01 — Overview
Mini Map visible in the right panel alongside the patent document. Concept groups load automatically from the search query.
A full end-to-end project I led solo — research, design, and dev handoff — during a period of significant organisational change.
During this project Clarivate went through significant redundancies. The PM was let go. The delivery manager was let go. I stepped up, doing the research, writing the first draft of user stories, managing stakeholder alignment, and making product calls alongside Matt Weber, our PM Director.
It wasn't the plan. It was just what the project needed.
Patent attorneys, IP strategists, professional searchers. People with decades of experience and very precise mental models. They don't change workflows for novelty. They change when something is measurably better.
Power users had built workarounds. Some had migrated parts of their workflow to competitor tools. New users expected speed and intelligence out of the box. Both groups had to be served.
Reading a patent isn't like reading an article. A single document can run to hundreds of pages of dense technical and legal language. Analysts aren't reading for pleasure. They're hunting for specific concepts, under time pressure, often across dozens of documents in a single session.
Highlighting is the primary navigation tool for that hunt. Without it, or with a version too basic to be useful, the work becomes slower and riskier. A missed term in a prior art search isn't an inconvenience. It can invalidate a patent claim.
One word per colour. No concept groups. No synonyms. No way to track "battery", "Li-ion" and "energy storage cell" together. A tool so basic that experienced users had stopped using it.
Richer highlighting, multi-term search, better document navigation. Users knew the gap existed. They were actively filling it with third-party tools and competitor products.
I ran every session myself. Professional searchers, IP analysts, attorneys, and senior internal stakeholders including a Patent Search Principal Lead and a Director of Patent Operations. We made a deliberate call to iterate fast. When something came up consistently across sessions, we fixed it immediately rather than waiting for the study to end.
The clearest example: icon-only buttons were invisible to users. After 8 consecutive failures I switched to labelled buttons mid-study. Failure rate dropped to zero.
Icons only. 8 out of 8 participants failed to find the feature unprompted. Common response: "I didn't realise that was clickable."
Labelled buttons. 0 out of 5 failed. Discovery was immediate and unprompted.
The Mini Map lives in the right panel. Always visible, always the exact height of the document viewer. Colour marks show where every concept group appears. Click any mark, jump to that exact line. Scroll the document, the indicator moves. Always in sync.
I used Claude to pressure-test spec logic, catch inconsistencies across documents, and draft sections of the handoff. Running this solo means no second designer reviews your calculations. AI helped close that gap.
I checked everything. Every number. Every assumption. The design decisions, the research synthesis, the prioritisation — that was mine.
Knowing how to use AI consciously, with proper oversight, is itself a skill. One I've deliberately built.
Where AI helped most:
The numbers below come from three different places and mean three different things. Mixing them would make the story dishonest.
Record View is the strongest retention driver in the product. Users who engage with it in the first 2 days are retained at nearly double the rate of those who don't. The highlighting feature lives inside Record View. A broken highlighting experience was actively undermining the stickiest part of the product.
Record View retention · 2-day golden window · 32.7% current adoption · 6,378 MAU
Multiple participants had already moved parts of their workflow to competitor tools specifically because of the highlighting limitation. This feature is a direct win-back opportunity.