Do shared commenters reveal shared repost audiences?

A reanalysis of 2 selected Weibo creator groups shows why comment overlap and repost overlap answer different questions.

An analysis of 2 selected SAGraph Weibo interaction files, released with research published in 2025. The files do not specify post dates.

Abstract

For 11 creators associated with a skincare product, 769 of 10,469 recorded commenters appeared under at least 2 creators. In an 8-creator toothbrush group, the equivalent was 556 of 5,035. The corresponding repost figures were 144 of 13,376 and 1,454 of 10,240. These historical records cover selected posts and do not measure campaign reach. [1][2]

A brand considering several creators may want to know whether it will reach new people or meet some of the same people again. Public interaction data can help describe a portion of that picture. It cannot tell the brand how many people saw a campaign.

Different interactions, different groups

Abstract conversation and sharing symbols in overlapping shapes.
Conceptual illustration. The shapes do not encode the study results.

What we compared

We analyzed the released SAGraph interaction records for 2 product groups: skincare and oral care. The source identifies people who promoted each product and records comments and reposts associated with their collected Weibo posts. We counted each responding account once per creator and interaction type, then counted how many appeared for at least 2 creators in the same group. [1][2]

The skincare list contains 14 entries but 11 distinct creator IDs; we deduplicated them. The oral-care list contains 8 distinct IDs. Each selected creator had comment and repost records. The source collected public posts through a selected network of accounts. The released interaction records do not contain post dates or identify campaign posts, so the figures describe collected interactions rather than a campaign period.

The answer changes with the interaction

In the skincare group, 769 of 10,469 distinct accounts in comment records appeared for at least 2 creators: 7.3%. Among accounts in repost records, 144 of 13,376 appeared for at least 2 creators: 1.1%. [2]

In the oral-care group, 556 of 5,035 commenters appeared for at least 2 creators: 11.0%. Among accounts in repost records, 1,454 of 10,240 appeared for at least 2 creators: 14.2%. Each percentage uses the distinct accounts recorded for that interaction in its own group.

The direction differs between the 2 groups. These selected creators and posts do not establish a typical rate or a rule about product categories. They show why a report should name the action and group behind an overlap percentage.

Accounts recorded for at least 2 creators

The percentage of distinct accounts recorded for at least 2 creators, calculated separately for comments & reposts in each group.

Recorded overlap by group & interaction
Product group & interactionDistinct accountsRecorded for 2 or more creatorsShare
Skincare: comments10,4697697.3%
Skincare: reposts13,3761441.1%
Oral care: comments5,03555611.0%
Oral care: reposts10,2401,45414.2%
Each action is calculated separately within its product group. Source: SAGraph interaction files; Influence IQ analysis. The source does not identify campaign posts or post dates.

Shared in both measures

Only 16 accounts in the skincare group appeared for at least 2 creators in both comment and repost records. In the oral-care group, 256 accounts met both conditions. The threshold applies separately to each interaction; the creators associated with an account’s comments need not be the same creators associated with its reposts. [2] Among accounts recorded reposting for at least 2 creators, 128 in skincare and 1,198 in oral care did not also meet the multi-creator comment condition. Those 16 accounts represent 2.1% of the 769 shared commenters in skincare; the 256 represent 46.0% of the 556 shared commenters in oral care.

The two populations also differ. Some accounts appear in comment records, some in repost records and some in both. A person’s absence from a recorded interaction set does not establish that they never viewed, followed or interacted with a creator elsewhere.

Shared accounts and repeated appearances

A shared account and an extra appearance are different counts. In the skincare comment records, the creator-level distinct-account counts sum to 11,525. Deduplicating across creators leaves 10,469 accounts, a difference of 1,056 appearances. Of those accounts, 769 appear for 2 or more creators. An account appearing for 3 creators contributes 1 shared account but 2 extra appearances. [2]

In the oral-care comment records, 5,751 creator-level appearances reduce to 5,035 distinct accounts. That is 716 extra appearances, while 556 accounts appear for at least 2 creators. Naming both measures prevents a percentage labelled simply “overlap” from concealing its denominator.

How to use an overlap report

Ask the provider what an account had to do to enter the analysis, which posts and dates were included, and whether a number represents shared accounts, repeated appearances or an estimated percentage. Ask for the base used in each percentage and the number of creators being compared.

For a creator shortlist, keep comment and repost results separate. If an account appears across several creators, that is evidence of recorded participation in the sampled posts. It does not show that the same account saw a paid placement, nor does a missing overlap show that the audiences are separate. Request deduplicated campaign exposure from an eligible measurement source when that is the decision you need to make.

Read the measure before the percentage

An illustrated person examines a report with conversation, sharing & group symbols.
  1. Interaction

    Are these comments, reposts or views?

  2. Scope

    Which creators, posts and dates are included?

  3. Counting

    Is the figure shared accounts or extra appearances, and what is the base?

Ask what action, posts, dates & denominator an overlap report uses.

How we analyzed the records

We used the product list and anonymized interaction files published with SAGraph. Within each product group we selected the distinct listed promoter IDs, retained comment or repost records, and excluded a creator’s interaction with their own account. We deduplicated actor IDs within each creator and interaction type. We then counted each actor’s number of represented creators and calculated the share represented by 2 or more. [1][2] The 2 selected interaction files are the smallest by file size.

The source describes multi-round interactions under posts, so a comment record can include participation in a discussion rather than a direct reply to the creator. The files contain no post identifiers or post dates. Collection coverage can vary across creators, and the source uses selected network-based sampling. Identities are pseudonymous account IDs; they are not verified people. The data do not measure views, follower overlap, conversion, or causation.

The results and analysis code are available in the accompanying package. The raw files remain with the source.

The downloads contain aggregate counts & code. The original interaction records remain with the source.

Sources & notes

  1. Zhang and colleagues, SAGraph: A Large-Scale Social Graph Dataset with Comprehensive Context for Influencer Selection in Marketing (2025).
  2. Influence IQ analysis of the released SAGraph product and interaction files.
  3. SAGraph is released under Creative Commons Attribution 4.0. The authors describe their selection, collection & anonymization methods in the study.

Every language edition describes the same selected Weibo records & calculations.

Related reading