Three weeks. One prototype. A $1M contract.

The situation

Conviva's analytics were built by data scientists, for data scientists, in the vocabulary of buffer ratios and bitrate. The platform reads more than five trillion events a day, and for years the people reading the output spoke the same language as the data.

Then the platform expanded from streaming into new verticals: retail, ecommerce, travel. The new buyers are product managers who carry revenue responsibility, have thirty minutes between meetings, and have no tolerance for cognitive overload. They don't need more data or more filters. They need answers. The legacy design, repurposed, fell flat.

A Fortune 500 retail account brought it to a head. The technology passed every test. The experience was the dealbreaker. I'd been making this argument to the CEO and CPO for months: the product wasn't failing on capability, it was failing on who it was built for. They asked me to build the reimagining, and gave me three weeks.

The reframe

I read the problem through Tesler's Law of the Conservation of Complexity. Every application has an irreducible amount of complexity, and the only real question is who has to deal with it. The legacy interface put that complexity on the user, because it was built by people who spoke the language of the data for people who spoke the same language. The new buyer doesn't, and shouldn't have to. So my answer, across every decision, was the same: the complexity of analyzing five trillion events a day belongs to the system, not the person.

The work

The premise was insight-first. The system proactively surfaces what matters, prioritized by revenue impact, in plain language. Intelligence is the starting point, not the reward for hours of analysis.

Navigation followed intent, not tool categories: see what needs attention, understand why, and only then dig into the data yourself. Progressive disclosure kept the complexity where it belonged. Answers first, evidence on demand, raw data only if you want it. The north star for the overview was a person seeing a good day or a bad day in ten seconds.

The structural artifact was a three-tier investigation model. Tier 1 is proactive intelligence: Monitor, answering “what needs my attention?” Tier 2 is assisted investigation: Investigate, answering “why is this happening?”, conversational, run with the product's AI assistant. Tier 3 is manual exploration: Analyze, full control for power users, for validation and edge cases. Most users should get their answer at Tier 1 or 2. If the majority need Tier 3 to do their jobs, the design has failed.

I built it as a working system rather than a set of screens, because the argument needed to be seen to be settled, and because a prototype that runs lets the client's reaction come back as a new build the next day instead of a new deck the next sprint.

The outcome

A stakeholder at the account said it in two words.

“10x better.”

Within days the account signed a $1M year-over-year contract, a starter deal expected to grow. It was a new contract, not an existing deal the prototype rescued, and the experience had stopped being the objection.