Comment by brianpan
3 days ago
"Universal design" or "design for accessibility" will give you lots of examples of constraints that are not "commonly" needed ending up having much wider application and benefiting many other people.
Some oft-cited examples are curb cuts (the sloped ramps cut into curbs for sidewalk access) and closed-captioning (useful in noisy bars or at home with a sleeping baby).
There are many examples from the web where designing with constraints can lead to broadly more usable sites- from faster loading times (mobile or otherwise) to semantic markup for readers, etc.
Ah, this raises 2 important nuances:
- How severe is the impact, and
- How close is the default state to the constraint
Kerb cuts help everyone. Kids, the elderly, disabled people, and anyone distracted by their phone are all less likely to fall on their face and lose a tooth.
Web accessibility helps websites go from unusable for disabled people, to usable.
On the other hand, when a dev puts a website on a diet it might make it load in 50ms instead of 200ms for 99.9% of users, and load in 2 seconds instead of 2 minutes for 0.1%.
So it doesn’t impact anyone meaningfully for the site to be heavy. And for that edge case 0.1%, they’ll either leave, or stick around waiting and stab that reload button for as long as it takes to get the info they need.
As shameful as it is, web perf work has almost zero payoff except at the limit. Anyone sensible therefore has far more to gain by investing in more content or more functionality.
Google has done Google-scale traffic analysis and determined that even a 100ms delay has noticeable impacts on user retention. If a website takes more than 3 seconds to load, over 50% of visitors will bail. To say that there is no payoff for optimization is categorically incorrect.
The incentives are there. Web developers are just, on average, extremely bad at their jobs. The field has been made significantly more accessible than it was in decades past, but the problem with accessibility is that it enables people who have no fundamental understanding of programming to kitbash libraries together like legos and successfully publish websites. They can't optimize even if they tried, and the real problem for the rest of us is they can't secure user data even if they try.
This test was a while ago - it’d be interesting to see if it’s still the case and if the results reproduce. But still let’s consider that Google is Google and most websites are just happy to have some traffic.
People go to Google expecting it to quickly get them info. On other sites the info is worth waiting an extra second for.
At Google scale, a drop in traffic results in a massive corresponding drop in revenue. But most websites don’t even monetize.
They’re both websites but that’s all they have in common.
1 reply →