Preqin~3 min read

An ESG score that meant nothing to the fund reading it.

Four weeks, one PM, one engineer, and two follow-on experiments that both failed.

Role: Product strategy, design, and research
Timeline: 4 weeks
Team: 1 PM, 1 engineer, me (plus investors and customer service)
Impact: 7 contracts, ~$560k of contract value
Platform: Web

Contracts closed

~$560k of contract value the modules contributed to

Pre-sales calls

Engagement

Contracts and pre-sales calls are commercial outcomes the sales team can verify against its records, counted over the two weeks after launch. Seven contracts at a pilot account value of about $80,000 comes to roughly $560k. A pilot group is not a random sample of the market, and the modules contributed to those subscriptions rather than constituting them, so read $560k as the contract value this work contributed to. Separately, 80% of survey respondents marked the modules very helpful. That is self-reported. Still missing: what the 60% engagement measures and across how many accounts.

Overview

Nobody could say what an ESG score meant for their fund, which made it hard to demo and hard to charge for.

Preqin was opening a new product line selling ESG data into private markets, presented as a static snapshot. Limited partners worried about greenwashing. General partners wanted to know what their peers disclosed.

Worked with: 1 PM, 1 engineer, investors, and customer service; 6 directors interviewed.

What I owned

I owned

Product strategy, including the read that an uninterpretable score was a commercial problem before it was a usability one
Three module versions and the reasoning behind each change
Both follow-on experiments, the decision to build them, and the read on why they failed
User research with both user types
Components contributed back to the design system

Decided with others

Feature sequencing, scored with the PM rather than set by me alone
What the underlying ESG dataset could support, with engineering

Research

Suspicion and benchmarking turned out to be the same question.

I interviewed 6 directors. One user is trying to catch a fund overstating itself, and the other is comparing. Both need the number against a relevant peer set, with the direction it has been moving. No data was missing. The work was moving from a snapshot to momentum and specific gaps.

Decision: heatmaps over tables

Three module versions failed the same question in different ways: where do I look first?

Version one used tables. They gave exact values and no priority. Version two showed risk across its real dimensions, with everything visible and nothing ranked. Version three used heatmaps to show where to look.

A heatmap does not show how much, and these users need exact figures for diligence. So the exact count stays in every cell.

Version one. A transparency score of 76% sits above three sub-scores for firm, portfolio, and asset governance, over a long table listing each ESG indicator with its pillar, theme, assessment tick, and source link. — Version one, tables. Exact to the indicator, and no help at all with what to look at first. Every row is the same weight as every other.

Version two. A bar chart counts portfolio companies against each environmental SASB issue, above a table listing 121 portfolio companies with the specific environmental, social, and governance issues attached to each. — Version two, risk across its real dimensions. Everything was visible and nothing was ranked, which is a different way of failing the same question.

The heatmap that shipped. Environmental, social, and governance SASB factors run in three columns, each row shaded from pale to deep amber by how many of the fund's thirty-two portfolio companies it touches, with a low-to-high legend above and the exact count still printed in every cell. — Version three, and the one that shipped. Shading carries where to look; the exact count stays printed in every cell, because these users still need the number for diligence.

Decision: preparation tools

I built a drawer and a wizard to help investors prepare questions, and both failed.

Gap analysis identified what to ask about and stopped, so every investor rebuilt the same brief by hand. The drawer put questions beside their gaps. The wizard walked through priority gaps and exported a PDF.

People could operate both. They rejected the idea underneath. Limited partners wanted their own thinking confirmed, and a finished brief took away the part they were paying for.

The guided preparation drawer, open over the transparency table. A meeting preparation guide lists three undisclosed indicators as numbered questions, the first expanded into a full suggested question citing peer disclosure, with a copy questions button and a persistent floating button in the corner. — The drawer: questions sitting beside the gaps that produced them, so a brief could be assembled without leaving the tab.

The wizard. A three-step rail (priority gaps, build questions, summarize and confirm) runs beside editable question fields, each pre-filled with a suggested wording that can be rewritten, skipped, or deleted. — The wizard: the same job as a sequence, with the questions editable and a brief at the end. Neither shipped, and both lost on the same point.

Scope

The heatmap module shipped, and nothing that recommended questions did.

Identifying a gap is evidence. Recommending questions is advice. These customers were buying something to check their reasoning against, not advice.

Results

Seven contracts closed and 50+ pre-sales calls came in.

I ran the right method against the wrong question, twice

Both follow-ons passed usability testing, which asks whether someone can do the task and never whether they want it done for them. Testing the changed positioning first would have been cheaper. I know two points on the wrong side of the line between evidence and advice, and not where it sits. Nobody revisited it while I was there.