The most common failure of a website is not a technical defect in any single layer. Usability, information architecture, performance — any of these can be the culprit, and it rarely matters which. What matters is that a visitor arrives with a task and cannot complete it. New browser capabilities and an ever-richer tech stack have made it cheap to keep piling complexity onto the average site, whether it presents a product, takes bookings, shows a portfolio or sells goods. Delighting the user has crowded out the simpler goal: letting them do the thing they came to do.

Make the essential actions first-class

If there is one improvement worth making, it is this: identify the most crucial actions users should be able to take on a page, then make them as easy and accessible as possible. Visual effects, elaborate graphics, polished interactions and tracking scripts all come second.

A lost suitcase, and a site that couldn't help

A personal example makes the cost concrete. On holiday in a remote area with very limited internet access, I lost my luggage and had few places to buy replacement clothing or cosmetics. The airline site would not load over my constrained data connection, so I could not learn where the bag was or when it would arrive — it wouldn't even surface a phone number to call. An email address I found elsewhere turned out to be outdated.

The site had no progressive enhancement and no graceful degradation. It only served privileged users with a good enough connection, because that was the only way to download the large amount of JavaScript responsible for constructing the entire experience. A simple form with two text inputs, plus basic textual information as a fallback, would have resolved my problem. The developers had undoubtedly spent countless hours making the experience delightful, and I could not even see it.

Why this keeps happening

Getting caught up in the moment is easy. So is following the milestones as they appear as tickets in Jira or another project management tool. Solutions we are used to get reused; code gets copied from earlier projects or from Stack Overflow. And it is easy to assume that something which "works on my machine" will work for everyone else.

The difficult part is stepping past the features that add new value and looking at the parts of the app that were overlooked along the way. It is hard to stay current as new features and browser APIs ship. It is hard to consider that someone might not share the privileges we have.

Audit your own assumptions

Familiarity makes this harder than it sounds. Once a solution is built, it becomes difficult to imagine how people fail to follow the instructions and clues we placed on screen, or how the page feels to an unsighted user or to someone who navigates with the keyboard alone. Edge cases and anything beyond the happy path go untested. We overlook that we are working on a powerful MacBook with a sharp display and a steady internet connection. We forget that not everyone is a native English speaker, and that a word which seems self-explanatory to us may mean nothing to a user who never uses it in everyday conversation.

Look at your site as if for the first time

Use it in production, with the stream of third-party resources that may be missing in development mode. Use it over a very poor connection and measure how long a simple task such as filling a form takes. Try it on a different device you have not used before.

Watch a real user at work

Find an actual user of your site and sit in on a user testing session. Whatever you assume causes headaches for your users, some of those assumptions will be challenged, and you will end up with a list of fixes you would not otherwise have considered.

Beyond the buzzword

Progressive enhancement is worth treating as more than a buzzword. For further reading, Jeremy Keith's presentations on the topic are a useful starting point, as is Aaron Gustafson's article that popularized the idea.