The Design Principle You Defended Last Week Probably Wasn't Science. Neither Was Mine.

The cost of treating convention as science, and why it matters more than you think.

There is a version of design that looks like science, with principles, frameworks, and guidelines, names attached to it like Nielsen, Norman, and Gestalt, documentation, rationale, and the quiet authority of being cited in enough articles that questioning it feels unprofessional. It also has very little controlled evidence behind most of it.

Not none, and that distinction matters. There is real science in design: contrast ratios in accessibility are measured, not guessed, Fitts’s Law has data behind it, and cognitive load theory comes from psychology research. These things are real, and they matter. But they are a small fraction of what designers present as truth every day.

Take the three-click rule. For decades, designers enforced it like physics. I enforced it myself in my early years, with full conviction. The idea was simple: users will abandon your site if they can’t find what they need in three clicks. Sounds reasonable, feels true, but a study by Joshua Porter analyzing 44 users attempting 620 tasks found no increase in drop-off after three clicks and no decrease in satisfaction either. The rule was never backed by data, it came from a book published in 2001 as a loose suggestion, and the industry turned it into law.

Why is the golden ratio still taught in design schools, referenced in portfolios, and built into Lightroom and Photoshop as an official composition tool used by hundreds of millions of people worldwide? An idea popularized in the 19th century that visual beauty follows a specific mathematical proportion. Decades of empirical research on the topic have produced conflicting results, findings have been mixed and disputed, and the evidence is far from settled. And that is before accounting for culture: an empirical study with 277 Korean participants found a significant preference for the root ratio (1:1.414) rather than the golden ratio: not a refutation, but a clear signal that aesthetic preference has a cultural component Western studies often ignore. Proportions are determined by mathematical logic, but aesthetic preference is shaped by experience and cultural context. Adobe shipped it as a feature, but the science never settled.

Skeleton screens versus loading spinners tell a similar story. I wrote about this in detail previously, but the short version is this: a 2017 internal study published by Viget, a design agency, found that skeleton screens actually performed worst in perceived duration compared to spinners, while a 2018 peer-reviewed study found a murkier picture: skeleton screens scored higher on perceived speed and ease of navigation, but users who saw a spinner completed the task faster, and the paper itself found no statistically significant difference across most of its comparisons. These are not equivalent sources, one is agency research, the other passed academic review. But even with that asymmetry, the question remains unresolved: the stronger study did not settle it definitively either. The two most cited studies on the topic point in different directions, under different methodological conditions, on the metric that matters most: whether users actually feel they waited less.

This is worth sitting with for a moment. The problem here is not that the science is absent, but that even the stronger evidence did not settle the question, and the industry decided to treat it as definitive anyway. Skeleton screens became standard practice not because the evidence supported them, but because they looked more modern and the right companies adopted them. The adoption was treated as a closed question when it wasn’t.

Three different patterns in different areas of design, with the same problem underneath all of them: we adopted them as truth without clear evidence. These aren’t the only examples, or the worst ones; they’re just the ones I could verify closely enough to stand behind in print.

These aren’t isolated mistakes, they are symptoms of a larger pattern. The examples above are conventions that spread organically, rules of thumb that traveled from blog post to blog post until they acquired the weight of fact. But the design industry also built something more deliberate: formal structures designed to codify decisions and give them institutional authority. Frameworks, methodologies, and guidelines that don’t just describe how design is done, but prescribe how it should be done. And those structures have the same problem, dressed in more official clothing.

Take Nielsen’s 10 usability heuristics. Heuristic evaluation as a method has legitimate research behind it, and the heuristics themselves are useful patterns for identifying potential problems. The issue is what happens to them in practice. A designer runs a heuristic evaluation, flags an issue under “consistency and standards,” and the report lands on a stakeholder’s desk with the implicit weight of Nielsen’s name behind it. The original finding, that these are starting points for identifying problems, not laws that determine correct design, gets lost somewhere between the method and the meeting. What survives is the authority, and the checklist becomes the conclusion.

Or take Design Thinking, a framework so broadly defined that it can mean almost anything, packaged by IDEO and Stanford Design School into a process that thousands of organizations invested heavily to implement. The research that exists tends to focus on educational settings, a 2023 meta-analysis found moderate positive effects on student learning outcomes. The organizational contexts where most of the money was spent have simply not been studied with the same rigor. That is not evidence that it does not work, but it is evidence that nobody has bothered to check. Part of why it sold so well without that evidence is that it solved a different problem entirely: it gave executives a vocabulary for innovation and a process they could point to. Whether it produced better outcomes was secondary to whether it produced visible activity.

None of this means these frameworks are useless, some of them are genuinely helpful as starting points. But a starting point is not a law, and the industry stopped treating them as one a long time ago. The authority of the source replaced the need for evidence. That is how convention becomes doctrine, and doctrine, when you internalize it completely, stops being something you follow, it becomes something you are.

This has consequences beyond bad design decisions. When you believe convention is science, every challenge to your work becomes a challenge to the laws that govern the universe, and you are not being stubborn, you are being correct. And when you are always correct, you stop being a person with opinions and become a person with a mission. I know how far this can take you, because I lived it. The field does not just teach you principles, methods, and frameworks, it teaches you who a good designer is, and you spend years trying to become that person. Every deviation from the doctrine stops feeling like a disagreement and starts feeling like a threat to your identity.

What follows is less an argument than an account.

I spent years treating every design decision like it had a correct answer which I was responsible for defending, where every pushback was a threat and any compromise was a failure. I worked overtime because there were never enough hours in a day to deal with everything that needed to be done.

Eventually I stopped going out on weekends since I was always tired from handling everything alone, as I believed that I was the one responsible for the truth. The signs of burnout were obvious to everyone around me, but I never saw it coming.

Then I burned out completely, and at the lowest point I thought about ending my life. I am sharing this not for effect, but because it is what happened, and because I think more people in this industry have been closer to that point than they admit. Burnout rarely has a single cause, and mine was no exception, but treating convention as the weight of scientific fact was the largest single factor, since it made eight hours of every day unbearable. I got fired shortly after, which freed me from the professional pain almost instantly, since all the worry I spent on every decision became irrelevant as I walked out the door. But the internal work of understanding what had happened took much longer.

The problem was never caring about design. It was treating convention as if it had the weight of physical law, and building an identity around defending it.

What came next was gradual. Losing what I was so desperately working on forced me to stop caring so much, first about my job, then about design, then about being right, and lastly about defending every decision like my identity depended on it. I started focusing on my own wellbeing more than my work, and somewhere in that process, without trying, I became a better designer. More focused on what actually mattered and less interested in winning arguments about things neither side could prove. I just wanted to be happy, and it turned out that was enough to make me better at my job.

The certainty was hurting my work. It was protecting a false belief that design had the right answers, and that I knew them. Letting go of that belief was the most beneficial thing I ever did as a designer and as a human being.

Understanding that design is mostly convention does not just change how you think about the field. It changes how you show up in the room.

What changed wasn’t a single realization, it was gradual exposure to decisions where the correct answer depended entirely on context. I remember a discussion with a CEO about chart types for a dashboard, whether to use a bar chart or a pie chart. I had research on my side, or so I thought: pie charts get harder to read as the number of categories grows, and bar charts handle that volume better, a fair summary of decades of research on angle and area as perceptual encodings. I was confident I was right, and perceptually I probably was. But the chart was illustrative, a visual entry point to data the user would actually consume in Excel tables. Whether they read the proportions correctly was irrelevant to what the dashboard was trying to do, so the battle I was fighting had nothing to do with the problem we were solving. Eventually I let it go, and the conversation ended in minutes. Ironically, while researching for this article, I discovered that a recent study designed to test exactly this question found that chart type barely moves real decisions once you leave the lab.

The same thing happened with a navigation redesign I was asked to review. I came in with a clear position: the current information architecture violated the principle of progressive disclosure, and I had the heuristic to back me up. The engineer I was working with pushed back. His concern was not about the architecture, it was about implementation cost: the redesign would require two weeks of work on a feature that analytics showed fewer than 3% of users touched. He was not wrong about the architecture, and neither was I. We were just looking at different perspectives. I dropped the heuristic and we agreed on a label change that took an afternoon. The 3% became slightly less confused, and nothing else changed. That was the right outcome for the business.

That shift showed up everywhere after that. Letting go of the need to be scientifically right made space for something more useful: the ability to reason about what actually mattered without needing to win.

None of this is an argument for doing whatever you want and calling it design. Conventions exist for good reasons: they provide a starting point, reduce friction, and give teams a shared language. The problem was never using them, it was forgetting what they truly are: not definitive answers, but reasonable defaults that can still be questioned depending on the context. The rest is judgment, context, and the honesty to know the difference between what you can prove and what you are assuming.

There is a practical test to see if your decision needs real data to back it up, think of it as a spectrum of proximity to your context. At one end, observed behavior from real users on your actual product: a metric that moved, a session that revealed a pattern, a test that ran with your audience. At the other end, convention, expertise, and educated intuition transferred from elsewhere. What matters is not just where a decision sits on that spectrum, but whether it actually moves the outcome you care about. Low-contrast text being hard to read does not need to be retested in every product, but whether a skeleton screen feels faster than a spinner in a loading page with a huge drop-off probably does. Knowing where you are on that spectrum, and how much it matters for the decision at hand, is the skill the field rarely teaches.

Letting go of certainty is not always the right call either. I’ve also seen teams dismiss a clear behavioral signal because a stakeholder’s intuition felt more senior than the data, and in those cases the data should have won. Yielding is not the lesson here, developing your judgment is.

Before defending a decision as certain, ask:

1. Where does this belief come from: my own data, or a borrowed convention?
2. If this belief is wrong, does it actually change something that matters for the outcome?
3. Only stop and test when the answer to both is borrowed convention and yes, it matters. If it doesn’t matter, go with whatever sits best with your stakeholders and move on to the decision that actually matters.

Most of the sources cited in this article, including the ones used to make this argument, sit somewhere in the middle of that spectrum. That is honest, and it is also my point. This spectrum itself is not a law either, it’s a convention I’m proposing to think more clearly, testable against your own experience, but not proven by it.

That honesty is harder than it sounds. It means looking at a convention you’ve cited a hundred times and asking, for the first time, whether you actually know it works. It means telling a stakeholder “I think this is right, but I can’t prove it” instead of performing certainty you don’t have. It means accepting that the person across the table who disagrees with you might not be wrong, they might just be working from a different set of unproven assumptions. The more comfortable you are in the room, the easier this becomes.

Design is mostly convention, and pretending it isn’t is a problem.

I blindly defended those conventions for years before I understood what I was actually doing. I don’t think I am alone.