top of page
  • Writer's pictureJakob Nielsen

(Experimentation + Iteration + Testing) × Learning = Great UX

Summary: Exploration, refinement, and evaluation are the keystones of exceptional user experience design. By engaging in repeated cycles of design trials and evaluations, designers maximize the potential of their work, leading to superior outcomes that stand the test of rigorous user testing. UX brilliance stems from persistence.


UX quality is driven by experimenting with many different design iterations. The more different versions of a design you create and test, the better your final outcome.

I usually avoid sports metaphors since different sports are popular in different countries, making expressions like “home run” bad for internationalization. However, please excuse me for deviating from this rule here. Many sports provide irresistible metaphors for the benefits of trying many times.

Here’s an example from soccer. You can retain possession and kick the ball around for 90 minutes, but you won’t win the match unless you score. And you don’t score unless you take shots at the goal. You’ll miss some shots, and the goalkeeper will save some. But a side with many shots at goal will usually score some goals.

The more shots you can take at the goal, the more goals you will score. (Midjourney)

There are several factors that sum to creating great UX:

  • Experimentation. You must have a design culture that encourages experimenting with many different design options, rather than locking yourself down easily in the process and sticking with whatever you initially think to be a great design direction. Trying many directions improves the probability that one of them will be better than your first idea.

  • Iteration. Don’t just try many directions, but refine the more promising design directions. Usability often lives in the details, and only by polishing a design through many iterations of trial and error will greatness result.

  • User testing. You need an evaluation function to determine which of the many design directions are the most promising and what changes to make during iterative design. Empirical data from observing real users serves as our evaluation function in UX design.

Diamonds look better when polished. The same is true for UI. The more you polish, the better the usability. The more you iterate, the more polished your design. (Midjourney)

For the first two of these bullet points, the more, the better. The UI design space is effectively infinite with so many different design dimensions you can tweak in so many different ways. More versions tested equals a better final result. (In contrast, the third bullet point, user testing, can be contained: there’s no real benefit to testing with more than 5 users when you are exploring the multidimensional design space. Better to finish each text round quickly and spend the time on more exploration of the far reaches of the design space.)

I can’t really show you the multidimensionality of the design space in a 2-D picture, but here’s an attempt. (Midjourney)

It’s a hard bullet to bite for many elitists such as myself that randomness and multiple attempts are major determinants for superior performance. But much research shows that they are. Not just for UX design, but for eminent intellectual performance in general. Whatever the field — art, science, engineering — even the most brilliant genius doesn’t always hit home runs. (Sorry for that sports metaphor. I mean that they don’t deliver their very best work every time. Some deliverables are better than others. Not all Beethoven’s symphonies are as good as his Fifth.)

The way to eminent performance is to try many times. There will be a distribution of excellence in the results, so the more attempts, the higher the score for your best work. If you’re a composer, some of your symphonies will be performed more often than the rest in 200 years. If you’re a UX designer, you luckily don’t have to wait that long — next week’s user testing will show whether you need a trip back to the old drawing board.

In UX, the quality distribution doesn’t matter. All that matters is the quality of your top attempt, because second-best and below will never ship. The more times you draw from that probability distribution of eminent performance, the higher the score of that number-one attempt.

Learn From Experience

And yet, there’s a tweak. Yes, no matter the shape of a probability distribution, the more attempts, the higher the score for the number-one best attempt. But you can also improve your performance by shifting the entire distribution to the right.

  • You’ll still produce some bad designs, but not as many as before.

  • Your average design will be better, but that doesn’t matter if you follow my advice to generate as many designs as possible. (But in some cases, you don’t have budget or time for even the fastest iteration, and thus your average design quality will sometimes matter.)

  • Most important, your best attempt will be better out of the N designs you have time to try.

How do you shift the quality distribution of your UX work to the right? By learning from past attempts: what works, what doesn’t. The more-skilled player will hit the bullseye more often in a game of darts.

Dare to fail, but learn from your failures. (And your successes.)

In a game of darts, throwing more darts will hit the bullseye more often, but the best player will have the highest percentage of hits. (Ideogram) As an aside, getting AI to depict a dartboard and dart players was amazingly difficult. This sport must be ill-represented in the training data.


Top Past Articles
bottom of page