TOCANAN
繁中Free GEO audit

Field notes

Xiaohongshu KOL Metrics and KPIs for Brands: CPE, Cost per Read, Engagement Rate and Viral-Post Rate

The KOL and KOC metrics that matter on Xiaohongshu (RED) — engagement rate, CPE and the KPIs brands should track for Chinese influencer campaigns.

By Eden Lau · · updated

Four numbers still decide whether a Xiaohongshu KOL campaign worked: what each read cost, what each interaction cost, how much of the audience interacted, and how many notes broke out. That has not changed since the 2024 version of this guide. What has changed is where the numbers come from and which benchmarks can be trusted. Xiaohongshu’s official creator marketplace, 蒲公英 (Pugongying), now shows most of these metrics on every creator’s card before you spend anything, and the practitioner consensus has shifted from cost per read towards cost per engagement paired with unique-visitor cost. Every benchmark below carries a source and a date. The unattributed thresholds from the 2024 version have been removed.

KOL, KOC and KOS on Xiaohongshu

Digitaling’s Xiaohongshu glossary (13 August 2024) draws the lines this way. A KOL (key opinion leader) holds more, and more accurate, product information than the group that trusts them, and can move that group’s purchasing. A KOC (key opinion consumer) is a consumer whose recommendations move their own friends and followers. A KOS (key opinion sales) is a content creator with real selling skill and deep category knowledge. The metrics below apply to all three; what changes is the scale, and therefore which statistic you should trust.

1. Cost per read (阅读单价, sometimes written CPR)

Formula: creator fee ÷ reads.

This is the screening metric. Pugongying calculates a creator’s 阅读单价 from their quoted price and their average reads, so it is visible before the brief goes out. Two cautions. Use the median reads (阅读中位数), which Pugongying defines as the middle value of a creator’s recent notes sorted by reads, rather than the average, because one viral note inflates an average for months. And do not stop here: cheap reads from an audience that never interacts are not cheap.

2. Cost per engagement (CPE)

Formula: creator fee ÷ (likes + saves + comments).

CPE is the metric most brands optimise for, because an interaction is a stronger intent signal than a read. One current practitioner benchmark comes from marketer Vic (赵子辰), writing on 人人都是产品经理 on 22 February 2025: S-tier notes at CPE ≤ ¥8 with an engagement rate of 5% or more, A-tier at CPE ≤ ¥12 with 3% or more, and anything above ¥15 cut. He pairs CPE with unique-visitor cost (spend ÷ unique visitors) as the two numbers to watch, and treats the pair as user-acquisition cost rather than media cost.

Those tiers are one practitioner’s cut-offs for 2025, not a platform standard, and the article does not split them by category. CPE differs by category, so set your tiers from your own last campaign and compare creators within a category, never across categories.

3. Engagement rate (互动率)

Formula: (likes + saves + comments) ÷ reads.

Pugongying shows this on the creator card and defines it the same way. A widely quoted pass line of 3.5% to 5% comes from a 2021 guide to the platform; it is old enough to treat as a floor rather than a target. More useful than the headline rate is its consistency: a creator whose last twenty notes sit steadily near the same rate is lower-risk than one whose average is carried by two outliers. Many practitioners weight saves above likes because a save is a plan to act; if the goal is 种草 (seeding), track the saves share separately.

4. Viral-post rate (爆文率)

Formula: viral notes ÷ notes published in the campaign.

Xiaohongshu publishes no official definition of a viral note (爆文). Qian-Gua, a major third-party Xiaohongshu analytics provider, uses 1,000+ likes as its working definition in its viral-note research. Agencies also use tiers by total interactions, commonly 1,000+, 10,000+ and 100,000+, but that is a convention rather than a standard, so state which definition a report uses. Pugongying lists a creator’s 爆文率 among its screening dimensions. The 2024 version of this guide put the campaign target at 20%; that figure was unattributed and is gone. Set the target from your own previous campaign or from the category median in your analytics tool.

Reading the four together

Pattern Likely meaning Action
High reads, low engagement rate Bought or off-target reach, or a weak fit between creator and product Drop the creator for this category
Low CPE and high engagement rate on a small account A strong KOC Scale with more creators of the same profile, not a bigger one
Good engagement rate, poor CPE Over-priced for the audience they deliver Renegotiate or replace
Good reads and CPE, low viral rate Steady but not breaking out Fine for always-on; not for a launch

For KOC programmes the same metrics apply, but reads per note are small, so judge creators in batches and lean on medians. CPE is comparable across account sizes; raw reads are not.

What 2026 adds to the four

Declared sponsored notes (报备笔记). Notes registered through Pugongying give the brand a dashboard with impressions and reads for each note, which the creator’s own view may not show in full. That makes click-through rate (reads ÷ impressions) a measurable campaign metric. See How to check your Xiaohongshu post CTR and reach.

Comments as data. With 60 million+ new comments a day on the platform (Qian-Gua, May 2026), the threads under campaign notes are a readable record of objections, questions and comparisons. Read them before the next brief.

AI visibility. Xiaohongshu notes are among the sources Chinese AI assistants such as DeepSeek, Kimi and Doubao draw on when asked about a brand. A 2026 campaign KPI can therefore include whether those assistants mention the brand for category questions after the campaign, and what they say. Tocanan’s Xiaohongshu brand guide explains the link, and GEO monitoring measures it across Western and Chinese engines.

Timing also feeds these numbers: a note published when the creator’s followers are online front-loads the early reads that trigger wider distribution. See Best times to post on Xiaohongshu in 2026.

Frequently asked questions

Which single metric matters most for a Xiaohongshu KOL campaign? Cost per engagement, read together with engagement rate. Cheap reads with no interaction indicate an audience that is not listening; a good CPE with a steady engagement rate indicates one that is.

What counts as a viral note on Xiaohongshu? There is no official threshold. Qian-Gua’s research uses 1,000+ likes; agency tiers by total interactions (1,000+, 10,000+, 100,000+) are common. State the definition a report uses.

Does Tocanan run KOL campaigns? Tocanan is a measurement and consulting practice, not a media buyer. It works on the metrics side: benchmarking creators, reading comment data, and tracking whether a campaign changes what Chinese AI engines say about the brand.

Sources

Find out what the engines say about you

The audit runs a category query set across every engine we cover and returns your baseline AIGVR, with the specific answers where you are missing or misdescribed.

Get a free GEO auditTalk to us first