Your KOL campaign got views and no ROI. Here is the fix.

The deck looked great. Millions of views, and a slide of warm comments to show people cared. Then you opened the sales report, and the line hadn’t moved.

A KOL campaign with no ROI has a cause, and you can find it. We book and audit creator campaigns on Xiaohongshu (小红书) and Douyin (抖音) from Shanghai and Hong Kong. Sometimes the agency got it wrong. Sometimes it did nothing wrong, and the product wasn’t ready. Start with the symptom you can see, then work down the six in order. Every rule and figure below was checked in October 2026.

What you see	Where it likely broke	Where to check
Big views, buyers in the wrong cities or ages	Wrong audience	Viewer profile in the platform report
Likes and saves, no store visits	No path to purchase	Was a store or product widget set at order?
Comments about the creator, none about the product	The post read as an ad	Comment text and completion rate
Views far above the creator’s usual posts	Inflated numbers	Natural versus paid traffic split
Sales arrived after the report closed	Wrong window	Searches and store visits 15 to 30 days out
Searches and visits rose, orders did not	Product or price	Store page, reviews, local prices

Views are not what you paid for

A view means a post crossed someone’s screen. On its own, it sells nothing. What you paid for is the next step: a search for your brand, then a visit to your store.

The platforms count those steps themselves. Xingtu (巨量星图), the creator marketplace of Douyin (抖音), tracks searches made right after a video and store visits in the two weeks that follow.

Xingtu’s campaign report counts store visits made within 15 days of watching a creator’s video. Its “search after viewing” counts a brand search made as the viewer’s next action. Source: Ocean Engine (巨量引擎), Xingtu help center, June 2026. https://www.xingtu.cn/help-center/demander/115182

Douyin has also admitted its old reports undercounted.

Xingtu’s earlier report counted only orders from videos with a Douyin store cart attached, so seeding value was underestimated. A viewer rarely watches once and then taps the cart to buy. Source: Ocean Engine (巨量引擎), Xingtu help center, March 2024. https://www.xingtu.cn/help-center/demander/133677

So name the deliverable before you book. A view has to become interest, interest a search or a visit, a visit an order. Each of those four links can break, and the table shows where to look for each one.

Link	What breaks it	What fixes it
View to interest	Wrong audience, or a post that reads as an ad	Creators picked on audience match; creative freedom
Interest to search	A brand name nobody can find	One Chinese brand name everywhere; search ads on it
Search to visit	No store, or no widget on the post	Store or product widget set when you order
Visit to order	Price, reviews, product page, stock	Fix the product page before the next round

Failure 1: you bought reach from people who can’t buy you

Follower count is the easiest number to put on a proposal. A Hunan court case about a Douyin (抖音) creator shows how little it can mean.

A brand paid 100,000 yuan for live selling by a Douyin creator with a million followers, with a 550,000-yuan sales floor. Actual sales were 8,209.26 yuan. The court ordered 98,507.41 yuan refunded and told brands not to choose creators on follower count and fame alone. Source: Red Star News (红星新闻), via Sina (新浪), August 2026. https://finance.sina.cn/tech/2026-08-26/detail-iniprytk0877893.d.html

On Douyin, most creator selling is done by smaller accounts.

Creators with fewer than 1 million followers made more than 85% of creator-driven sales on Douyin in 2025. Source: The Beijing News (新京报), reporting Douyin E-commerce’s 2025 data report, January 2026. https://m.bjnews.com.cn/detail/1768473381129936.html

Up close, the mismatch is sometimes plain. The followers are students, and the product costs a week of their budget.

Both marketplaces let you check: Xingtu (巨量星图) on Douyin, and Pugongying (蒲公英) on Xiaohongshu (小红书).

Xingtu’s custom report can compare the people a video reached with the creator’s followers, to see if they match what the brand expected. Pugongying, Xiaohongshu’s creator marketplace, gives each note a viewer profile to judge the match with the brand’s target buyers. Source: Ocean Engine (巨量引擎), Xingtu help center, August 2026; Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), October 2026. https://www.xingtu.cn/help-center/demander/119060 ; https://pgy.xiaohongshu.com/help/home

Failure 2: there was no path from the post to a purchase

A Western brief often ends with “link in bio.” In China that route is closed, by rule. Xiaohongshu (小红书) now ties the penalty to Pugongying (蒲公英), its creator marketplace.

Xiaohongshu’s revised trade-diversion rules took effect on September 18, 2026. Penalties now reach commercial rights, including suspending business cooperation and Pugongying order-taking. Source: Xiaohongshu e-commerce rules center (小红书电商学习中心), September 2026. https://school.xiaohongshu.com/rule/detail/17/119931

On Douyin (抖音), Xingtu (巨量星图) screens for the same thing before a video airs.

Since March 14, 2025, Xingtu rejects at pre-review any creator video that steers users off Douyin. Source: Ocean Engine (巨量引擎), Xingtu help center, December 2025. https://www.xingtu.cn/help-center/demander/144149

So the path has to sit inside the platform, set when you order. On Xiaohongshu (小红书), a creator note can carry a pinned comment widget that opens your store, a product page or a search results page.

The Pugongying comment widget works for products in a Xiaohongshu store, and a note that is already published cannot have one added. Source: Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), October 2026. https://pgy.xiaohongshu.com/help/home

The search version opens the results for a keyword you set, so a buyer lands on more notes about you instead of a dead end. A link to an outside store exists, but not on ordinary notes.

Xiaohongshu’s direct-to-purchase link works only on ad notes, not on every note in the community. Source: Jiemian (界面新闻), May 2025. https://www.jiemian.com/article/12752238.html

Douyin’s marketplace has a catch of its own.

On Xingtu, each task can carry one add-on widget, and it cannot be changed after the video is published. Source: Ocean Engine (巨量引擎), Xingtu help center, June 2026. https://www.xingtu.cn/help-center/demander/109142

If the campaign ran without a store or a widget, the views had nowhere to go. Blame the setup for that one, not the creators.

Failure 3: the post was an ad, and the audience knew it

Ask anyone who scrolls Xiaohongshu (小红书) at night why they’re there. They want reviews from people like them, and they scroll past scripts.

A creator’s review video was the most common way consumers got “seeded” on a product, at 61.49%. Soft-sell advertorials came last, at 24.91%. Source: iiMedia Research (艾媒咨询), June 2026. https://www.iimedia.cn/c1094/111612.html

Hiding the ad doesn’t help either.

The top reason consumers got seeded on a product was a real, existing need, at 49.10%. “I thought it was pure sharing and didn’t notice it was an ad” came last, at 32.85%. Source: iiMedia Research (艾媒咨询), June 2026. https://www.iimedia.cn/c1094/112164.html

The platforms have started punishing the hard sell, too. Xiaohongshu began a clean-up on October 1, 2025.

Xiaohongshu’s clean-up targets stiff ad insertion and templated look-alike notes. Penalties include making a note visible only to its author or followers, and limiting its traffic. Source: Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), October 2026. https://pgy.xiaohongshu.com/help/home

Douyin (抖音) and its Xingtu (巨量星图) marketplace followed six weeks later.

From November 12, 2025, Xingtu bars four abrupt brand placements. Two are a jump into the ad with no lead-in and a stock line that breaks the story. The others are a placement unrelated to the story and brand footage over half the frame. Source: Ocean Engine (巨量引擎), Xingtu help center, June 2026. https://www.xingtu.cn/help-center/demander/147608

Look at your brief. Did it run to three pages of key messages? Was there a list of banned words, and a product shot required in the first second? Did every post end up saying the same sentence in the same order? Then the creator was reading your ad copy aloud. People follow these creators because they don’t sound like ads.

The audience heard it.

Failure 4: the numbers were inflated, and nobody checked

Some of the views in that deck may not have been people at all.

In 2022, 53.2% of the average KOL’s followers were invalid, Miaozhen found. Source: Miaozhen (秒针系统), March 2023. https://www.miaozhen.com/all-reports/marketingeffiency/806.html

We found no newer share in print. Enforcement on Xiaohongshu (小红书) shows the trade is still busy.

Between August 22 and September 20, 2026, Xiaohongshu acted against 76,000 accounts and 357,000 notes for interaction cheating. It removed 170 million fake follows and 980 million fake likes and saves. Source: Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), September 2026. https://pgy.xiaohongshu.com/help/home

Inflation also gets into reports through what is counted, and not only by creators. In a Jiangxi court case reported in October, the agency did it.

An agency claimed 15.752 million plays of a related topic. The court went by the brand’s own films: 38,900 exposures. Likes the agency bought after losing the first trial were not counted, and its claim for 507,000 yuan failed. Source: New Rule of Law Daily (新法治报), via Phoenix Jiangxi (凤凰网江西), October 2026. https://jx.ifeng.com/c/8x1euV2KjUG

Our guide to spotting padding in a China social media report lists ten questions that catch this before the invoice is paid.

You don’t need special software to check. Xingtu (巨量星图) splits each video’s traffic into paid and natural, so you see what the creator brought alone.

If a creator’s estimated plays rise once Xingtu splits out marketing traffic, a large share of that creator’s traffic comes from DOU+ boosting. Source: Ocean Engine (巨量引擎), Xingtu help center, June 2026. https://www.xingtu.cn/help-center/demander/142935

Pugongying (蒲公英) does the same job on Xiaohongshu.

Pugongying’s note report includes both natural and promoted traffic. Its audience analysis has a switch to count only users brought by natural traffic. Source: Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), October 2026. https://pgy.xiaohongshu.com/help/home

Ask for the natural view. Then check how each post was booked, because deals done off the marketplace now hurt the brand as well.

Since January 12, 2026, Xiaohongshu’s brand trust system targets creator deals made off the official platform. It can show a risk warning to users who search the brand. Source: Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), October 2026. https://pgy.xiaohongshu.com/help/home

Failure 5: you measured the wrong window

This is the cheapest failure to fix, because the fix is mostly waiting.

If your deck closed the books a week after the last post, it may have missed the buyers who were still deciding. Xiaohongshu (小红书) has the cases.

At Xiaohongshu’s 2024 WILL conference, its marketing chief said a robot vacuum’s return turned at about 45 to 60 days. A body oil its brand thought was a 7-day decision took many buyers 60 to 90 days. Source: Huxiu (虎嗅), via Sina (新浪), December 2024. https://finance.sina.com.cn/wm/2024-12-20/doc-ineacsym5596846.shtml

The platforms build their attribution around that long tail. A snack is a faster call than a 3,000-yuan appliance. Our rule: read at 15 days and at 30, and at 60 for anything a buyer has to think about. Set the window before you order, because some tools fix it then.

Platform tool	Windows it offers	Set when
Xingtu (巨量星图) product seeding data	1, 3, 7, 14 or 30 days	Product link added at order
Pugongying (蒲公英) Taobao link (小红星)	15 or 30 days	Taobao or Tmall store linked
Pugongying JD link (小红盟)	15, 30 or 60 days	Brand is a JD advertiser
Pugongying audience analysis	Up to 60 days, 30 to 45 suggested	Product bound before analysis
WeChat (微信) creator marketplace (互选)	Click attribution, brand-reported	Fixed when the order is created

Xingtu measures product seeding over 1, 3, 7, 14 or 30 days. Pugongying’s Taobao link reports at 15 and 30 days, and its JD link at 15, 30 or 60. Its audience analysis runs up to 60 days, with 30 to 45 suggested. On WeChat’s creator marketplace, an order uses click attribution, and that setting cannot be changed once the order is created. Source: Ocean Engine (巨量引擎), Xingtu help center, June 2026; Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), October 2026; Tencent 互选 help center (腾讯营销互选平台), August 2026. https://www.xingtu.cn/help-center/demander/140449 ; https://pgy.xiaohongshu.com/help/home ; https://huxuan.qq.com/trade/free/help_detail?post_ids=8348

Failure 6: the product isn’t ready for China yet

Nobody likes this one, and it’s the one an honest agency owes you. If searches and store visits went up and orders stayed flat, the creators did their job.

Seeding speeds up demand that’s already there. The iiMedia survey above put a real need first among the reasons people get seeded. And price gets weighed against what the product does.

Only about 20% of Chinese consumers buy because of low-price promotions. Most weigh whether the price matches the product’s core value and function. Source: NielsenIQ (尼尔森IQ), reported by China Daily (中国日报网), December 2025. https://cn.chinadaily.com.cn/a/202512/02/WS692eb055a310942cc4994948.html

Local rivals have raised the bar. Take cosmetics, one category where the data is public: domestic names now lead.

Domestic brands held 57.37% of China’s cosmetics market in 2025. Source: Xinhua (新华网), from China Consumer News (中国消费者报), February 2026. https://www.news.cn/fashion/20260203/fb1fcc73fb41483ebc1f41fa296a5ec9/c.html

Sometimes the product is fine and the pitch is wrong. At Xiaohongshu’s (小红书) latest WILL conference, a speaker told of a cream sold as a moisturizer, in a crowded category. Seeding showed makeup artists used it as a primer. Repositioned, it became a best seller.

“It was not a bad product. It was the wrong battlefield.” Source: Beijing Business Today (北京商报), on Xiaohongshu’s 2026 WILL conference, December 2025. https://www.bbtnews.com.cn/2025/1223/579681.shtml

Read the comments on your best posts. If buyers asked about price or delivery and got no answer, fix that before the next round.

How to audit a KOL campaign with no ROI in an afternoon

Those are the six. Here’s how to find yours without waiting on anyone.

You need the platform reports, your store data and four hours. Do it soon, because Xingtu’s (巨量星图) report on Douyin (抖音) stops updating.

Xingtu play and like counts update for at most 45 days after a video goes live. A custom report updates daily and stops 45 days after it is created. Source: Ocean Engine (巨量引擎), Xingtu help center, June and August 2026. https://www.xingtu.cn/help-center/demander/115182 ; https://www.xingtu.cn/help-center/demander/119060

Then work through the eight steps, one creator at a time.

Step	Where	What to look for	Red flag
1. Export every report	Xingtu (巨量星图) task report, Pugongying (蒲公英) data center	All posts, all metrics, as files	Screenshots only
2. Switch to natural traffic	Xingtu traffic split, Pugongying natural view	Views the creator brought alone	Natural views far below the total
3. Compare audiences	Viewer profile, followers, your buyer	Age, city and interests that match	Reached people match nobody
4. Check each post’s path	Widget or cart on each post	Clicks on the widget	No widget, or “link in bio”
5. Read the comments	The post itself	Questions on price, use, where to buy	Praise for the creator only
6. Pull searches and visits	Report search data, store backend	Lift 15 and 30 days after posting	Nothing measured past day 7
7. Price each creator	Fee divided by store visits	A cost per visit you can compare	One blended number
8. Confirm how it was booked	Pugongying or Xingtu order records	Every post booked on platform	Cash deals with no order

Once the table is filled in, sort the creators by cost per store visit. Pugongying tells brands to choose on exactly that when the goal is sales.

For a sales goal, Pugongying’s creator guide says to favor creators with a low past cost per store visit and many store visits. Source: Xiaohongshu Pugongying help center (小红书蒲公英帮助中心), April 2026. https://pgy.xiaohongshu.com/help/home

We’d drop the bottom third and give the top third the next brief.

If the review leads to an agency change, see how to switch China agencies without losing your accounts. Secure the creator order history and usage rights before the handover.

How to set up the next campaign so you can measure it

Most of the fixes happen before anyone posts.

First, write the deliverable into the contract. The Hunan court above told brands to spell out the sales floor, how valid orders are counted and the refund ratio. Name the window, and name the metric you’ll judge it on, whether that’s store visits or orders.

Second, build the path at the order. Open the store first. Paste the product link into the Xingtu (巨量星图) order and set the widget. On Pugongying (蒲公英), add the comment widget before the note goes live.

Pick creators on fit and history, not follower count. Ask each one for a viewer profile and their past cost per store visit (a good creator has both to hand).

Then brief the idea and let the creator say it: one product truth and one use case, in their own words.

And test small. Five to ten creators is where we’d start. Read them at 15 and 30 days, then put the money behind the ones that moved visits.

Our influencer marketing service page explains how we vet, brief and report on creators. Our pricing page shows what we charge to run a creator campaign. Our strategy, campaigns and analytics page covers setting KPIs before the first post. The China KOL pricing guide and the KOL vs KOC cost comparison cover what each creator should cost.

FAQ

Why did my China KOL campaign not drive sales?

Six things can break. The creators may have reached the wrong buyers, or the posts had no store or widget to click. Maybe the content read like an ad, or the numbers were inflated. The report may have closed before buyers decided. Or the product, or its price, didn’t suit China yet. Check them in that order, using the platform’s own report data, not the agency’s summary.

How do I check if a KOL has fake followers?

Book through Pugongying (蒲公英) or Xingtu (巨量星图), not off-platform, then use their data. Xingtu splits paid from natural traffic, so a creator who depends on boosting shows it. Compare views with the creator’s usual posts, read the comments for product questions, and check the viewer profile. Miaozhen found 53.2% of the average KOL’s followers were invalid in 2022.

What attribution window should a KOL campaign use?

Match it to how long your buyer takes to decide. Xingtu (巨量星图) offers 1 to 30 days, Xiaohongshu’s (小红书) Taobao link 15 or 30, and its JD link up to 60. In Xiaohongshu’s own cases, buyers took 45 to 90 days. Read at 15 and 30 days for most products, and at 60 for anything expensive. Set the window before you order.

Send us the deck

Send us the agency’s report and your store data. We’ll run the audit above, post by post, and send back the filled-in table. It shows which of the six broke, and which creators are worth booking again. If it was the product, we’ll tell you.