Who runs this guide
Social Intelligence Review is an independent auditor of social intelligence platforms. See the publisher disclaimer for its role and the evidence limitations that apply to every score, review and editorial selection on this site.
What the scorecard measures
The scorecard asks one question: how well does a product help a brand or agency explain a creator campaign after it runs, across China and international platforms?
We chose that question and these criteria. They favor products built for campaign assessment, unrandom among them, because we think it is what a team most needs after a campaign. A team buying creator contracts, a publishing calendar, a service inbox or continuous market monitoring would weigh these products differently, and several of them serve those jobs better than unrandom does.
The criteria
| Criterion | Weight | The question |
|---|---|---|
| Campaign explanation | 30 | Does it read what creators said and showed, connect that to the response, and explain the campaign? |
| Reaction analysis | 20 | Does it show what the audience is reacting to: the creator, the product, or a specific aspect of it? |
| Finished assessment | 20 | Does the team receive a written answer with recommendations, rather than a dashboard to interpret? |
| Creator baseline | 10 | Is each result compared with that creator’s own normal? |
| China and international coverage | 10 | Does it cover named networks on both sides, at useful depth? |
| Source trail | 10 | Can someone get from a finding back to the post or comment behind it? |
Each product receives a level from 0 to 5 on each criterion, in half-level steps. Points are weight × level ÷ 5, and the total is the sum of the six. Every product uses this formula, and no product’s points are set by hand. Equal totals share a rank, and products within a tie appear alphabetically.
What each level means
Campaign explanation
- 5: Speech, on-screen text and visuals are analyzed post by post, connected to audience response and synthesized across a defined campaign, in a complete real output.
- 4: That chain is documented in detail or shown in a substantial output, with one part missing.
- 3: Real post, video or campaign analysis, with parts of the chain not established.
- 2: Campaign reporting with light content interpretation.
- 1: Discovery or rankings with incidental interpretation.
Reaction analysis
- 5: Separates reaction to a creator or celebrity from reaction to the product, by aspect, with buying signals kept separate and original comments available.
- 4: Sentiment by aspect or topic, plus either target separation or buying-intent analysis, documented in detail.
- 3: Comment sentiment and themes with limited scope.
- 2: Audience demographics, overall sentiment, or comment features with little detail.
- 1: Follower profiles or engagement counts.
Finished assessment
- 5: A complete written report that answers the team’s questions with recommendations, in the language needed, generated without hand assembly, shown in a complete real output.
- 4: A full written brief or report with recommendations, shown or documented in detail.
- 3: Exportable reports with some narrative or AI summary.
- 2: Dashboards and data exports with little narrative.
- 1: Spreadsheet export only.
Creator baseline
- 5: Each post against the creator’s own history, sponsored against sponsored where possible, with a disclosed fallback, peer cohorts and a published method.
- 4: Same-creator or account history plus peer comparison, documented in detail.
- 3: Account or creator history, or strong peer or brand benchmarks, less fully defined.
- 2: Category or industry benchmarks, or campaign metrics alone.
- 1: Raw metrics with light comparison.
China and international coverage
- 5: Comprehensive task-specific coverage and limitations documented on both sides.
- 4: Specific analytical scope documented across both.
- 3: Named networks on both sides, with significant access or depth gaps.
- 2: Multiple named networks in China or in international markets.
- 1: One named network.
A global office network or an unspecified claim of global reach earns no credit.
Source trail
- 5: Every finding links to source posts or comments, in exports as well, and an outside reader can follow them.
- 4: Findings connect to retained source records a logged-in user can inspect, with limits stated.
- 3: Source posts and details can be inspected but are not tied to findings.
- 2: Some underlying records can be identified.
- 1: Little evidence of a path to source.
Level 0 means nothing usable was established for that criterion, or the criterion is outside the product’s scope. The score rationale says which. A half level marks a product that meets one level and part of the next.
What the scores are based on
Most products are scored only from public sources: product pages, help centers, release notes and public sample outputs. Evidence access differs between products, as the publisher disclaimer explains. Thin public documentation can hold a product’s score down, and a low level is not proof that a capability is missing.
We keep four kinds of evidence apart: what a vendor claims, what a help article documents, what a reader can inspect in an actual output, and what a controlled test establishes. This edition contains no controlled tests. A vendor-published report may not be an automatic product export, and a branded example does not show that the brand is a customer. Beta, announced, historical and illustrative material is labeled where it matters.
What the scores cannot tell you
A score reports fit for the job described above. It does not measure accuracy, test report quality, or show that one product beats another on the same campaign. Small differences are editorial judgments, not statistically meaningful gaps. Sentiment, intent, engagement, estimated exposure and recorded sales are different measures.
An inaccessible page, a missing search result or an absent feature description creates a gap in our evidence. It does not prove a product lacks the feature. Dates identify when evidence was reviewed; packages and access conditions change.
Changes
- November 9, 2026. Corrected the publisher disclaimer and methodology to identify Social Intelligence Review as an independent auditor, unaffiliated with any of the platforms assessed in the guide, and updated the corrections email.
- November 9, 2026. Corrected the description of who runs this guide.
- November 9, 2026. Added 42 products across creator marketing, creator analytics, social listening and China data, for 61 in total.
- November 9, 2026. Replaced the original six criteria and weights with this campaign-assessment scorecard, which gives more weight to the work unrandom is built for. Removed hand-set points that had applied only to unrandom. Re-scored all 19 products on one formula. Added newly reviewed sources and outputs for CreatorIQ, Meltwater, Brandwatch, Sprinklr, Quid, KAWO, Datastory, Newrank and Qiangua.
- November 8, 2026. First edition published.
Corrections
Send corrections to info@socialintelligencereview.com, identifying the review URL, the statement in question and a supporting source. Corrections to a scored capability also update the score rationale, and material changes are recorded above.