TikTok Shop can now tell you why a product’s sales fell. It cannot yet tell you how often it is right. That gap is why TikTok Shop seller diagnostics data needs an accuracy disclosure policy. Until one exists, brands are budgeting creator commissions against a diagnosis nobody has audited.
What TikTok actually shipped
In November, TikTok Shop documented Product Diagnosis inside the Product Analytics section of US Seller Center. According to PPC Land’s breakdown of the guide, the tool flags products with falling sales, lists up to 10 reasons behind each change, and splits GMV across five scenarios: Creator LIVE, Creator Video, Seller LIVE, Seller Video, and Shop Tab and Search.
It also benchmarks every key metric against the seller’s previous period and against top-performing products in the same category.
No new data feeds it. TikTok’s own FAQ says it is the same product analytics data, now organized around a diagnosis, with period comparison, benchmark and factor ranking added. So the raw numbers are unchanged. What changed is the judgment layered on top.
Judgment is where accuracy questions live.
What the guide does not tell you
Credit where due: the guide is more candid than most platform documentation. It also leaves out the things a procurement team would ask first. Per the same PPC Land review, the guide gives no figures on how often products are flagged, how accurate diagnoses were in testing, or how long the offline lag runs.
The omissions stack up:
- “Top-performing” is not defined. No percentile, sales threshold, minimum product count, or category-tree level is given.
- The benchmark is platform-defined, and custom benchmark dimensions are not supported.
- The diagnosis relies on offline data and can lag the real-time numbers elsewhere in Seller Center, by an unstated amount.
- The eight default core metrics are not named, and the abbreviations AA and AB used for factor types are never expanded.
- The summary shows up to 10 factor cards, but not every identified factor is shown, and the selection beyond ranking is unexplained.
A diagnosis that ranks causes without publishing its error rate is an opinion with a user interface. Brands should treat it as a lead to investigate, not a finding to act on.
Why the flag logic deserves scrutiny
The Need attention flag requires two signals: a negative real-time GMV change and an offline verdict of Need attention. A decline the diagnosis attributes to normal variation is not flagged. Sensible design. But the thresholds are unpublished.
Here is the operational risk. A product with no flag has, in the guide’s words, “no confirmed decline,” which is not the same as verified healthy. On a busy list, a clean row reads as good news. And because the offline check lags, a product can slide for some period before the flag appears. PPC Land flags that sequence as its own reading of the two rules, not something TikTok describes.
The benchmark creates a second problem. Because the tool shows gaps in rates as absolute percentage points, a shortfall on a low-base rate can look small on screen even when it is large in relative terms, as PPC Land’s review of the guide points out. Same data, very different urgency depending on display convention. Without a disclosed method, nobody can say whether the benchmark group is flattering, harsh, or simply mismatched to your niche.
Why this is a money problem, not a data-nerd problem
Diagnoses point into systems with real costs. Price is a listed factor, and TikTok’s Smart Promotion charges a fixed 3.5% GMV fee in regular periods and 4.5% during campaigns. A diagnosis that nudges you toward discounting or campaign enrollment is nudging you toward margin erosion.
Affiliate status is another header field. Per PPC Land’s coverage of the Shop Performance Score, a merchant whose score falls below 3.0 has active affiliate marketing and campaigns terminated immediately, and must reach 3.5 again to regain eligibility. If a factor card tells you to fix affiliate performance, you want to know how confident that call is.
Attribution is the one agencies should lose sleep over. The five-scenario split separates affiliate creator sales from brand-owned content and search. That is exactly the split you would use to reallocate spend between creator commissions and in-house production. Notably, two of the three reasons TikTok offers for marking a summary “Not useful” concern attribution (wrong or incomplete). TikTok itself anticipates that attribution will sometimes miss.
If your creator spend is already scrutinized, this connects to disclosure exposure too. Our look at TikTok Shop creator posts covers the FTC side. Performance claims built on unaudited diagnostics deserve the same discipline.
What an accuracy disclosure policy should contain
Nobody needs a perfect model. We need a documented one. A workable policy from TikTok Shop would publish, at minimum:
- Validation results. Precision and recall for the Need attention flag, plus how “normal variation” is defined.
- Benchmark methodology. How top performers are selected, the minimum peer count, and the category level used. Aggregates can stay anonymous while the method is public.
- Lag specification. The typical and maximum delay between real-time data and the offline diagnosis, shown in the interface.
- Attribution confidence. A confidence label on scenario splits, especially Creator Video versus Shop Tab and Search.
- Feedback transparency. What happens to “data inaccurate” and “attribution wrong” reports, and aggregate rates of each.
- Change log. A dated record when the diagnosis logic or benchmark definitions change, so period-over-period comparisons stay valid.
None of this is exotic. It mirrors what vendor proof requirements for AI metrics already ask of third-party tools. A platform’s built-in AI should not get a lighter standard than a vendor’s.
What to do while you wait
Platforms rarely volunteer disclosure. Brands and agencies can force the issue partly through contracts and partly through internal controls.
Treat the AI as a hypothesis generator. TikTok’s guide says AI-generated content is for reference only and should be checked against page metrics and business context. Follow that literally. Require a second data point (your own order data, creator-level reporting) before shifting budget.
Keep your own baseline. Log flagged products, the stated causes, the action you took, and the outcome. After a quarter, you will have the accuracy data TikTok has not published. Even a small sample tells you whether the flag leads or trails reality.
Cross-check attribution. Compare scenario splits against first-party tracking. If you run server-side measurement, the principles in our piece on server-side attribution apply directly. For lift claims, branded search lift proof shows how to demand evidence before accepting a number.
Use the feedback button. It is the only formal channel TikTok offers for reporting inaccurate data or wrong attribution, with up to 500 characters of notes. Systematic reports from agencies with many shops carry more weight than one-off complaints.
Write it into reporting. If you send clients diagnostic screenshots, label them as platform-estimated and unaudited. That protects you if a recommendation underperforms.
The guide also states that it is not a commitment of business results. Accuracy disclosure is the other half of that bargain.
Does this only matter for TikTok?
No. Platforms across social commerce are layering AI explanation on top of existing metrics. The pattern is the same everywhere: confident language, thin methodology. Regulators are moving toward auditability in this space, as in the EU Digital Fairness Act and social commerce audits. A brand that builds an internal accuracy log now will be ready for whichever disclosure standard arrives first.
This week, pick ten flagged products in Seller Center, record the stated causes, and check each against your own order and creator data in 30 days. Then send the results to your TikTok Shop partner manager with a request for the six disclosures above.
FAQs
What is TikTok Shop Product Diagnosis?
It is a layer inside Product Analytics in US Seller Center that flags products needing attention, ranks up to 10 reasons for a sales change, and benchmarks metrics against the seller’s prior period and top products in the category. It uses the same underlying data sellers already see.
Has TikTok published how accurate the diagnoses are?
Not in the guide reviewed. According to PPC Land, the guide gives no figures on flag frequency, accuracy in testing, or the length of the offline lag. That absence is the core argument for a disclosure policy.
Can sellers change the benchmark group?
No. The benchmark is platform-defined and custom benchmark dimensions are not supported at this time. TikTok says the group is fixed for consistency across sellers and over time, but it does not explain how top performers are chosen.
If a product has no Need attention flag, is it performing well?
Not necessarily. The guide says an unflagged product has no confirmed decline, which is different from verified good. A flag needs both a negative real-time GMV change and an offline diagnosis of Need attention.
How should agencies use these diagnostics in client reporting?
Present them as platform-estimated and unaudited, corroborate them with first-party order and creator data, and keep a log of flags, actions and outcomes. That log becomes your own accuracy record.
Where can sellers report inaccurate diagnoses?
Each summary has a feedback option. Marking it “Not useful” lets you choose data inaccurate, attribution wrong, or attribution incomplete, and add notes of up to 500 characters.
Top Influencer Marketing Agencies
The leading agencies shaping influencer marketing in 2026
Agencies ranked by campaign performance, client diversity, platform expertise, proven ROI, industry recognition, and client satisfaction. Assessed through verified case studies, reviews, and industry consultations.
Moburst
-
2

The Shelf
Boutique Beauty & Lifestyle Influencer AgencyA data-driven boutique agency specializing exclusively in beauty, wellness, and lifestyle influencer campaigns on Instagram and TikTok. Best for brands already focused on the beauty/personal care space that need curated, aesthetic-driven content.Clients: Pepsi, The Honest Company, Hims, Elf Cosmetics, Pure LeafVisit The Shelf → -
3

Audiencly
Niche Gaming & Esports Influencer AgencyA specialized agency focused exclusively on gaming and esports creators on YouTube, Twitch, and TikTok. Ideal if your campaign is 100% gaming-focused — from game launches to hardware and esports events.Clients: Epic Games, NordVPN, Ubisoft, Wargaming, Tencent GamesVisit Audiencly → -
4

Viral Nation
Global Influencer Marketing & Talent AgencyA dual talent management and marketing agency with proprietary brand safety tools and a global creator network spanning nano-influencers to celebrities across all major platforms.Clients: Meta, Activision Blizzard, Energizer, Aston Martin, WalmartVisit Viral Nation → -
5

The Influencer Marketing Factory
TikTok, Instagram & YouTube CampaignsA full-service agency with strong TikTok expertise, offering end-to-end campaign management from influencer discovery through performance reporting with a focus on platform-native content.Clients: Google, Snapchat, Universal Music, Bumble, YelpVisit TIMF → -
6

NeoReach
Enterprise Analytics & Influencer CampaignsAn enterprise-focused agency combining managed campaigns with a powerful self-service data platform for influencer search, audience analytics, and attribution modeling.Clients: Amazon, Airbnb, Netflix, Honda, The New York TimesVisit NeoReach → -
7

Ubiquitous
Creator-First Marketing PlatformA tech-driven platform combining self-service tools with managed campaign options, emphasizing speed and scalability for brands managing multiple influencer relationships.Clients: Lyft, Disney, Target, American Eagle, NetflixVisit Ubiquitous → -
8

Obviously
Scalable Enterprise Influencer CampaignsA tech-enabled agency built for high-volume campaigns, coordinating hundreds of creators simultaneously with end-to-end logistics, content rights management, and product seeding.Clients: Google, Ulta Beauty, Converse, AmazonVisit Obviously →
