Organised around the workday, not the data.

The screen they work in, not one they visit.

Relationship managers (RMs) open this platform first thing and close it last. The brief was to rebuild the RM-facing side of it for a wealth management desk.

There are no invented numbers on this page. The decisions are the story.

Six screens and one account.

Screens I saw

6, over a live screen share

Accounts I saw

1

Documentation

None

Analytics

None

Product context

None

Domain knowledge

None. On day one I couldn’t have defined a fixed income product.

That last row matters, because the findings below came out of the instruments, not the interface.

It followed the data model, not the day.

The things an RM checks first sat at the same visual weight as the things they rarely touch. But the sharper version of the problem only showed up once I read the account rather than the screen.

What the data said that the screen didn’t.

All four can be worked out from data the platform already held. None of them needed a new data feed.

And what was wrong on the surface.

That last one isn’t a design opinion. A profit-and-loss column was headed in dollars and filled in yen.

Real constraints, and a real clock.

• 4 days, start to finish.
• About 40 minutes with the live product, over a screen share. No hands-on access.
• No real users, so no interviews and no usability testing.
• A test environment, so some content was stale or placeholder. I noted it rather than designing around it.
• Real banking data can’t be published, so the client name is changed and account details are blurred.

In the order it happened.

1. Read the data before the interface.

Before touching the layout, I went through the actual portfolio behind the screens. The four findings above came out of that, not out of looking at the UI.

It also showed which asset classes the interface was failing. Structured products, fixed income and derivatives had all inherited the equity column set, which is why those screens looked empty.

2. Audited the hierarchy.

I sorted everything into two buckets: what an RM checks first, and what they rarely touch. Without users to ask, this was the closest I could get to their day. I leaned on how wealth desks work, and said so, rather than pretending otherwise.

That gave me 15 findings, ranked by what they cost the user rather than by how easy they were to fix.

3. Reorganised around the workday.

Density and grouping now follow the daily workflow, not the account structure. Alerts, drift and upcoming actions moved up to where the first glance lands. The rarely used things moved down, not out. Nothing got deleted just to look tidy.

4. Kept the table dense, on purpose.

The obvious move on a cramped data grid is to add whitespace. I didn’t. An RM comparing 38 positions wants them all on screen at once. Halving the rows would have made the tool worse while making the screenshot better.

Same row height, same column count. What changed was alignment, number formatting, one value per cell instead of two, and a grid that sorts, filters and totals.

5. Replaced the pie chart with a sorted bar.

A pie answers “what is the mix”, which the RM already knows. A sorted bar answers “what is out of line”, which is the reason they opened the screen.

The currency chart kept a circular form, but it leads with the number itself, because that one figure was the largest unhedged risk in the account.

6. Added one screen, and only one.

Every existing screen was about a single client. Nothing answered “which of my clients needs me this morning?” So I added an Overview, built entirely from data the platform already holds.

Everything else stayed where it was. I didn’t want to move things that hundreds of trained people already know how to find, on the strength of four days and no user contact.

Taking inventory first.

The product had no linked text styles or published tokens. So before restyling anything, I inventoried what was actually on screen: type roles, spacing, colour and gradients, with usage names inferred from where each one appeared. The new components were built from that.

No users means no validation.

I can’t say this tested better, because I never tested it. The whole redesign rests on assumptions about how RMs work. They’re reasoned assumptions, but they’re assumptions.

There are no impact metrics here, on purpose. Everything above is what I found or what I built. None of it is what happened next, because nothing happened next. Making a number up would undo the rest of the page.

How I’d have measured it.

Since I can’t show results, this is what I’d have baselined before shipping anything:

Put it in front of RMs.

• Watch where they look first.
• Check my “checked first vs rarely touched” sort against how they really work. I expect to be wrong about a few of them.
• Find out what the two undocumented columns actually mean. I designed around them rather than removing them, but I was guessing.

Looking back.

Sorting by workflow instead of by data sounds obvious in a sentence. It took longer than a sentence to get right.

The strongest findings came from reading the portfolio, not the interface. I’d start there every time now.