Web3 Marketing Dashboard: 0 to 1

The logic behind our ad targeting is complicated

Before this, the team built audience packages by hand, analyzing wallet behavior and putting together a list for each client individually, then handing it off for them to run ads. My job was to turn that service into something clients could do themselves, so the company could actually scale like a SaaS product, not an agency.

Build the first prototype

First, I used Material Design UI patterns to quickly piece together a first version of the prototype to test with users.

I broke the whole journey down into 3 simple steps:

Designing the cohort selection UI

The hardest part was the logic behind how we found audiences, helping the marketers understand this and guiding them through the process of setting their cohorts.

Marketers stacked filter conditions one after another, and only found out at the end if the audience worked. Too narrow, too broad, either way, they'd already built the whole thing before finding out it didn't.

My first instinct was to show live wallet data on every change. But that was slow, and it cost us money in API calls.

So I went back to a few core decisions:

Pre-selected category pools, not a blank slate. Marketers started from pools scoped to the type of project they were running, GameFi, NFT, DeFi, each already narrowed to what actually mattered, instead of building a cohort from zero.

A slideout, not a full-page flow. People could keep an eye on the estimate while they worked. As they added filters, the number updated in real time, pulled from our own data, not a fresh API call every time.

A way back, not a restart. If someone still hit a dead end at the last step, they could open the slideout again, jump back, and watch the estimate update live from there.

A percentage, not an exact number. An exact count implies precision we didn't have, the data was cached, not live. "1,284 wallets" sounds like right now. It might've been an hour old. A percentage told users what they actually needed to know: too narrow, or too broad.

Design trade-off

My engineer had limited time and couldn't get deep into UI details, so the design system I'd built in Figma didn't always make it into the final build cleanly.

That's part of why I now build my own React components with AI, so nothing gets lost between what I design and what actually ships.

See what I've built with AI →

Feedback & Impact

Clients later described the experience as "surprisingly easy," even though the underlying workflow, mapping wallet behavior to real identities, was highly complex.