How Andrex Took Affiliate Compliance Violations to Zero on TikTok Shop

The Client

Andrex is a UK household brand with dominant retail presence but no direct-to-consumer channel. TikTok Shop represented their first serious D2C play, and their first time putting brand claims in the hands of people they do not employ.

For a household CPG brand of that size, the creative upside was never the hard part. The hard part was that 87 creators would be making claims about hygiene, sustainability and product performance under the Andrex name, and under UK advertising law the brand carries the liability for every one of them. Andrex partnered with TikTok Shop agency SoSquared, powered by Vidsight's compliance and creative intelligence layer.

The Challenge

You cannot gate affiliate content

Affiliate creators are not an agency. They write their own scripts, post on their own schedule, and do not submit for approval. Any compliance model that depends on pre-publication sign-off does not survive contact with an open affiliate program. The question is not how to stop non-compliant content being posted. It is how fast you can find it and pull the money off it.

The highest-risk claim territory in the category

Household paper goods sit on top of three of the ASA's most actively enforced areas: environmental claims, health and hygiene claims, and flushability. These are exactly the claims creators reach for instinctively, because they sell.

Paid amplification turns a creator's mistake into the brand's ad

An unsubstantiated claim in an organic affiliate video is a problem. The same claim running under GMV Max with brand budget behind it is the brand publishing that claim. The exposure window between publication and spend removal is the entire risk.

Zero baseline, zero playbook

100% of revenue through retail meant no existing creator relationships, no performance benchmarks, and no established rules for what creators could and could not say. Nothing existed to brief against.

The Vidsight Approach

The system runs autonomously across four steps.

1. Connect

Andrex authorised TikTok Shop access in one click. No agency or partner setup required, and no dependency on creators submitting anything.

2. Ingest

Every affiliate video was pulled automatically, paid and organic, whether or not the creator had ever engaged with a brief. Coverage was the whole program, not the compliant subset of it.

3. Score

Each video was checked frame by frame and against the full audio transcript, against Andrex's own do's and don'ts, and rated by severity:

Severe:

• Unsubstantiated environmental claims (plastic free, fully biodegradable, sustainable, eco)

• Health and hygiene claims (antibacterial, kills germs, safe for sensitive skin conditions, dermatologist approved)

• Flushability and drainage claims (won't block your pipes, safe for all plumbing)

• Denigration of named competitors

Light:

• Missing, buried or late commercial disclosure

• Price and promotion misrepresentation (cheapest in the UK, fabricated urgency, discount claims without a reference price)

• Brand guideline breaches (prohibited product contexts, off-brand humour, misuse of brand assets)

• Vague superiority claims without a stated basis

4. Act

Detection triggers enforcement without waiting for a human. The creator is warned with the exact offending phrase quoted back to them, paid spend is pulled off the video, and repeat offenders are removed from the program under a configurable three-strike system.

Severity is configurable against the strike count, so a severe violation could be set to carry more weight than a light one. Andrex ran severe violations at double weight, meaning two severe flags were enough to trigger removal.

Creators who did engage with briefs also had drafts scanned before posting, and 94% of published videos went through at least one pre-upload iteration. That improved creative quality. It was never the compliance control, because it only ever covered the creators who chose to use it.

What Was Caught

Across 189 published videos, Vidsight detected 10% light non-compliance and 3% severe non-compliance, 25 videos in total.

Paid spend was removed from 100% of non-compliant content within one hour of publication. No non-compliant Andrex video carried paid amplification for longer than that.

By Wave 4, no non-compliant content was posted at all.

Not every creator corrects

Most creators fixed their behaviour after the first warning. Some did not. Four creators reached three strikes and were removed from the program, all during Waves 1 and 2. Between them they accounted for a disproportionate share of severe violations, and their removal is a large part of why Wave 3 severe violations hit zero.

Removing a producing affiliate is a revenue decision as well as a compliance one, which is why the strike thresholds sit with the brand rather than with Vidsight. Andrex chose to enforce. The performance numbers below are what the program did after those creators were gone.

Violation mix

Environmental claims accounted for the majority of severe flags. Creators consistently reached for sustainability language as a differentiator, in the most heavily enforced claim category in the ASA's current guidance. Disclosure failures accounted for the majority of light flags, and almost all of those were placement rather than absence: the disclosure existed but appeared after the hook, past the point where most viewers had already formed a purchase intent.

Why the rate fell rather than the volume

Publishing volume held flat across all four waves. The decline came from two things: persistent offenders leaving the program, and every violation producing a specific quoted correction that fed straight back into the next wave's briefs. Creators stopped making the mistakes they had seen quoted back at them.

Compliance-Safe Creative Still Won

The constraint did not cost performance. The formats that stayed compliant were also the formats that converted, because they sold on situation rather than on claims.

Pantry restock POV outperformed standard product demos by 156%, at 4.2% conversion against a 1.6% category baseline. It makes no product performance claim at all.

Bulk value framing ("stock up and save") delivered 3.8% conversion and £24.60 AOV against 1.9% and £18.40 for premium quality messaging. Price and quantity framing is substantiable. Quality superiority framing usually is not.

Pattern-interrupt hooks lifted watch-through 127% over product-first openings, and moved disclosure into a compliant position as a side effect, since the hook no longer needed the first three seconds for the product name.

CTA timing mattered more than CTA wording. Mid-roll CTAs at 18 to 22 seconds beat end cards by 67% in sub-30s videos, and dual CTAs beat single by 89% in 30 to 60s videos.

Vidsight's competitor scan of 2,847 household CPG videos, including Cottonelle, Charmin and Who Gives A Crap, generated 31,200+ datapoints scored for both conversion potential and claim risk. It showed how routinely competitor affiliate content was making claims the ASA would not accept, and supplied labelled examples of every violation category before Wave 1 launched.

Results (4 Waves)

Campaign totals: $225K+ GMV, 20M+ views, 200K+ engagements, 189 published videos, 100% of non-compliant content defunded within one hour, zero non-compliant content posted in Wave 4.

Key Takeaways

You cannot approve your way to affiliate compliance. Creators post without asking. The only control that covers the whole program is automatic ingestion of every video, paid and organic, after it goes live.

Speed of defunding is the real metric. A violation you catch in an hour costs a brand almost nothing. The same violation running on paid budget for a week is the brand publishing the claim.

Enforcement has to have teeth. Warnings alone move most creators. A minority ignore them, and a three-strike system with actual removal is what stops them setting the tone for everyone else.

Environmental claims are the primary risk in household CPG. Creators reach for sustainability language unprompted because it differentiates, and it is the claim category regulators are watching hardest.

Quoting the specific violating phrase is what changes behaviour. Generic warnings produce repeat violations. Quoted corrections produce learning, which is why the violation rate hit zero by Wave 4.

Compliance and conversion are not a trade-off in this category. Situational formats outperform claim-led formats, and situational formats are the ones that stay clean.