Guessing and empirical data as a basis for design principles

absrtact : Even less empirical data, such as observing 2 users, can greatly improve the chances of making the right UI design decisions.

Provides the ability to adjust the font size to the user, or simply rely on the browser's own features? This issue has recently been raised in an interactive design discussion group.

12 people responded to the question. Most people simply talk about the idea of solving the problem. It makes sense: Everyone is an expert in their field. But there are 6 posts on what is best to express their views.

Of these, 2/3 of the threads are purely imaginary, while the other 1/3 are based on user-observed empirical data.

guessing faction :

• "Now in this age, [...] Most people are familiar with how to make fonts larger in browsers. " wrong
• "Users who need to adjust the font are implemented through the browser; it's not difficult."
• "It's not 95," he said. "Not all users over the age of 50 are like amateurs, and don't know or want to know how to adjust the font size in the browser. " wrong
• "The majority of users who need to increase their fonts are older people over the age of 65, who are often the least able to adjust their settings."

Data Faction :

• "I had to set it up for my parents manually, but for more than 65 years old, more and more abilities, such as the ability to adjust the size of the font of such difficult to find features, are at an alarming rate away from them." " to
• "I ' ve observed usability studies on sites" included text resize widgets [...] Most, if not all, the participants [...] had no idea what it is. " Right
• "I've done usability research on sites, including text-resizing controls [...] In most cases, participants [...] There is no concept of it. " to

The data faction is winning the guessing pie

The guideline is roughly this: use a font of relative size to make it easier for the user to adjust (if they know how to do it), but the font should be large and legible in the default state. Based on a large number of observations that many elderly users do not have the ability to adjust the font size.

The case in our discussion group,

• The designer who provided the extra data got a 100% correct rate.
• And those who rely on individual subjective judgment of the designer the correct rate is only 25%.

Most striking of all, 75% of all guesses are wrong. That is to say, advice to these people is better than tossing a coin to get the answer.

In this simple case, the correct rate is doubled based on the design recommendations for the minimum number of real-world experience observations.

Beware: Although getting data from parents is better than no data, I do not recommend that you build your decisions on family members because they are usually smarter than the average user. (because you are one of the wise people who know the usability knowledge.) We have learned from our studies of children and adolescents that, as opposed to, children and adolescents in general are more likely to encounter more difficulties in using the Internet.

Two user tests are also better than guesses

Although our case for font size is based solely on a very small set of data feedback. But another set of cases based on more data feedback confirms the same conclusion.

We tested two different ways of showing bank account information, each of which was completed by 76 users, and a total of 152 testers were involved in the benchmark test. We ask users to complete tasks such as checking their account balances and finding out the current level of interest rates offered by the bank. The results are as follows:

Usability Standards	Design A	Design B
Success rate (through four tasks)	56%	76%
Time to complete four tasks (minutes: SEC)	5:15	5:03
Subjective satisfaction (1~5 level, 5 is full score)	2.8	3.0

Of all three availability attributes, version B has a better design score, although only a statistically significant difference in the success rate. But in general, B is superior to a is beyond doubt.

(compared with the above research, sometimes different designs win or lose on different usability aspects.) For example, a design might give the user a better chance of achieving the goal, while the other is giving the user a hand at the speed of completion. In many cases, you might want to make a trade-off, or, where possible, you need to create an extra third design to combine the advantages of the first two. ）

In this case, I'll design A and b for the 21 participants who are in the interactive design class and ask them which option they will choose for the bank. The probability of getting the best design advice is 50%, relying entirely on guesswork to choose who is the best design. It's not as good as flipping a coin. (Believe me, a convenient way to save on consulting fees is to ask for coins in your pocket.) ）

I then asked another group of 38 students on the same course to test--two designs, each with two users. Sure enough, the success rate was 76%, depending on the experience of the 2-user behavior.

Looking at the result from another perspective, only 2 users in each design saw the error rate drop from 50% to 24%, a full half. Of course, if you're talking about a high ROI design decision, the 24% error probability is passable, so in this case it's clear that we need to test more users per design (I usually recommend 5 users). ）

Although this is small to no small study, but in each design test 2 users, far more than toss a coin-like speculation, greatly improve the quality of design recommendations.

(In this study, two versions showed similar importance to measurement research.) If you compare a rough prototype with a complete effect chart, you will tend to have data. ）

When the guesswork is wrong

Comparing our two case studies, we guessed that the performance of one side was poor in the case of text size. If someone makes a design decision based on these conjectures, the 3/4 decisions made will be wrong. In the case of the bank, they were half wrong.

So, why do these poor discussion groups speculate? The answer is in the following two sentences:

• "It ' s not 1995 ..."
• "It's not 95 years ..."

Unfortunately, too many web designers reject the belief that usability research will work for a long time. The idea that the difficult things of the past must be much simpler now leads to the destruction of many websites.

When we really study real users, we find that the speed with which users learn technology is so slow that the ability to use fancy web users is so small. In addition, most importantly, we find that there are so few users who are interested in learning many fancy features. Users just want to log on to the page, get what they want, and then exit. They didn't want to learn.

The wrong reason for guessing is that many designers want to believe in the potential of advanced design without hesitation. They would not have imagined that their beloved skill was highbrow and highbrow.

(yes, in recent tests, we did find that users have made some progress in skill mastery, but progress is slim; In the next more than 10 years, you'd better continue to believe that simplicity is the key to winning.) ）

A little bit of data helps a lot

In the two examples I've listed, even with very little empirical data, the likelihood of design correctness has increased dramatically: whether it's your loved ones or 2 user tests.

Of course, a little more data is definitely better, but no data is better than nothing. (ask yourself) How many design decisions have you made when you don't have any empirical data on the user's behavior at hand?