Case study 01 · Institutional fintech · AI

Edgefolio: eight tabs into two pages

Fund evaluation hid the decision inside eight tabs. My research found two readers with opposite jobs, so I gave each one a page of its own and put their questions in the order they ask them

Read
Vantage Global Macro Fund
Vantage Capital Partners LLPAs of 31 Mar 2026

The old profile in miniature. Click the tabs, then collapse them into one page. That was the first answer.

7 of 7allocators reached the data tier unprompted in testing
420+fund managers onboarded at one tier-1 prime broker
20%projected conversion lift, on a buy cycle of months to years
Owned
I led design on the fund profile, end to end
  • The problem was cognitive overload, so I proposed the fix, took it to the head of sales and got the buy-in
  • The patterns I added to the design system were adopted across the company fast
  • I built the framework for working with AI, a knowledge base for design and research and AI in the daily work, and proposed it to the CPO in January. By April the product and design teams were AI-first
Scope
  • Fund profile, onboarding, Investor Hub
  • Information architecture, new design-system patterns, the AI prototyping framework
Method
  • Session recordings and allocator interviews
  • Personas rebuilt as AI research agents, then tested against real allocators
  • UX psychology for the architecture
  • AI-built prototypes for stakeholder presentations and user testing
ProblemNo brief beyond "redesign the fund profile". I defined the problem from research: cognitive overload. A fund hid inside eight tabs, and nothing told you which tab held the answer.
The callI turned the eight tabs into two pages, one for each reader. The first walks the analyst through five questions in the order they ask them. The second, the Investor Hub, holds the proof for the committee. The complexity never disappeared, it moved off the reader and into the structure. Designers call that Tesler's law.
ResultIn testing, 7 of 7 allocators reached the data tier on their own. The onboarding I rebuilt brought 420+ managers in at one tier-1 prime broker.

Finding the real problem

There was no brief beyond four words, redesign the fund profile. No problem statement, no definition of the user. So before designing anything I went looking for what was actually broken. I mapped the information architecture, watched session recordings of how each reader moved through the product, and interviewed allocators about what slowed them down. The same pain kept coming back: cognitive overload, too many clicks between a question and its answer. I also rebuilt the personas as AI research agents and tested what they said against real allocators, which nobody there had done before.

Overview
The Overview tab: fund introduction, fundamentals, strategy, team and share classes.
Pitch
The Pitch tab: the manager's sustainable-investment approach.
Analysis
The Analysis tab: return statistics, cumulative growth and monthly returns.
Exposures and holdings
The Exposures and holdings tab: performance and breakdown charts.
Peer analysis
The Peer analysis tab: a dense comparison table against the peer group.
News and comments
The News and comments tab: a blog-style feed of fund updates.
Data room
The Data room tab: a folder tree and file table of fund documents.
Contact
The Contact tab: a message form to reach the fund manager.
The eight tabs of the original profile. This is what the research worked from.

I turned the research into personas, then rebuilt them in NotebookLM as AI agents grounded in the research itself. Between iterations I put my questions to the agents, and what survived went into sessions with real allocators. The agents sharpened my questions. The users settled the answers.

Who uses a fund profile A top-of-funnel timeline. A prime broker introduces the fund and the fund manager keeps the profile current. The allocator's analyst screens it, the investor builds the case in the data room, and an IC memo goes to the committee for a yes or no. This case study covers both sides, the highlighted stages: the fund manager who builds and maintains the profile, and the analyst and data room on the investor's side. The fund profile The decision Prime broker Fund manager Analyst Data room Committee introduces the fund builds & keeps current discovers & screens DDQ, ODD, the IC memo decides, yes or no

Then the research stopped being about navigation. Two people were reading the same profile, and they arrived with opposite jobs. The analyst screens funds out, hunting for the reason to say no and move on. The portfolio manager comes later, building the case that goes to the investment committee. One reader wants to leave quickly, the other to stay and dig. The eight tabs were never the real problem. One profile was trying to do both jobs at once, and it did neither well. So each reader got a page of their own.

Why one long page lost

The first answer was one long page. Every tab stacked in order, nothing hidden, scroll to whatever you came for. Jobs to be done is what killed it. The first page has one job, catching interest, and it works the way a pitch works: put the cards down and earn the next scroll. The Investor Hub has the opposite job, and it only matters once someone is already interested. On one page the pitch and the proof compete for the same scroll. Two pages let each start where its job starts.

Five questions, in the analyst's order

The analyst came first. In the recordings they bounced into a tab, found nothing that answered their question, and went straight back out. No tab told them which question it answered. These are readers deciding where institutional money goes. They will not hunt for the number that settles it. So I mapped the order a screener actually asks their questions and made that order the shape of the page. Five questions, each section earning the next. I capped it at five because past that, testers lost the thread. I called the page the Conviction Funnel, because that is its job: build the conviction to keep reading, or hand the analyst a fast no. For a screener, both count as a win.

Each call maps to a named principle. Tesler's law: the complexity never disappears, it moved out of the reader's clicks and into the structure. Cognitive load set the cap at five. Information scent: every heading promises the answer behind it. Progressive disclosure: reasoning before numbers. Jobs to be done: the analyst's job, in the analyst's order.

  1. The strategy and the manager's reasoning, before any number.

    In the profile, in miniature
    Structural dislocations in rates and FX, in liquid macro positions, sized to survive the tail, not just to win the base case.

  2. Returns, key metrics, the audited track record.

    In the profile, in miniature
    +87.3% since inception · net
    Top quartile, trailing three years

  3. Drawdown, volatility, exposures, how the fund behaves under stress.

    In the profile, in miniature
    Rates
    FX
    Equity
    Max drawdown −6.2% · Ann. vol 8.4%

  4. Team, structure, service providers, the compliance posture.

    In the profile, in miniature
    Vantage Capital Partners LLP
    UCITS, Article 8 SFDR
    Eight-year audited track record

  5. Liquidity, minimums, the data room and the path to allocation.

    In the profile, in miniature
    Due-diligence questionnaire.pdf
    Audited financials, 2025.pdf
    Minimum ticket · $5M

A page this dense only feels real when you can click through it, so I built the prototypes with AI tools, for stakeholder presentations and for user testing. In testing, allocators went looking for the data on their own. The order did the persuading.

The Conviction Funnel fund profile, five sections from hook to access
The top of the Conviction Funnel fund profile: the track record, the investment strategy and the manager's commentary
The fund profile. Scroll inside it, the nav and fund title stay put
"I don't trust a number I can't see the reasoning for."
What an allocator told me in a research session I ran. It became the rule the whole rebuild followed.
Before · eight tabs, no order
  • Overview
  • Pitch
  • Analysis
  • Exposures
  • Peer analysis
  • News
  • Data room
  • Contact
1Conviction Funnel

The screener · invites you in

  • Hook, proof, risk, trust, access

from Overview · Pitch · Analysis · Contact, re-sequenced

2Investor Hub

The proof · when you're ready

  • Analytics, peers, data room, gated docs

from Exposures · Peer · News · Data room

Eight tabs, re-sequenced into two pages. Solid lines build the decision, hollow lines hold the detail.

A second page that holds the proof

When the analyst is convinced, the portfolio manager takes over, and the job changes from persuasion to proof. That work needed its own page, so the Investor Hub holds the deep half. The hub gives them two things: access to the data room, requested and granted rather than handed over so compliance could sign it off, and a memo tool that turns the data room and the investor's activity into a formatted committee memo.

The Investor Hub evidence layer, performance analytics, data room and gated documents
The Investor Hub's evidence layer: risk analytics, exposures and the data room's folders and documents
The Investor Hub, what the PM walks into committee with. Scroll inside it

What the fund manager sees

There is a third reader in this story. The fund manager watches the same profile from the supply side, and their Fund dashboard is the Conviction Funnel run in reverse: who is looking, what they read first, what still needs work to win the allocation.

The manager's Fund dashboard for Vantage Global Macro: live allocator activity, the deal pipeline, introduction requests and profile health
The manager's Fund dashboard. The live session shows the commentary being read before the numbers.
Who uses a fund profile A top-of-funnel timeline. A prime broker introduces the fund and the fund manager keeps the profile current. The allocator's analyst screens it, the investor builds the case in the data room, and an IC memo goes to the committee for a yes or no. This case study covers both sides, the highlighted stages: the fund manager who builds and maintains the profile, and the analyst and data room on the investor's side. The fund profile The decision Prime broker Fund manager Analyst Data room Committee introduces the fund builds & keeps current discovers & screens DDQ, ODD, the IC memo decides, yes or no

Getting a fund listed in the first place was manual and piecemeal. I rebuilt it as an automated journey for the capital-introduction teams, from the first invite email to a manager teaching themselves the product.

01

Reach

Email sequence

Triggered invite from the capital-introduction team·staged reminders, no manual chasing.

02

Activate

In-app onboarding flow

Guided profile build, step by step·benchmark and holdings set up in place.

03

Retain

Learning hub

Teaches the product as the manager goes·self-serve, so the team scales without headcount.

How a profile earns an investor's trust is an essay of its own: Who uses a fund profile?

What it changed

This was mine end to end. I ran the research, mapped the information architecture and restructured the profile around what it showed. I built the variations, tested them with allocators, and presented the strongest to the head of sales, who backed it. From there I set the direction and led the rebuild, users and stakeholders with me the whole way. The rebuild needed patterns the design system did not have, so I introduced them. Other teams adopted them fast.

In testing, 7 of 7 allocators reached the data tier on their own. The onboarding I shipped brought 420+ managers in at one tier-1 prime broker. Institutional allocation runs on months to years and the profile did not ship on my timeline, so the 20% conversion lift is an honest projection. The framework for working with AI that I proposed to the CPO in January was how the product and design teams worked by April. The onboarding journey became the template for every other client bank.

Reflection

The brief said redesign the fund profile, nothing more. The session recordings said two jobs, each needing its own surface. Once I read the sessions for intent rather than friction, eight tabs stopped being a navigation issue and became a structural one.

There was no product manager either, so where to look and what to ask were mine to work out. I went deep on session recordings and allocator interviews, and only past the middle of the project did I think to sit down with customer success and sales. They are in those conversations every day. They know what people ask for and what the tickets keep saying. They also know what customers will argue about, which turned out to be the useful part. I affinity-mapped what they gave me. It overlapped with what the research had already found, but it would have set my thinking on the right track sooner and saved a week or two. On a project shaped like that one, I now book that time in the opening weeks.

Next

02 · A symptom checker 5.5M people use