Founder / Product Lead · 2024–Present

I turned a TikTok creator pain point into a product used across 100+ countries.

NoViolation scans TikTok Shop videos for policy risk before creators post.

7,000+

users

100+

countries

20,000+

scans

99%

organic

The problem was hiding in plain sight

TikTok Shop creators were constantly asking each other "Can I post this?" Violations could cost them videos, products, or revenue, but TikTok rarely made it clear what actually caused them.

I read through hundreds of creator discussions and found the same pattern: people were posting and praying.

"Any way to test a video to see if it violates community guidelines before uploading?"

"Got my first tiktok shop violation. What does this mean?"

"i don’t think there is any way of telling if ur vids gonna get hit"

"No, unfortunately I didn’t find a tool."

"It is almost impossible to figure out what is ok and what will cause a violation strike."

The first MVP was too big.

I scoped the full product: accounts, payments, video analysis, transcription, affiliates, and an admin panel.

The estimate: $2K–$10K and 2–3 months to build.

PRD — Overview, problem, goal, user, flow

Before the estimate stopped me, I'd already designed the full product in high fidelity — desktop and mobile.

Desktop — high fidelity MVP screens

Mobile — high fidelity MVP screens

The landing page

The landing page is as far as I got. I realized this MVP wasn't feasible to build, and I needed to pivot.

I didn't need the full product yet. I needed to know if people would actually use it.

So I cut everything down to one test.

I built the MVP in a week

I used Wix, Zapier, and OpenAI to fake the first product. Creators uploaded a video, the workflow analyzed the transcript, and a result landed in their inbox within two minutes.

No accounts. No dashboard. No database.

The MVP — a single upload form and a button.

The Zapier workflow — Wix form → transcribe → delay → JavaScript → split into paths.

Path A / Path B — OpenAI conversation, then Gmail back to the creator.

The results email — violation likelihood, status, and an upgrade nudge.

500 users in the first two months. 100% organic.

Then the MVP broke

Transcript analysis worked until the violation was visual. NoViolation could call a video safe while completely missing a prohibited product, claim, or text shown on screen.

That wasn't a prompt problem. The model couldn't analyze information it couldn't see.

Rebuilding NoViolation with a CTO over Google Meet — Bringing on a CTO and rebuilding the product to actually watch and listen to the video.

I brought on a CTO and rebuilt the product to actually watch and listen to the video.

The real design problem was trust

A risk score wasn't enough. Creators needed to understand what was wrong, why it mattered, and what to change.

I redesigned the results around plain language explanations, severity, and clear next steps instead of making users interpret an AI score.

The checker — script and video scans from one place.

Interactive regulated-goods check — tap the risky phrase.

Live analysis — "Listening to audio" while the scan runs.

Video check results — status, probability, and confidence.

Plain-language verdict — audio and visual analysis, no jargon.

Once people paid, I called them

I interviewed five paying users to understand when NoViolation actually fit into their workflow.

Those conversations led to three bets: scan packs, script checking, and performance checking.

Only one worked.

User research plan — objective, participants, interview guide (p.1)

Interview guide — violation experience, discovery, usage (p.2)

Interview guide — opportunity, closing, principles, output (p.3)

Scan packs doubled revenue

Users didn't scan every day. They scanned in bursts, usually around launches or important posts.

Monthly subscriptions didn't match that behavior, so I introduced scan packs.

Scan pack revenue doubled within 30 days and stayed there.

Scan packs surfaced in-product — remaining count and a buy-more nudge.

The other features taught me something bigger

Script checking was barely used. Performance checking was barely used. The Appeal Assistant I'd already built was barely used.

The pattern became obvious:

People used NoViolation to scan a finished video before posting it.

They didn't want NoViolation to do everything around that job.

That changed the roadmap

Instead of adding more features, the focus became making the core job better.

Faster scans. Clearer verdicts. More trustworthy results. Easier fixes.

Everything else is secondary.

Results

017,000+ users across 100+ countries
0220,000+ scans run
0399% organic traffic, near-zero paid acquisition
04Scan pack revenue doubled in 30 days, and held
05Built and run by a two-person team

What I learned

Test features like products.

I validated NoViolation cheaply, but later overbuilt features before validating them.

Behavior beats interviews.

Users genuinely cared about scripts and performance. That didn't mean they wanted NoViolation to solve them.

Don't confuse more product with better product.

NoViolation has one core job: scan my video before I post it. The next version shouldn't do more, it should do that one thing incredibly well.