Financial Tool
Financial analysts are under enormous pressure to produce actionable insights. Last year analysts spent over 26.5 billion dollars on research software, but the successful hit rate for investment ideas is still only 52%.
The Watson powered Financial Research Tool was designed to address specific pain points for equity researchers. On this projects was the UX lead, the primary UI designer, and I also conducted a series of interviews with both buy and sell side financial analysts. Working with a team of product managers and a handful of developers within IBM Watson, we built a functioning prototype of the tool in order to pursue a development partnership with one of the leading financial data aggregators.
Discovery
An equity research report is a document prepared by an equity analyst. It is a form of communication between financial experts and investors. The analyst conducts an in-depth analysis of a company, industry or even an economy and explains his findings in the form of a report. The purpose of preparing such reports is to provide investment recommendations to the clients (buy, sell or hold). Before speaking to any financial professionals, I immersed myself in the documents, interfaces, and tools used by analysts to produce a report. With a sufficient understanding of the domain to ask intelligent questions, the product manager and I began setting up end user interviews.
Research Findings
Our interviews took us to Ronin Capital, Vanguard, GW Capital, and RCN. We conducted a series of interviews with 9 volunteers. These volunteers remained sponsor user for the duration of the project and were able to contribute both initial thoughts and detailed feedback.
General Findings
Everyone we spoke to used Bloomberg Terminal (watch lists and news filters)
They found Social sentiment indicators to be very noisy and unreliable
Pain: Difficult to finding meaningful information within the large amounts of data each with a different format
Development
Analysts described a common and compelling pain point. They spent hours reading news and opinions. They sought information to prove their investment thesis, but often ignored refuting arguments. We had arrived at a use case for an AI. We returned to our sponsor users with wireframes of a News Triage application.
Specific Findings
New Triage would not reduce time spent reading news, instead it would improve productivity and effectiveness
Researchers spend ~2h/day reading news
All necessary content available in Bloomberg Terminal (#1 SEC filings, #2 sell side reports, #3 normal news, #? Messages/Chats)
The weighting of information source is key
System need to distinguish facts from opinions and weight these appropriately
Must add ability to trend thesis performance
Users are loath to update their news filters
Watson may need to understand cascading implications in order to tie news to theses
