When users encounter familiar conventions, they already know how to use them, they feel safe and secure, they can use the product (or the service) efficiently and effectively, and in most cases they also feel satisfied.  Breaking the conventions, on the other hand, can easily lead to significant usability issues.

That’s why UXers and human-computer interacton experts usually encourage the use of “UX Patterns” as much as possible to avoid creativity while designing micro-interactions, so as not to surprise users and confuse them.  Any professional that has ever conducted usability testing can indicate that deviations from conventions will probably result in a negative effect on system status acquisition, reaction time, and on satisfaction.  We also know that breaking conventions binds the user to learn a new pattern that sometimes competes with existing skills they’ve already mastered, and that learning in that case is expected to be slower.

But how extensive is breaking conventions, per micro interaction?  What is the performance “price” we pay when we utilize disruptive patterns?  Since I wanted to study this issue from a quantitative perspective, I developed a small experiment.

Experiment Goal

The goal of the experiment was to study and demonstrate the decrease in human performance when working with a non-standard pattern, with respect to reaction time and mistakes.  For that purpose, I wanted to measure the difference in reaction times between a well-established UX pattern and a slightly different pattern that was still reasonable and logical. I then wanted to compare the number of mistakes – at least the ones noticed and corrected by the user (and to some extent the unnoticed ones, as well).

The conventional pattern I chose for the experiment was a standard telephone keypad (or “dialpad”), and the disruptive pattern was non-standard (yet still logical) layout of the same keypad.

123 147
456 258
789 369
*0# *0#
Call Call
Standard Keypad Disruptive Keypad

Using a standard telephone keypad is a well established psycho-motor skill, and we can assume that it is already autonomic for most of the adults.  Once the standard keypad is presented, most people will use it naturally with minimal effort.

Breaking UX conventions can easily lead to significant usability issues

The disruptive keypad, on the other hand, is expected to consume more cognitive processing resources, since users are not accustomed to using it, and since using it competes with the existing pattern of the standard layout.  Using the disruptive keypad is expected to take longer, and more mistakes are expected to occur.

The Experiment

I composed a small webapp based on a jQuery keypad (client side) and on a MySQL (server side), on which the experiment’s participants (subjects) were asked to dial a ten digit phone number they can easily recall.  They were asked to dial the same number twice – the first time using a standard dialpad, and the second time using a disruptive one.

Prior to each of the dialpads shown above, there was a 3-2-1 countdown screen that served two purposes: (1) to make sure the user is ready to dial the number as soon as the dialpad is displayed and (2) to make sure the jQuery is totally loaded and functioning as soon as it is exposed.

The performance time for dialing the ten-digit number (from dialpad reveal to tapping on the call button) was measured, and the number of taps on the backspace button was counted.

Results and Discussion

The experiment was limited to Android users with Google Chrome browser, due to unexpected performance of the jQuery plugin on other platforms.  During November 2015, around 150 subjects participated the experiment by invitation via WhatsApp, in person and during my talk at DevFest Romania.  I had to eliminate some of the results due to easy-to-type numbers (like 12345…), due to input that was , and due to unreasonable performance time (that probably occurred as a result of technical issues).  The following table displays the statistics of 130 valid pairs of inputs.

 Performance time (seconds)Number of taps on [Backspace]
St. Dev.4.285.340.972.28

From the table above, we can see that the average time it took to dial the ten-digit number was 35% longer for the disruptive dialpad when compared to using the standard one.  The median time to dial for the disruptive dialpad was 10 seconds, which mean that 50% of the subjects took more than 10 seconds to dial the number. A closer look at the performance of the standard dialpad shows that only 20% of the subjects took more than 10 seconds to dial in that case.

Another important finding was that 28 (22%) of the number pairs didn’t match.  Assuming that subjects did succeed in dialing the correct number with the standard dialpad, we’ll have to conclude that the disruptive dialpad was so confusing that they didn’t dial the right number while using it.

Now, since the subjects of the experiment used their own devices, we could not avoid delays in the performance that occurred due to technical issues. For that reason, I Iet myself clean the data of the seven subjects that took more than 16 seconds to dial the number with the standard dialpad (16 seconds is about Average+2*St.Dev).  The following table displays the statistics of the 123 valid pairs of inputs that remained.

 Performance time (seconds)Number of taps on [Backspace]
St. Dev.3.305.460.862.34

From the table above, we can see that the average time to dial the ten digits number was 46% longer while using the disruptive dialpad, when compared to using the standard one. The  median time to dial for the disruptive dialpad was 10 seconds, which means that 50% of the subjects took more than 10 seconds to dial the number. A closer look on the performance with the standard dialpad shows that only 15% of the subjects took more than 10 seconds to dial in that case.

The following graph displays the experiment results sorted by the performance time with the standard dialpad.  It’s obvious from the graph that longer performance times with the disruptive dialpad is correlated with the number of taps on the backspace button, and indeed, statistics results here with a correlation coefficient value of 0.77.  There were 16 cases in which the performance time with the disruptive dialpad was better than the performance with the standard dialpad, but as can be seen on the graph, they all occurred with subjects that were above the average with the standard dialpad – most of them high above the average.  This can imply that these subjects either suffered from some kind of a technical problem or a distraction on the first part of the experiment, or that a few subjects are less sensitive to the keypad layout and they could even improve from the first part to the second one (learning curve).


From the above results, we can conclude that a non-standard keypad – that is not very different from the standard one in layout and in logic – can lead to an increase of 30%-50% in performance time, and can significantly increase the likelihood of mistakes – both noticed and corrected and unnoticed.

When users are accustomed to using a pattern, even a minor change in that pattern can be very expensive in performance terms.  In fact, the minor change is confusing in two ways: (1) since the difference is minor, it can go unnoticed, hence , and (2) since the user is already well-trained with the standard pattern, their knowledge masks the new pattern and inhibit learning.

Image of seamless triangle pattern courtesy of Shutterstock.

Article No. 1 326 | October 15, 2014
Article No. 1 247 | June 2, 2014


Most 'UX designers' haven't got a clue about user interface design - which is doubtless why they call themselves 'UX designers', not 'User Interface Designers', of which only a handful exist in the world today. Hence this website is full of meaningless articles about nonsense which any intelligent person would know the answers to without having to think hard at all. Most modern interface design is atrocious and actually makes computers harder to use, as MOST people are well aware. Look at Windows 8 - full screen apps, the Start screen, the 'Ribbon', etc.etc.


Its an interesting test and I agree that patterns exist for a reason - similar to interstate signage or roads....if those patterns were disrupted there would be chaos. 

However, in interface design just because its an established pattern does not mean its correct - we should continue to strive for a "better mousetrap" albeit released in a way that does not feel disruptive.

Nice article though. 

There is nothing innovative or radical on UI component if the component needs an explanation to be used in a right way.



As a researcher, I would have to agree with the aforementioned problems with this study. Evan's suggested "title" and following assertions are an appropriate critic of the article and study. Yes, it is easy to demo, but does that truly prove the point? I would recommend finding a more appropriate comparision, such as navigating with hover drop-downs vs. a drop-down that requires you to click once to open the list, or how the fact that you can't utilize hover menus on mobile sites affects the efficiency of navigating quickly through a complex site. Or, you could use my favorite, the hamburger menu. It's a design pattern that continues to be utilized, yet many users (esp dependent on the demo) don't even recognize it as a menu, and then what happens when you add a drawer to that. 

A better title for the article might have been: "The Price Of Not Using Non-Standard Telephone Keypads In Your App"

But taking this one "experiment" (which was obviously set up to confirm an existing bias) and trying to convince your audience that it demonstrates "The Price Of Not Using UX Patterns" is deceptive and pseudo-scientific (no matter how many graphs and data tables you want to include).

Separately, there's also a price to pay for being conservative and risk-averse... there are many situations in which you'd want to ignore a prevailing "pattern" and invent something new.

The statistical metrics look solid.  As a comparison, I would like to see this experiment run with other modes of "disruptive pattern[s]", such as arranging the numbers with 7-8-9 on the top and/or changing of alignment of the characters within their boxes.

A better test variation would have been a calculator keypad. It's a standard UI pattern used in a non-standard context. I hypothesize it would have provided more realistic results.

IMO this doesn't prove that using any non standard patterns has a performance tax. This tests seems purposely designed to trick a user, you even label it as disruptive. I get that it's a well established micro interaction but maybe its too micro?.

Real tests are often hard to create, but wouldn't it be better to test something that breaks a convention but is intentionally designed to be better and has been tested to be at least a good experience. Then we have two GOOD experiences to test. We could assume the convention became so because it is better, We got to this point because someone at some point challenged the status quo with a better idea.

Often we see non conventional interfaces that are just bad UX, It doesn't mean it was bad because it was not a convention. Eventually things around a convention change and relying on best practice becomes sloppy practice. I think we need to keep analyzing and trying to improve

A fun test to see though.

I did not try to prove that non standard patterns have a performance tax. That's a known fact. My purpose was to demonstrate it and to quantify it. As I explained, the disruptive pattern in as logical as the standard one, which make this example great to demo how our habits set the rules. There is no "good" or "bad" experience here, only habits.

And when a designer wish to use non-standard pattern, they do need to explain to themselves why that’s important and why is it better than the standard way. Then test it, and decide… Take it small steps :-)

I am sorry, but there is indeed a "good" and a "bad" UX here. 

The way we are trained to read is frome left to right, when we reach the end of the line, go back to the full left, one line under. reading per columns like you have in the "disruptive" keyboard is harder for our brain. Even more when you see the second column, from top to bottom "4, 5, 6, 0", which does not make any logical sense. There's been tons of research that goes in this direction saying that a 2 or more column layout is less performant, because your eyes go through a weird path. 

I would have love to see the same experiment with another kind of design like having the number as a circle (like old time phones for examples). 

Agreed!  Please share the results of your experiment, after running it.

I would also say that social context applies to the nature of the keypad. You might find for example that Chinese or Japanese test subjects would have less of a delay since the written languages they use can be read right to left or top to bottom giving less inclination towards the standard western/latin route of left to right.

This same kind of social context issue is something that plagues a number of different tests (IQ test are known for this sort of fallicy, such as having questions based around something only a select social group would have familiarity with).

As such just remember who your target market is when looking at such distruptive tests. (Windows 8 is also a good example. all of my non computer relatives love it over windows 7, the ones used to the old windows paradigm... not quite so much).