I started writing the Alertbox column about user experience in May 1995. Now, 25 years later, we havepublished 1,330 UX articles— all for free and available to the world-wide UX community. While these numbers average to one article per week, the actual publishing frequency has increased from one per month in the beginning to two per week in recent years.

I thought it was worth looking back at the early articles to see how well they have held up to later developments in the fields of UX, computers, and the Internet. (Early articles focused mainly on web usability, whereas later articles have ranged widely across the UX spectrum.) As a concrete exercise, I reread my 250 first articles, written from May 1995 to December 2001.

(For a different take on UX history, see themanbetx官网手机登陆1998 - 2018年的文章是最受欢迎的with readers each year.)



  1. 需要速度(1997)。Response times are crucial for web usability, and sites with fast loading times have high转换率。仍然像以往一样真实。不幸的是,当我们revisited the page-speed issue in 2020, we found that website response times are still pretty bad for desktop users (at 6 seconds) and are unacceptable for mobile users (at 20 seconds), given that the响应时间建议are the same as ever: 1 second for great usability and 10 seconds for the longest acceptable wait.
  2. How Users Read on the Web(1997). 我们的研究表明,web写作的主要指导方针保持不变关于人们如何在网上阅读的最新跟踪研究
  3. 搜索你可能会发现(1997)。As I said back then (a year before Google was founded), “Search is the primary interface to the Web for many users. Search should be global (not scoped to a subsite) and available from every page; booleans should be made intimidating since users usually use them wrong.” All still true.
  4. Why Advertising Doesn't Work on the Web(1997)。Banner blindnesshas been confirmed again and again over the decades, including in our recent eyetracking studies. Worse, evenuseful page elements are ignored if they look like ads
  5. Nielsen的互联网带宽定律。带宽每年增长50%(1998年)。我的原始分析基于15年的带宽数据,从声学调制解调器到ISDN,但模型也适合未来22年的数据。特别是结论带宽呈指数级增长,但摩尔定律所特征的计算机电源背后滞后仍然有效。
  6. 网络用户越来越保守(1998)。People want it plain and simple and don’t care about the latest technology, leading to the recommendation to be a few years behind the curve for optimal usability. Still the case.
  7. WAP Backlash.(2000)。1999年,我说WAP应该代表流动性错误的方法,第二年我们得到了一个可怕的结果早期WAP手机的可用性研究。Yet, all the big phone companies were huge WAP promoters and strongly criticized our research. But I was right: WAP phones never took off. (I was also right in 2001 when I asked “手机:欧洲下一个MINITEL?“由于当时的移动电话主导地位,欧洲将导致移动互联网的当时普遍存在的叙述。相反,我在1997年被我站在了“Voice Is Just Another Datatype。” My 2001 analysis ofJapanese mobile products在gre是正确的at expectations for mobile photography and definitely right in praising high-quality Japanese screens with better pixel density than the low-quality screens of American and European mobile devices.)
  8. 新设备AGUR体面移动用户体验(2000)。Mobile devices should dedicate their entire surface area to the screen (not a keyboard) and embrace a deck-of-cards form factor. 7 years later (2007), the iPhone finally made this design recommendation happen. [However, the title of that article was too optimistic: yes, wefinally have decent mobile UX,但这并没有我在2000年预期的那么快。]
  9. Flash 99% Bad(2000)。Flash软件本身很长时间来自大多数网站,所以在那种感觉中,我被证明是正确的。但以一种更基本的方式,这篇文章仍然是相关的,因为大多数网站animation continues to be bad可用性。对,animation can be used for good UX, but rarely is.
  10. Avoid PDF for On-Screen Reading(2001)。正如我在2003年的关注文章中所说的那样,在线阅读,PDF is the monster from the Black Lagoon。Our最新研究has basically the same conclusions, though the technical problems with PDF have declined over the years.

2000年,我发表了这篇文章为什么你只需要用5个用户测试。但是,我不包括在1995 - 2001年的上述10位最重要的洞察列表中,因为它是基于我的早期工作折扣可用性从1989年到1993年。尽管如此,small-N用户测试的观点仍然被证明是正确的,任何做定性可用性测试的人都发现了同样的事情:在第一批用户之后,你对你的设计没有学到多少新的东西。最好从事iterative designto improve the user interface and then test it again with a new handful of users.

7 Biggest Fails

The following are my 10 most egregious failures from the 1995–2001 period: things I predicted but that definitely didn’t turn out the way I thought.

  1. The Death of File Systems(1996). File systems didn’t die, and files and directories are still the main way in which personal information is organized. One point remains as true as ever: that users are bad at generating good file names and have difficulty recognizing a name and remembering what it stands for at later times. Local search has emerged as the main solution to this naming problem. Some information spaces use different organizational schemes — notably, camera rolls and social media walls are simply sorted by time. The deeper problems in personal information management remain unsolved.
  2. 1997年网络发展趋势(1997)。语义数据编码没有发生!网ite data is as stupid and unstructured as ever. Instead, there’s now a trend to use AI to interpret natural-language information on websites and aggregate it on outside “scraping” services — sometimes causing hilarious mistakes and sometimes generating useful content. Unfortunately, it’s unrealistic to expect either individual users or even professional business operations to invest time upfront to structure data in order to facilitate later reuse. People always focus on the immediate use case and neglect the future. (My second 1997 prediction, “taking the web seriously as a business tool”, did happen. It’s interesting that the web was so unprofessional in the early days that stating that things would have to change was a provocative prediction at the time.)
  3. 在网上的忠诚(1997)。I said that the cultivation of loyal users was one of the beat ways of growing business value for a website, and I still believe this. However, very few sites have loyalty-supporting features like the ones I predicted.
  4. Do Websites Have Increasing Returns?(1997)。I predicted that the biggest websites would be huge (which did happen) but that the cumulative value of smaller sites would still be bigger. While some small sites definitely have disproportional business value, it has unfortunately turned out to be possible for web-wide aggregators to skim off more value than they deserve from the work spent creating smaller websites.
  5. The Reputation Manager(1998)。I expected the web to be heavily based on automated ratings of how well various sites and services had served users in the past. Search engines do sort results based on a form of reputation ranking, but that is not a true user-derived score. The main home for reputation managers is currently on marketplace platforms like eBay and Amazon’s third-party sellers, which are ranked by user feedback, and only the brave (or reckless) dare place an order with somebody with too few stars.
  6. The Case for Micropayments(1998)。I expected the web to evolve toward a model where users pay for service, but mainly on a per-use basis instead of subscriptions. Micropayments would have supported a wider and more diverse set of high-quality content services than the mainly advertising-dominated lower-quality services we have today.
  7. 请求营销(2000)。人们应该能够告诉公司他们感兴趣的东西,然后公司将在可用时销售这些产品。20年后,此功能几乎不存在。通常,当前系统对未来面向的互动提供糟糕的支持,用户在以后对事件表明兴趣发生。请求营销的主要示例是许多电子商务网站确实允许用户在储存备用项目可用时会要求通知。这一想法可以做得更多。



Right When Conservative, Wrong When Optimistic

Looking back at the 10 wins, 7 fails, and 1 half–half prediction, I was mainly right when I was being conservative: when I relied on users being resistant toward fancy technology and preferring simplicity. It’s striking how many of these long-lasting insights were based on a small base of qualitative user research. You don’t need a lot of users if you really look deeply at what they’re doing and understand why.

Conversely, I mainly failed when I was being optimistic and expecting big changes. Complex systems, such as the internet and the business ecosystem, are hard to move and often resist improvements that seem valuable but are too hard to make happen.

My most aggressive prediction of all, “In the Future, We'll All Be Harry Potter,“来自2002年,超出了本分析的范围。这也很早就告诉它是否会发生。但我在这种情况下保留了一点乐观主义 - 这对我来说肯定更加安全,以便在“未来”中而不是“10年”。特异性容易使预测错误,即使一般推力是对的。