Role
Advertising
Timeline
1 months, redesigned full existing app
Community Truth Tool
In 1 month, led the redesign of McCann's research tool into an enterprise product for Omnicom.
Problem
McCann's specializes in finding real insights in culture trends, their in-house research unit, McCann Truth Central, runs large-scale consumer studies (The Truth About Youth, The Truth About Privacy) and feeds insights into creative strategy. However, finding the truth is not always easy. Strategists relied on data scientists to interpret findings for them. But the best ideas come when people can explore the data themselves. Last year, McCann self built an internal research tool for strategists with some part-time developers and it started to gain interest company wide. Leadership wanted the whole Omnicom network to access this data. That's where my team came in: we had one month to audit the original tool, identify what wasn't working, and transform it into a product that strategists of all experience levels could actually use.
Process
Through 10+ user interviews to identify problem with existing app, we learned that the first step to build a community (aka advertising audience) where a lot of users gets stuck. Depending on their skill level, users either got lost in advanced settings or missed key functionalities entirely. The UX was misleading and lacked information hierarchy. We decided this was the part we should spend the most time iterating.
Solution
Community Building
We asked Claude Design to suggest interface approaches for making audience building selections easier and more discoverable. We iterated extensively as a team, and once the Claude prototype was in good shape, we brought it directly to users for testing. With real user feedback, we ensured humans remained the ultimate gatekeepers of the product's usability. The biggest benefit of working with AI was the rapid prototyping: we could iterate on different UX approaches in the morning and A/B test them with users in the afternoon. We were able to validate designs in 24 hours rather than weeks. Another huge benefit was that when we were ready for hand-off, the prototype was so close to production that the engineers could extract Claude's front-end code directly, cutting development time in half.
Landing Page
Added a landing page so it is easy for agency users to collaborate and reuse communities others have created. It's also creates space for existing users to access their previous generated outputs for sharing.
Comparing communities
We made it super simple to compare communities with baseline community already pre-selected for user based on their target market (ex. US general population), and they can add others to compare with a simple dropdown. Once selected the chip with colors acts as legend for the infographic below. It not only saves space in UI but also makes data fast to read.
We proved that a small team could ship a sophisticated data product in a month by rethinking our prototyping workflow. The AI prototyping process became a standard for the wider design team.
Result
What started as a single-agency tool became enterprise infrastructure. Data democratization meant strategists across Omnicom could build audience insights into strategy from day one—earlier in the planning process, with more confidence because they're working from facts, not secondhand interpretations.
"This is much easier to use! I love it!"
- Aaron K, McCann London
Reflection
Using Claude Design as a real-time testing partner changed how we work. My colleague San and I could debate a UX change, iterate it live with Claude AI, and watch users interact with the result within minutes. That speed compressed months of traditional iteration into days. We also explored how to assist users further in finding the nuggets across data sets. The tool is still chart-heavy, which works for data people but can overwhelm strategists new to consumer insights. The next iteration is to add AI guidance that highlights interesting patterns (over-indexes, under-indexes) without introducing bias. I am interested in how AI could work alongside the data vis, not replace it.This project deepened my conviction that the best design tools aren't about removing complex options — they're about making complexity easy for people to navigate and use .