The brief was “add data visualization”. I decided what that meant.

It came in as a business idea

Charts would make the comparison tool better. Which data, in which format, and where a chart would actually help was the work.

BlackRock runs the iShares fund websites for investors in the UK, Germany and the US, and the comparison tool is where they put funds side by side. Regulation defines that data: what can be shown, how, and with which disclaimers. There were a lot of Compliance rules, so nothing could be hidden or summarised away.

These are technical users. They know what they are looking for and they need every figure on the page. In a regulated product you can't remove the complexity, so I changed how it was shown. All of this happened while the team moved from Sketch to Figma, a migration I was responsible for on our product, so every component I designed also had to work as a decision for the whole library.

Start where the company was looking

I ran round after round of unmoderated tests to find which sections of the table benefited from a visual comparison. The answer was: not all of them.

I set up the tests on UserTesting.com, wrote the flows, ran them and analysed the results myself. The question underneath was narrower and more useful than "add charts": where does a chart earn its place, and where is it just decoration?

Sustainable funds were a firm-wide priority, and it was the right call on the merits too: their indicators reward being compared side by side. Then the format. Pie charts looked richer and answered a question nobody was asking. Comparing several funds on one indicator is what bars do best, and I tested it instead of assuming.

The numbers never leave

I placed each chart under the row it draws, next to the figures and never in their place.

Compliance needs the number, the investor needs the pattern, and both stay on screen. The table keeps every figure the rules require, and the chart underneath shows how four funds compare on that one indicator. I defined the chart formats with engineering, instead of handing over a picture of a chart, and that is what made them buildable.

A second signal for every bar

Four funds, four colours, and colour was the only thing telling you which bar was which.

That leaves out anyone with a colour vision deficiency, and it stops working the moment the page is printed in black and white. Contrast ratios do not fix it, because the problem is identification, not legibility.

So the bars carry a second signal: fill patterns, one per fund, matching the legend, that a user can turn on. Colour still does the fast, comfortable work for everyone who can use it, and it stops being the only thing that does. WCAG AA covered the rest of the screen too: contrast, type sizes and focus order.

Specs for every breakpoint, market and state

Engineering told me which questions kept coming up while building, and I started answering those before they were asked.

Most of the format was the BlackRock standard: a company with a lot of rules and real design maturity, especially around accessibility, and being held to that bar taught me what a good delivery looks like. What was mine was the migration from Sketch to Figma for iShares Compare. Rebuilding every component was the chance to update things that had not been touched in years, and to make accessibility part of what "migrated" meant.

The specifications lived in a Figma master: you changed them there, and the change flowed down into the library. I took that idea with me, and the checkout library I later built at Jimdo works the same way.

The doubt we kept saying out loud

We knew comparing several funds on a phone might not be worth much, and we built it anyway.

The reasoning held up for what we knew: our users were overwhelmingly on desktop, that is where every round of testing ran, and parity across breakpoints was already the plan.

What I would change is one round spent finding out. A doubt the team keeps repeating out loud is already a research question, and usually the cheapest one on the list. When we finally revisited it, the answer matched the doubt we had started with, and mobile comparison was deprecated after I left.