Advertiser Investigation & Lookup

Unified two fragmented investigation tools into one system for TikTok's Integrity team, lifting monthly cases per user 33% and cutting time to decision 40%.

Role

Product design lead

Team

1 PM, 2 Software Engineers

Timeline

~6 weeks

Focus

Research, Information Architecture, UI Design, Prototyping

Impact

-40%

Time to decision

+33%

Monthly cases per user

4.3

CSAT, up from 3.6

-38%

Reversal rate

The problem

Investigators had to switch between two tools to complete a single investigation: Jedi for breadth and Actor 360 for depth, and the tools didn't always agree with each other. Every case meant re-verifying data mid-investigation, and different teams sometimes reached different conclusions from the same advertiser.

The reframe: it's one motion, not three tasks

An investigation isn't three separate steps. It's one motion, narrow from a network to a decision where breadth tells you where to look, depth tells you what's true, and action closes it. The tooling had split that motion in half; the fix wasn't a better version of either tool; it was one tool shaped like the motion itself.

What research told us

Investigators follow a predictable path

View the actor → 80% go straight to Customer Complaint Rate → 60% go to Actor Association → take action. The overview page needed to be built around this path, not an alphabetical tab list.

Only 7 of 15 tabs saw real use

Most of the tool was noise investigators had already learned to ignore.

Data disagreed across tools

Actor 360 and Jedi could show the same advertiser differently, so investigators re-verified before acting on it — and without a shared source of truth, different teams sometimes reached different conclusions from the same case.

One overview, not fifteen tab

Eight of fifteen tabs were cut or merged, using one rule applied evenly: redundant with something else, or fewer than 10 daily active users. What remained was rebuilt around the order investigators actually use — overview first, in-app, collapsible, no tab-switching required.

Trust needs an audit trail

The deeper problem wasn't the number of tabs — it was that nobody fully trusted the data in front of them. Putting the qualification breakdown, full audit history, and the ability to overturn a decision in one place, not two, is what let investigators act with confidence instead of re-checking everywhere.

Search shouldn't stop at one ID

Investigators work across strategy teams, often chasing several advertiser IDs from different leads at once. Search accepts a single ID or many, so an investigator can pick up exactly where a teammate's list left off instead of running searches one at a time.

A mode built for exactly one task

Not every moment in an investigation needs the full interface. A full-screen focus mode strips away navigation and secondary panels, leaving only the attributes and extraction an investigator needs to finish the task in front of them.

Extraction without leaving the case

Investigations often need data spanning multiple layers, actor-level and feature-level detail that used to mean jumping to a separate tool. Extraction pulls that detail directly into the investigation, so the case stays whole instead of scattering across tabs and tools.

One enforcement action, not twenty

Investigations often need data spanning multiple layers, actor-level and feature-level detail that used to mean jumping to a separate tool. Extraction pulls that detail directly into the investigation, so the case stays whole instead of scattering across tabs and tools.

Outcome

-66%

Strike volume per advertiser

75%

Completion rate, ~8 min avg

69%

Quiz pass rate, first attempt

4.7

CSAT, +14% vs. platform

-7%

Overall ad rejection rate

~3%

Increase in active ad revenue

The number that mattered most wasn’t completion - it was the drop in strike volume. That’s the loop actually closing: design → learning → behavior change → platform integrity.

Reflection

Fixing the tool meant fixing what investigators could trust: one source of truth, not two tools quietly disagreeing.

Learn more

There's so much more behind the scene! Reach out to see additional design explorations.

©️ 2026 Shibi Siddharth Karuppusamy.