Improving MATLAB's Plot Viewing Experience

MATLAB is used by engineers and scientists at Tesla, NASA, aerospace companies, and research labs worldwide. It's where they analyse simulation data, compare test results, and make sense of complex outputs. That work produces plots, sometimes dozens per session.

The standard experience: run the code, watch new plot windows appear on screen, each one floating separately, stacking on top of each other. To compare two plots, you drag and resize windows by hand. Run the code again and every arrangement is gone. Start over.

It's not a dramatic problem. Nobody files a ticket saying "this is broken." It's just slow, repetitive, and accepted as part of the job.

The engineering team had already solved it: Tab and Tile View. All your plots in one window as tabs. One click to arrange them all into a grid automatically — no dragging, no manual sizing, no losing your layout when you re-run the code.

Every engineer in the study reacted the same way the moment it was shown to them. The feature worked.

But before Beta, the researcher had given those same engineers one task: find the feature that arranges plots into a grid. Five minutes. Only 1 of 7 found it without help.

The feature was buried in a 3-dot menu under the label "Tile All." That phrase meant nothing to someone looking for a grid arrangement. And even engineers who found it hit a second wall: a submenu with radio buttons and a 5×5 grid picker — Single, Left/Right, Top/Bottom, then choose your exact layout. Like inserting a table in Word. Nobody connected that interaction to "arrange my plots." The label sent them to the wrong place, and the submenu confirmed they were lost.

Before proposing anything, I tested the current experience end-to-end. Then I looked at how engineers work in the tools they already use daily — VS Code, Figma, Chrome DevTools, Power BI, Tableau. The goal: any improvement should fit what they already know, not ask them to learn something new.

The benchmarking surfaced more than UI patterns. MATLAB's tab view was missing things that felt standard everywhere else: no way to recover a closed tab, no pinning, no grouping. The only way to get a plot back was to re-run the code.

None of these were the original brief — but they were real gaps. I ran an Impact-Effort Matrix with the researcher, PM, and engineering lead to put every candidate improvement on the board and decide together what we were actually solving. "Quick switching between views" landed High Impact, Low Effort. Unanimous. Everything else moved to the backlog.

Since we were all aligned on making the view switch simpler to find and use, I tried a few concrete ideas, worked through each one, and learned from what didn't work.

I stopped asking what the ideal solution looked like and started asking what was achievable: the 1980s codebase, existing design system components, and patterns MATLAB users already knew.

Not glamorous, but a checkbox was already used throughout MATLAB's menus. Engineers knew how to read it. It fit the design system. The codebase could support it. Check it and the system picks the best grid layout automatically. No configuration. No choices.

The redesign shipped with the next Beta build. Feature discovery went from 1 in 7 to all 7. Plot arrangement dropped from 35–45 seconds to 5–8.

When I walked the engineering team through the final design, the developers who'd maintained that codebase for years reacted first — before the PM, before the researcher. They'd watched users work around this problem for a long time. They knew the fix was right. That meant more than any test metric.