By Sacha JibrilWhat we collect, and what we drop
Carouscale indexes TikTok Shop photo carousels that pass 5,000 views across 6 markets (US, UK, France, Germany, Spain, Italy), with their stats and the linked product, then recreates those winning formats with your product using AI, text baked into the image in your language. From €29 per month.
The pipeline described here is built and maintained by Antoine Marchi, co-founder and CTO. The editorial guides on this site are signed by Sacha Jibril, co-founder and CEO, who sells on TikTok Shop himself.
Carouscale collects public TikTok photo posts, carousels of at least 2 images, from niche searches and from a list of seller accounts the pipeline follows. Two hard filters run before anything reaches Trends:
- Shoppable only: the carousel has to be tied to a TikTok Shop product. A viral carousel with no linked product proves nothing about whether it can sell, so it is dropped.
- 5,000 views minimum: a hard bar, applied at collection and rechecked before publication. Under that threshold, engagement rates are noise.
Duplicates are merged by post ID, since TikTok will return the same post through several paths. We keep the most-viewed version and aggregate the regions the carousel showed up in. The recency filter uses the real publication date of the post, not the date we found it.
The score: purchase intent, not sales
TikTok does not expose revenue per post, so we rank carousels by purchase intent instead: engagement rates per view, weighted by how strong each signal is, then multiplied by a logarithmic reach factor. A save on a carousel with 500,000 views counts for more than a save on one with 5,000, but not 100 times more.
| Signal | Weight | Why |
|---|---|---|
| Saves | 50% | The strongest purchase-intent signal available: people save what they are considering buying. |
| Shares | 25% | Sending a product to someone else is an active recommendation. |
| Comments | 15% | Sizing, shipping, where's the link: a purchase gets prepared in the comments. |
| Likes | 10% | The weakest signal. People like plenty of things they will never buy. |
Written out: score = (0.5 x save rate + 0.25 x share rate + 0.15 x comment rate + 0.1 x like rate) x reach factor. Rates are computed per view, which is what lets a carousel with 8,000 views and one with 800,000 be compared on equal footing.
How we decide which market a carousel belongs to
Trends covers 6 markets: the United States, the United Kingdom, France, Germany, Spain and Italy. Classification leads with language. Text that is clearly French, in the caption or in the product title, files the carousel under France regardless of the region it was discovered in. Same logic for Spanish, Italian and German.
When language does not settle it, we check whether the product is available in the US catalog, then fall back on the region of discovery. Leading with language corrects a real bias: plenty of European products are also listed in the US, and a catalog-availability test on its own pushed far too many carousels into the US bucket.
This is also why a carousel found abroad stays usable at home. The format is what travels: replication rewrites the slide text in your language, whatever market the original came from.
The metrics we display, defined precisely
- Views
- The number of plays on the post (TikTok's play_count). This is the entry-bar metric: below 5,000 views, a carousel never enters Trends.
- Saves
- The number of times the post was saved (collect_count), meaning people who put the carousel aside for later.
- Save rate
- Saves divided by views. It is the first number we look at, because it measures purchase intent independently of how big the audience was.
- Shares
- The number of times the carousel was sent to someone else (share_count).
- Score
- The composite index described above: a weighted intent rate multiplied by a logarithmic reach factor. It ranks carousels against each other. It does not predict revenue.
Known limits, stated plainly
Every dataset has a shape, and its shape is what it cannot see. Here is ours. If one of these matters to the decision you are making, weigh our numbers accordingly, or ignore them.
- TikTok does not expose revenue per post. Our score measures purchase intent through saves and shares, not actual sales. It is an engagement ranking and we call it one.
- We only ever see winners. Nothing in the database tells you how many carousels used the same format and went nowhere, so a format that shows up often is a format that worked at least once, not a format that works reliably. This is survivorship bias and no amount of collection fixes it.
- The 5,000-view bar is a reach filter, not a sales filter. It keeps out posts that only reached a creator's own followers, and it also throws away carousels that converted well on a small audience. That trade-off is deliberate; it is still a trade-off.
- Coverage is uneven across the 6 markets. The US and France are the best served; Italy and the UK are thinner, so a niche can look empty in a small market when it is really just under-sampled. Read cross-market comparisons with that in mind.
- Every metric is a snapshot taken on the day of collection. A carousel keeps gathering views after it enters Trends, so the numbers are refreshed on each pipeline run rather than frozen at discovery. Two carousels collected weeks apart have not had the same time to accumulate.
- Market detection leads with the language of the text. That is reliable on clear cases, but a carousel with no caption and an English product title can land on a neighboring market. Niche and category labels are assigned automatically too, and automatic labels are wrong sometimes.
- The sample is not exhaustive. It covers the niche searches and the seller accounts the pipeline follows, not the whole of TikTok. A carousel that sells can slip through, and a niche nobody on our list posts in will look smaller than it is.
- We do not measure what happens after you replicate. We have no controlled comparison showing that rebuilding a proven format outperforms starting from scratch, and we will not claim one until we can run it. What we can show is which formats already earned attention on a real audience.
One consequence worth spelling out, since it decides whether you should buy from us at all: this is not a market research database. We do not estimate revenue, rank shops, track creators or keep long catalog histories. Tools like Kalodata, FastMoss, EchoTik and Shoplus do, and on that ground they are ahead of us. Our data is collected to answer one question, which post format earned attention on this product, and it is poor at answering anything else.
Using our numbers in your own work
The public aggregates we publish, the per-niche and per-market figures on the TikTok Shop trends barometer, are released under the CC BY 4.0 license. Reuse them in an article, a deck or a video, with credit to Carouscale and a link back to the page you took them from.
What is not covered by that license is the carousel-by-carousel detail: individual posts, their stats and the linked products sit behind a subscription, and the underlying content belongs to the creators who published it.
How we rate the tool reviews
This site publishes reviews of Kalodata, EchoTik and Shoplus, plus head-to-heads such as FastMoss vs Kalodata. The ratings out of 5 rest on four criteria, judged from the point of view of someone who publishes carousels to sell:
A rating is not an average of four numbers. Every review shows the arithmetic in the open, criterion by criterion, and says why the score is not half a point higher or lower. Where we have landed so far: 3.5 out of 5 for Kalodata, 3.5 for EchoTik, 3 for Shoplus. If you read one of those and the reasoning does not hold up, tell us, and we will either defend it or change it in public.
Prices and terms, verified on a stated date
Pricing is read off the vendor's own page, and the review displays the date we read it. When a vendor does not publish a public grid, which was the case for Kalodata on July 25, 2026, the ranges come from third-party sources and are labeled reported. They are never presented as verified.
Market coverage
What the tool actually shows of each market it claims, and how deep it goes. A ranking that only reflects US volume is a different product from one that holds up in Germany or Italy.
Usefulness to someone who publishes
What is left to do once you have the data. Does the tool stop at a table of numbers, or does it help produce the content that does the selling? In practice every analytics tool we have reviewed scores the same here, which is to say nothing at all, so this criterion separates none of them and never carries a rating on its own.
Vendor transparency
Public price grid or a form to fill in, trial and renewal terms stated plainly or buried, refund policy findable or not. Also what we could not test: we do not buy every plan of every tool, so quotas and limits are what the vendor states, and we say so on the page.
Conflict of interest, said out loud: Carouscale publishes these reviews and sells a product that competes with part of what they cover. You should read them with that in mind, and we build them so you can. Every one shows the date we checked the pricing, separates verified prices from reported ones, and states in writing where the compared tool is stronger than we are, typically global data depth, historical archives, sales estimates, creator discovery and live analytics. Every one also says plainly who should buy the other tool instead of ours. What we do not do is buy every plan of every competitor: quotas and limits are reported as the vendor states them, never as something we stress-tested, and no vendor pays us for a placement or a score.
Frequently asked questions
- Where do the carousels in Carouscale Trends come from?
- From automated collection of public TikTok photo posts, the carousels, restricted to shoppable posts, meaning posts tied to a TikTok Shop product. A carousel with no linked product never enters Trends, however viral it is.
- Why a 5,000-view minimum?
- Below that, engagement is not meaningful: a handful of saves on a post with 300 views produces a spectacular rate that reproduces for nobody. The bar is applied at collection and rechecked before publication, so no carousel under 5,000 views appears in Trends.
- Does the Carouscale score measure sales?
- No, and we do not claim it does. TikTok does not expose revenue per post. The score measures purchase intent by weighting engagement rates per view (saves 50%, shares 25%, comments 15%, likes 10%) and multiplying by a logarithmic reach factor to separate carousels that share the same rates.
- How do you rate tools like Kalodata, FastMoss, EchoTik and Shoplus?
- On four criteria: prices and terms verified on a date we display, market coverage, usefulness to someone who publishes carousels, and vendor transparency. Each review shows its own arithmetic, criterion by criterion, next to the score. As of July 2026 we have published 3.5 out of 5 for Kalodata, 3.5 for EchoTik and 3 for Shoplus. Carouscale publishes those reviews and sells a competing product. That conflict of interest is stated on every one of them, any price we cannot verify is labeled reported, and every review names at least one thing the tool does better than we do.
- How do I report an error in a number or a price?
- Through the contact page or at contact@carouscale.com. A price that changed on a vendor's page, a carousel filed under the wrong market: we fix it and update the checked-on date of the page concerned.
Anything else you want to check, or a number that looks wrong? Get in touch.
Keep reading
- Kalodata review & pricingWhat it costs, what it doesn't do.
- FastMoss vs KalodataThe honest comparison, plus a third option.
- TikTok Shop winning productsFind what to sell, with data.
Guides by niche