Product

A campaign report with recommendations, in about 30 minutes.

Add posts from a campaign on Xiaohongshu, Douyin, TikTok, Instagram, YouTube or X. unrandom analyzes the content and audience response, then compares performance with available benchmarks. Generate the report in English or Chinese.

Intelligence report

Illustrative campaign · not client data
Autumn cleanser launchSkincare · Cleansing
Campaign Verdict

The double-cleanse routine performed well in this launch. Creators who filmed it from start to finish beat their own baselines. Comments praised the cleanser's gentleness but repeatedly questioned whether it left skin tight after rinsing.

Performance71Strong
Audience+36Positive
Posts54
Creators33
Comments analyzed9,870
Key Findings
  • +Full-routine videos beat their creators' own baselines
  • −Comments questioning tightness after rinsing attracted the most likes among critical comments.
Recommendations

Use full-routine videos in the next campaign and address tightness after rinsing on camera.

How brands and agencies use unrandom

From links to a finished report, while there's still budget to move.

Add campaign posts by link or from your library. You can also set up Tracking to collect posts on a schedule. unrandom analyzes the campaign and produces a report in English or Chinese, with findings for your team to review.

The campaign workflow
  1. 01Prepare

    Search the library for campaigns in your category, including competitors' past work.

  2. 02Track

    Add campaign posts by link or from your library, or set up Tracking to pull them in on schedule.

  3. 03Understand

    Review the content analysis and audience feedback alongside performance benchmarks and signs of unusual engagement.

  4. 04Deliver

    Generate a report with findings and recommended actions in English or Chinese.

app.unrandom.co/dashboard/campaigns/summer-spfIllustrative
Summer SPF Launch
PerformanceCommentsPostsLabelsAudit

Intelligence report

Campaign Verdict

Audiences responded positively to the claim that the sunscreen leaves no white cast. Mid-tier creators generated the saves, and comments praised the texture.

Performance74Strong
Audience+41Positive
Posts4
Creators4
Comments analyzed3,412

From links to a finished report.

Research the category before your next campaign or pitch.

Search more than 30,000 campaign teardowns by brand, category, platform or tactic. Examine how a campaign used creators and how audiences responded before drawing lessons for your own brief.

app.unrandom.co/dashboard/archivesIllustrative

Archive whitepaper

Beauty · Sun care

Summer sunscreen mega-launch

Reach48M
Posts213
Creators87
The takeaway

Outdoor test videos performed better than studio videos. Mid-tier creators accounted for most of the reach.

Engagement rate

6.8%

ObjectiveAwareness to trial
DifferentiationProduct demonstrations outdoors
FormatVideo-led
OccasionSummer peak
Price tierMass premium
DistinctiveOutdoor stress test (11%), ingredient deep-dive (7%)
StandardBefore/after, get-ready-with-me
Watch-outDiscount depth crept up in week 3
01

Playbook

One claim, proven outdoors: no white cast, even after six hours in the sun.

Ideas to test in your next campaign
  1. 1Show the product claim in a demonstration.
  2. 2Build the creator roster around mid-tier accounts.
  3. 3Repeat the same test across creators
02

Top posts

412K reads

286K reads

198K reads

154K reads

03

Patterns in the top-performing posts

Opening hook

“I wore it on a six-hour hike so you don't have to.”

Shared by the top posts
  • Proof shot inside the first three seconds
  • Creator's own face, outdoors, natural light
  • Pinned comment carries the full routine
04

Measured from posts

Compare the campaign metrics with their benchmarks.

Median reads86.4KTop 9%
Engagement rate6.8%Top 12%
CPE¥1.8Top 24%
CPM¥42Top 31%
05

Casting

Creator mix

Mostly mid-tier creators, with no celebrity lead

Head
Mid-tierDrove results
KOC / long-tail
06

The evidence: distinctive vs typical

Compare campaign choices with patterns in the category. Percentages show how often each choice appears. Common choices are summarized below.

Messaging themesDistinctive

Ingredient deep-dive 7% · Outdoor stress test 11%

Typical Before/after 62%

Content formatsDistinctive

Long-form field test 9%

Typical Get-ready-with-me 57%

Promo mechanicsConventional

Typical Limited-time bundle 71% · Free-gift sampling 64%

Trend hooksDistinctive

Skinimalism 13%

Typical Ingredient literacy 58%

07

Conversion signals

Demand signals

“Link please, going to Sanya next week”

Open questions

“Does it pill under foundation?”

Primary CTA

Tap the product card

08

Audience Voice

Positive, with price pushback

+

“Finally one that disappears on darker skin”

−

“Wish the 50ml wasn't priced like a serum”

No advertising-law risks detected.

Read what worked before you spend.

Campaign posts, watched, heard and read.

unrandom transcribes speech and extracts text shown on screen. It analyzes comments by topic and sentiment, including signals of interest in buying, and checks for unusual engagement. Where enough of a creator's posts are available, it compares their campaign performance with their usual results. Historical benchmarks provide context for the campaign's performance where coverage is available.

app.unrandom.co/dashboard/campaigns/sunstick-summer/commentsIllustrative
Sunscreen stick summer push
Posts48
Comments analyzed12,438
Net Sentiment+38
unrandom Score86Exceptional

The campaign score, shown in context.

Why unrandom

Frontier AI, built on real campaign evidence.

Your team already uses AI. unrandom adds the campaign data and historical comparisons needed to assess performance.

A conductor monkey leads rows of monkeys, each with one job. A chat window is one instrument. This is the orchestra.
Read what the film shows

What the film shows

Every campaign, reply and figure in this film is illustrative, apart from the 3 million posts, the count of posts unrandom has measured (September 2026).

  1. Your team already has AI. So why unrandom?
  2. A plain chat window is asked “How did our spring serum launch do?” It answers: “Your launch performed strongly: high engagement, positive sentiment and good creator fit. Next time, consider working with more micro-influencers.” It answers in seconds. It sounds right. It never saw a single post.
  3. Asked again, it answers: “Results were mixed: engagement slowed in week two, and several comments questioned the price.” Ask it twice. Two answers. Which one goes to the budget review?
  4. The monkey tries to give the chat window the whole campaign. Three screenshots fit; the rest piles up and spills over. The screenshots show only part of the campaign. A few screenshots. Some text. 62 videos, 14,206 comments, 41 creators’ histories.
  5. Enough monkeys. The monkey multiplies, and each copy takes one job: scenes, speech, on-screen text, brand, comments, translation, baseline, reach. Each with one job. The campaign’s posts ride past them and collect a mark at each station; one monkey drops a card, and another catches it and sends it through again. The view pulls back to rows of lines running at once under one conductor. A chat window is one instrument. This is the orchestra.
  6. Is 4.6% engagement good? Compared with relevant benchmarks. 3 million posts measured by unrandom (September 2026). The campaign lands at 58, middle of the pack (unrandom Score, skincare launches).
  7. The intelligence report, Spring serum launch, flight 1: 990 people asked where to buy, and nobody answered; ¥1.2M went to two accounts below their own baseline; one 48-second demo drove 38% of purchase questions. Ask why. It shows you the posts, the comments, the figures. Ask again. Same evidence. Kept for the next review.
  8. We use frontier AI too. The same chat window, given the campaign analysis, writes: “Lead flight 2 with the 48-second demo, answer ‘where to buy’ in every post, and move ¥1.0M to mid-tier skincare specialists.” Using the campaign analysis, it drafts the recommendation.
  9. Frontier AI, built on real campaign evidence. unrandom. The social intelligence platform. unrandom.co

  1. The hard part is the data

    unrandom brings campaign metrics together with transcripts, on-screen text and comments. As posts are added and refreshed, the analysis can draw on the updated campaign data.

  2. Scores need real benchmarks

    A high engagement count needs a relevant comparison. unrandom uses historical benchmarks to show how a post performed against comparable content on its platform. Benchmark coverage varies by platform.

  3. Creator performance needs context

    A creator's own posts provide another comparison. Where enough are available, unrandom compares the campaign post with the creator's own posts from around the same time, to show whether it performed above or below their usual results.

  4. Compare performance in the context of each platform

    Platforms count attention differently. unrandom uses relevant platform and market benchmarks where available, then presents the results in a consistent report. This lets regional teams review the evidence without treating unlike metrics as interchangeable.

  5. Recommendations must survive questions

    When a client or colleague questions a finding, review the supporting posts and comments with them. unrandom keeps the evidence behind important findings available for inspection.

  6. Campaign history you keep

    Return to a campaign's analysis when planning the next quarter or reviewing a budget. The earlier findings remain available alongside the campaign's creators and comments.

5 years
supporting Amorepacific China
30,000+
campaign teardowns in the library
3 million
posts measured by unrandom

AI where it is strongest.

unrandom uses AI to interpret campaign content and draft the report. The analysis draws on structured campaign data and historical benchmarks. Those inputs let your team examine the evidence behind the findings.

See it work on one of your own campaigns.

In the demo, we work through the campaign with you, from adding posts to reviewing the report.

Request a demo