
Rules
Part of Creator marketplaces: five questions to put to the terms
Creator marketplaces: five checks to run during a trial
Creator marketplaces and their platforms, tested in a trial: five checks on coverage level, revoked connections, live share, data model and source changes.
A trial is the cheapest moment to learn what a creator marketplace can and cannot show you. A comparison grid says that a venue supports a platform, and a trial lets you test the claim. Five checks, run on creators whose numbers you already know, show what a demo leaves out.
The venues this applies to are the platforms agencies shortlist, CreatorIQ, Aspire, GRIN, Upfluence and Traackr among them, and the marketplaces creators join directly, such as TikTok Creator Marketplace and Shopify Collabs. None of them is ranked here, because the products are not measured on the same thing, and the affiliate platform questions take the same question-first approach.
What to take away
- Supports is not a level. Ask for each platform whether the data is authorized, public, modeled or typed in by the creator.
- Permissions set the ceiling. What a venue can show you depends on what the creator granted, not on how good the product is.
- Test with evidence you can opena creator's own analytics, a revoked connection, a full export.
- Every venue in this category is downstream of decisions it does not control, so ask what happened the last time a source changed.
The five checks in order
Run these during the trial, before any contract, and write down each answer. A vague answer is itself a result.
- Place each platform in a level, one row per platform.
- Watch what the interface shows when a connection is revoked.
- Ask what share of the roster holds a live connection.
- Open the data model and a full export of the record.
- Ask what the venue did the last time a source changed.
Check 1: place each platform in a level
Ask the venue to place each platform it claims, Instagram, TikTok, YouTube or Twitch, into one of four levels: authorized, public, modeled or declared. Give one row per platform rather than one row for the vendor. Vague answers here are the single most reliable warning sign in the category.
Authorized means the creator connected their own account and granted access, which supports audience composition and reach over a stated window. Public means collected from pages anyone can see, which supports follower counts, posting cadence and visible engagement, but not audience makeup.
Modeled means inferred from public data and a sample. It gives an estimate with error bars, useful for ranking a shortlist, and never a fact about one person. Declared means typed into a profile by the creator, and it proves nothing until corroborated.
Four Levels of Coverage
Authorized
- Data source
- Creator connected account
- Supports
- Audience, reach, history
- Cannot support
- After access revoked
Public
- Data source
- Visible pages
- Supports
- Follower counts, cadence
- Cannot support
- Audience makeup
Modeled
- Data source
- Public data sample
- Supports
- Estimates with error bars
- Cannot support
- Claims about a person
Declared
- Data source
- Creator typed profile
- Supports
- Nothing on its own
- Cannot support
- Everything until corroborated
Then test the answer. Pick three creators you know and ask each for their own analytics over the same window. Put those figures beside what the venue shows. Where a figure is modeled, treating it as measured is the quiet error described in the guide to creator discovery tooling.
Public collection can mean web scraping, which is legitimate but fragile, and the law around it has been contested. In hiQ Labs v. LinkedIn, the Ninth Circuit took up whether LinkedIn could stop a competitor from collecting information that users had shared on their public profiles. A number gathered this way can stop being available without warning.
Checks 2 and 3: permissions set the ceiling
None of this is decided by the venue's engineering. It is decided by whether a creator connected an account, what scope that grant covered, and whether it is still valid.
An authorized connection can be revoked at any time by the person who granted it, so a profile that was rich last month can be thin today without anyone doing anything wrong.
For check 2, ask a creator to disconnect during the trial and watch the interface. A venue that silently falls back to estimates is presenting two different kinds of number in the same place.
Scope varies too. On Instagram, permission to read a creator's basic profile through the Instagram Graph API is a separate grant from permission to read audience insights.
A venue's coverage of a platform is coverage of one permission there, so ask which fields need which grant.
Connections expire, which is check 3. Ask how often creators are prompted to reconnect, and what proportion of the roster currently holds a live connection. That last figure is the honest version of the coverage claim, and a venue can produce it if you ask plainly.
Check 4: the data model and the full export
The structure underneath decides what you can get out later, and it is the least examined part of any evaluation. Two questions settle most of it.
- Is a creator one record across platforms, or one record per platform? Both are defensible. Only one lets you ask what a person did across their whole presence, and the reconciliation involved is a record linkage problem the venue has either solved or ignored.
Data Model Questions
- One record across platforms or per platform?
- Rights terms as fields or attachment?
- State for neither active nor rejected?
- Export whole record, human-readable?
- Audience figures stored with date and source?
A third question follows from the export: what you keep if things go badly. Historical data already collected under an authorized connection is a different asset from live access, and keeping it is a contract question. The discovery platform contract checks list what to settle before signing.
The record also matters beyond switching vendors. The FTC says a company that pays and directs influencers needs programs to train and monitor them, and a readable record of messages and approvals is one thing such a program can draw on.
Check 5: what happened when a source changed
Instagram, TikTok and YouTube tighten, loosen and redefine what they expose, sometimes with notice and sometimes without. Every venue in this category depends on that access, and none of them controls it. The five depths of platform support show how far the same dependency runs in campaign tools.
The useful question is historical: what did you do the last time a source changed what it exposed, and how did customers find out?
A vendor who has been through it will answer with specifics, including how long the gap lasted. A vendor with no answer has not been through it yet, which is a fact about their age rather than their quality, and it is still worth knowing.
Put the question on the first screening call. The screening call questions show how an agency can ask before the checks begin.
Coverage is not audience quality
Excellent authorized coverage tells you nothing about whether an audience is genuine, because the two are different problems with different evidence. Coverage is about access. Quality is about interpretation, and the guide to fraud evidence behind a coverage claim shows what each kind of check can and cannot establish.
Keep them apart in the trial too. A venue that answers a coverage question with a quality claim, or the reverse, is either confused or hoping you are.
Common questions
A venue claims full coverage of a platform. What should I ask?
Ask which level applies to each platform. Public data may be collected for everyone while authorized data exists only for the creators who connected, so ask for the proportion with a live connection and the claim resolves itself.
Should we require authorized data for every booking?
For anything consequential, ask the creator for their own analytics over a stated window and keep the file. That works regardless of which venue you are on and it survives a connection being revoked.
Is breadth or depth better?
Whichever matches where you actually buy. Breadth is worth paying for when you are exploring categories, and depth is worth more when your spend concentrates in one place.
The estimates from two venues disagree. Which is right?
Possibly neither, and the disagreement is informative. Ask both what the estimate is built from and over what window, then treat the range as the answer rather than either number.







