More statistics on user clicks and browser settings

by Dan Zambonini

Following on from Plotting the exact X/Y coordinates of clicks on a page, we've now aggregated data from a number of similar websites (including the Imperial War Museum). As mentioned in my last post, we've been collecting the X/Y coordinates of clicks, together with other relevant data (including: screen resolution, text size settings, user agent and "time to click").


I've been looking for patterns in the data, and although there isn't much to report, I thought it would be interesting to share what we've found so far:


Text Size statistics


I was surprised by the number of users who choose non-default text size settings in their browser. Although this may be due to a selection effect (this data comes mainly from Museum websites, which may have a particular audience), the figures are still surprising:






























Font Size (IE) % of users
Largest 4.1%
Larger 2.7%
Medium 88.9%
Smaller 3.1%
Smallest 0.9%


(give or take rounding errors)


With regards to 'time to click', I checked for a pattern, but found that there is no strong relationship between Text Size and Time to Click. I had hoped to find that users with larger text sizes took longer to find what they were looking for (either because there was less on the screen, or because they tended to be older, less experienced users), but found no evidence to support this.


I say 'strong' relationship in the above statement because there was a little evidence: users with 'Medium' took (on average) 12.4 seconds to click, 'Larger' took 14.4, and 'Largest' 13.7 (for interested readers: Smallest = 14.5 and Smaller = 13.6). Although you could argue that non-Medium text sizes took longer to click, the difference is so small that a much larger data sample would be required to conclusively prove this.



Screen Resolution statistics


I was expecting to find a trend of some sort, but found that there is no relationship between Text Size and Screen Resolution. So, a user with a resolution of 640x480 is just as likely to use 'Largest' text size as a user with 1280 X 1024.


Using data from a single, 'tall' site (in design), I also checked the y position of clicks against screen resolution. Again, I had a preconceived pattern that I was hoping to detect - that users with larger resolutions were more likely to click further down the page. However, my findings seemed to prove the opposite (although again the figures are so close that we can't conclusively prove anything):






















Resolution % of clicks under y=768
800x600 7.6%
1024x768 7.0%
1280x1024 5.9%



Browser statistics


I'm not sure what I was hoping to find, but a comparison was run of Browser against average "time to click":






























Browser Time to click (seconds)
IE 5, PC 14.7
IE 5.5, PC 13.1
IE 6, PC 12.6
FireFox, PC 11.8
Safari, Mac 11.7

So, the obvious conclusion is that Safari is the most usable web browser! Seriously, though, I can't think of any browser-related reason why users of one product would be able to 'find their link' quicker than users of another browser. Perhaps FireFox and Safari render the pages sooner than the 'onLoad' page event (which is when we start recording for 'time to click') than in IE. Or maybe IE users are just plain dumber, so take longer to read and find content. I am an IE user, so I think that's hard evidence to back this hypothesis.


7 Comments

jim_mac
2006-03-14 07:39:11
Validity?
Not to get overly analytical, but it would be interesting to know the sample size you used for your data analysis. You quote your stats down to a tenth of a per cent. Was the sample size large enough to validate the use of that level of accuracy?


Better yet, you should perform some statistical testing to check if the differences you think you are seeing are significant or merely due to chance. Merely compiling some stats and "eye-balling" them is worse than useless, as it can be quite misleading!


JIm

Dan_Zambonini
2006-03-14 07:46:57
Re: Validity?
Hi,


Most of these stats were created from 25,000 data points. The only exception is the 'screen res' section (where I say I limit it to a single site), where the sample size drops to about 15,000.


Yeah, the statistical significance would be useful, wouldn't it? I'll see what I can do...


Thanks,


Dan

Erich Schubert
2006-03-15 04:43:35
I have suggestions how to explain the last two statistics.


First the clicks below y=768. Pages usually have most links at the top. However, when a user scrolls, he will not see these links any more, and thus maybe pick a link on the lower part of the page.
Users with a lower screen resolution are more likely to have scrolled so far that they don't have easy access to the navigation at the top any more, instead of scrolling *back* they pick a link in the lower part more often.


As for the browser time-to-click stats: it's not that much a question of usability, but also of rendering speed, rendering stability (you can only click when the page has "stopped moving") and of course, audience. Firefox users are probably power users. Users with IE 5 never even upgraded their IE...
I can't really make up my mind on Safari, though. Doesn't strike me as power user. But maybe their clock starts when the page is done loading, whereas other timers start earlier?

Derek O'Brien
2006-03-15 10:53:47
I pretty much agree on what Erich Schubert has mentioned above.
I would say that the time between the onload event and the content fully loading can differ dramatically between browsers.
FireFox was built with a superior rendering time compared to IE. The IE stats only go to show the improvements over version releases and Mac stats probably show the stability of the underlying OS the browser is running on. Which does not surprise me as the Mac was designed to run smoothly for graphics use and most rendering time within browsers is taken up calculating image rendering.
Robert Drózd
2006-03-16 11:57:15
Hi Dan, thanks for these numbers. I wonder how did you get the information about non-default text sizes, I thought this was not possible with JS. :-)


Regarding "time to click", it would be good to compare distribution of these times for different browsers (particularly IE and Fx/Safari).


Maybe IE users split into two groups: "normal users" with time of about 12 seconds, and "newbies" (they mostly use IE, don't they?) with time that was much longer?

kirk
2006-03-21 08:00:32
Under the browser stats it may not be that Safari is the most usable browser but may be an indication of the skill level or technical ability of the user. If you look at the stats, the "new" browsers, Firefox and Safari, and the most up to date version of IE demonstrate the best click times. The people using these are more likely to be "up" on the lastest trends and skills, while someone who continues to use IE5, well, they are more likely to be a little "slow" on the tech curve.
Laurent
2006-04-06 15:16:20
If you re interester, i conducted a study about real window size vs screen resolution statistics ... it was interesting to see how user use their desktop space.


window size distribution stats have been published here
http://www.gwix.net/web_gwix/article.asp?filID=144