Baidu Ocean Cube and the Baidu Union display network. You build audiences from search behaviour, from site visitors captured by the Baidu Tongji tag, and from uploaded lists, then serve them across the Union publisher network. Opening a Baidu Union account is the step that opens display inventory outside search results.
One warning. Uploading a customer list you collected in Europe into a Chinese ad platform is a cross-border transfer under PIPL. It needs its own consent and, depending on volume, an assessment. Build your Chinese list in China.
3. Behaviour Retargeting Inside Douyin’s Closed Loop
Douyin does not do retargeting the way Google does. It does something arguably stronger inside its own walls.
Every interaction with your content is a signal the algorithm keeps: a 3-second watch, a comment, a profile visit, a product card click, a live-stream entry, an abandoned cart. Through Qianchuan (千川), Douyin’s ad platform for merchants, you can build audiences directly from those behaviours and serve them a different creative. Someone who watched 80% of a video but never clicked gets the product demo. Someone who entered the store and left gets the price-anchored offer. The whole sequence happens inside the app, so nothing depends on a cookie surviving anywhere.
This is why a growing number of brands are shifting budget across. The mechanics behind that shift are covered in our piece on brands migrating from Tmall to Douyin.

When Does Retargeting Work?
Retargeting is a branding and conversion tool. It works when it sits inside a wider strategy, not as a standalone tactic.
It pairs with content marketing and paid display. Those drive traffic but do nothing for conversion. Retargeting can lift conversion but it cannot bring anybody to your site in the first place. It is always the second step.
Your best odds come from using one or two channels to generate traffic, then recontacting that traffic several times before you write it off.
How Do I Drive The Traffic In The First Place?
Set up a proper Mandarin landing page hosted so it loads fast in China, then run Baidu PPC for foreign businesses on a tight set of 15 to 30 character keywords. Banner buys can supplement it. PPC remains the fastest way to test demand.
In parallel, build organic visibility: Chinese content, backlinks, presence in forums, Baidu Baike, Zhihu, and news sites. SEO is slow but it brings the most qualified traffic, which makes every retargeting sequence downstream cheaper.
One 2026 addition. A large share of Chinese product research now starts in Xiaohongshu search or in an AI assistant like Doubao or DeepSeek rather than on Baidu. That traffic arrives with high intent and no cookie history at all. If your only capture mechanism is a pixel, you lose it entirely. A QR code on the landing page catches it.

Baidu vs Google Re-Targeting
Structurally the two are similar: campaigns, ad groups, keywords, cost per click, keyword quality score. A marketer coming from Google Ads will recognise the layout.
The differences matter. Baidu’s conversion tracking is weaker, and it does not tie clicks back to outcomes as cleanly, which makes keyword-level optimisation harder. The interface is clunkier. Reporting arrives with more lag. On the other hand Baidu allows product-level remarketing that is genuinely precise: if a user searched 滑雪板 to research winter sports, a sports retailer can serve that exact category with a ski-specific offer on Union banners afterwards.
Baidu Index also lets you compare keyword demand over time, split by device and by region, and pull the related-query growth indexes around a campaign. Use it before you build the keyword list, not after.

A Client Case: Marc and the Belgian Jewellery Brand
Marc runs a Belgian jewellery brand, mid-price, strong in Benelux, six years of trying to crack China with nothing to show for it. When we started he was spending about 70,000 RMB a month on Baidu search and converting 0.8% of visits.
What failed first: a pure display retargeting campaign on Union banners. We ran it for two months, spent 40,000 RMB, and it produced 11 orders. The reason is simple in hindsight. Jewellery at his price point is not an impulse purchase in China, and a banner cannot answer the two questions buyers actually had, which were about authenticity and after-sales service.
What worked: we rebuilt the landing page around a WeCom QR code offering a free engraving consultation. Scan rate hit 14% of landing page visits. Those contacts went to two Chinese-speaking sales staff who answered questions in the chat. In parallel we uploaded the scanned contacts as a Tencent DMP audience and served them Moments ads showing the workshop, not the product.
Seven months later: conversion rate on the same traffic volume at 2.6%, cost per acquisition down 44%, and 31% of revenue coming from repeat orders placed inside the mini-program by people who had first landed through a Baidu ad the previous quarter. The paid traffic did not change. What changed was that he stopped losing it.
What To Put In Place On Monday Morning
If you take one thing from this article, take this list. It is roughly the order I give clients.
- Put a WeCom or Official Account QR code on your Chinese landing page. Give a real reason to scan. This single change usually captures 8 to 15% of traffic you were losing at 100%.
- Fix your consent layer. Separate checkbox for marketing use, a working opt-out, a Chinese-language privacy policy naming what you collect. Without it your audience data is legally unusable.
- Install Baidu Tongji and the Tencent conversion tag properly. Not just the homepage. Product pages, form pages, thank-you pages.
- Create your first Tencent DMP audience from existing contacts, then build a lookalike from it. Budget 5,000 RMB to test before you judge.
- Build the Douyin behaviour audiences if you already run content there. Video viewers at 75% completion are your warmest list and it costs nothing to segment them.
- Decide your sequence, not just your ads. Day 1 reminder, day 3 social proof, day 7 offer. Rotate creative every two weeks or frequency kills performance.
- Measure at 90 days, not 14. Chinese consideration cycles for foreign brands run long. Judging a retargeting programme after two weeks is how good campaigns get killed.

FAQ
Can I run retargeting in China without a Chinese business licence?
Partly. You can buy Baidu and Tencent media through an authorised overseas reseller account without a WFOE, and you can run a .com site aimed at Chinese users. What you cannot do without a local entity is get an ICP licence, host inside China, open a WeChat Official Account with payment and mini-program functions, or use WeCom for sales. Since private domain is where most of the value sits now, the licence question is really a question about how serious you are.
How long before I see results?
You need traffic volume before retargeting means anything. Below roughly 3,000 monthly visitors your audience lists are too thin for the algorithms to work with. Assuming you have the volume, expect four to six weeks to see the first movement in conversion rate and around three months for a stable cost per acquisition. Private domain takes longer to pay off and pays off larger, because the contact you capture this month keeps costing zero next year.
Does PIPL actually get enforced against foreign brands?
Yes, and the pattern is consistent. CAC enforcement waves have targeted apps, advertising, education, healthcare and online finance, and foreign companies have been among those cited for cross-border consent failures and for not appointing a data protection officer above the processing threshold. Nobody gets fined for a first minor issue, but a brand collecting Chinese customer data with a translated European privacy policy and no separate marketing consent is carrying a real risk. Fix it before you scale, it takes a week.
Re-targeting in China is no longer a cookie trick. It is a first-party data discipline, and the brands that treat it that way pay a fraction of what their competitors pay for the same customer.
About GMA
GMA builds and runs China retargeting systems for foreign brands: Baidu and Tencent audience setup, WeChat private domain capture, Douyin behaviour targeting, and the PIPL consent layer underneath it all.
We work mostly with mid-size European and North American brands already spending on Chinese media and unhappy with what comes back.
Send me your Baidu account numbers and I will tell you what you are losing. Contact us through marketingtochina.com or reach me on LinkedIn.
Sources: CNNIC 57th Statistical Report on China’s Internet Development (February 2026), QuestMobile 2025 mobile internet half-year report, China Briefing PIPL compliance guide.