Quick Analysis of Web Browser Stats
by Jon Peltier
Thursday, April 1st, 2010
Peltier Technical Services, Inc., Copyright © 2012.
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
In Chrome takes a bite out of IE and Firefox, Ed Bott describes how Chrome has started eating into the browser market share of both Internet Explorer and Firefox. Ed’s analysis looks right on, but he was rather light on data visualization. I figured I’d help out with some enhancements to Ed’s data presentation.
Ed looked at his web analytics reports of browser usage for roughly the months of September 2009, March 2009, and March 2010. He showed separate screen shots of the tabulated data, and a pie chart which I think was just decoration for his post.
The first improvement I’d make is to combine the data into a single table. For this amount of data, a table actually was sufficient for his conclusions.
The new table formatting introduced in Excel 2007 can be used to make attractive tables. I used manual formatting in Excel 2003 for this table.
Anyone tempted to use pie charts should pay heed. The first two pies below are essentially indistinguishable, and the data is nearly so. The third pie shows a large increase in Chrome and a smaller decline in IE and, but it is not possible without reading the percent labels that Firefox has also declined. So these three charts are not very effective.
How should we plot data by time? Where do you think a “time line” got its name? Use a line chart. At a glance you see all of the data, using a single pair of axis scales. It’s obvious that Ed looked at data at irregular intervals, first six months, then twelve. Probably not relevant to the final analysis, but good to show it. You can also see the increase in Chrome that corresponds to the decline in IE and Firefox.
What you also see, but which Ed didn’t mention, is that the increase in Chrome From September 2008 to March 2009 seems to match a similar decrease in Firefox. This wasn’t evident in a quick look at the table: For Chrome, the eyes see an obvious difference between 3% and 4%, because the first digit changes. But the eyes look at Firefox and say “37%, 38%, no difference” because the first digit is unchanged. The chart doesn’t care about digits, it plots the points according to the underlying values.
Let’s revisit the steps:
- Put the data together. Combine the numbers into a single table.
- Plot the data. Use an appropriate chart type.
- Let your eyes see what’s going on.
Something like 70% of our sensory neurons are located in our retinas, and 50% of our cerebral cortex is dedicated to analyzing the visual information received by these visual neurons. This awesome processing power is concentrated on graphical images, not conversion of text to data.
Related Posts:
- Web Browser Stats: Problems With Data Gaps
- SPC Approach to Browser Stats
- Bar Chart Value Axis Scale Must Include Zero
- Sales “Funnels”
- Spiraling Down the Drain
- Who’s to Blame for the Mess in Washington? Analysis of Chart Types
Posted: Thursday, April 1st, 2010 under Charting Principles.
Comments: 11
Comments
Comment from Jean Paul
Time: Thursday, April 1, 2010, 4:07 am
Hello Jon,
From September 08 à March 09, the growth of Chrome is approximately equal to the decrease of Firefox.
From March 09 à March 10, the growth of Chrome is approximately equal to the sum of the decrease of Internet Explorer and Firefox.
One can also see that for the first time Internet Explorer is dropped below 50% (first digit is changed).
Quick analysis but absolutely convincing
Best regards
Comment from Calvin Graham
Time: Thursday, April 1, 2010, 5:39 am
I find a stacked cluster works quite well in this situation. You have one column with IE and a stacked cluster of the other browsers. This technique works well for data series where one input makes up ~50% of the total
Comment from Jon Peltier
Time: Thursday, April 1, 2010, 6:02 am
Calvin -
Good idea. Here’s a clustered column chart:
and here’s the clustered-stacked version:
It doesn’t replace the line chart, but the drop of IE below 50% stands out, and it’s not hard to see that Chrome has increased substantially.
Comment from Ed Bott
Time: Thursday, April 1, 2010, 7:54 am
Very nice! Thanks for the visual assist.
One tiny correction: I did mention the change from Sept 2008 to March 2009. “The slight increase in Chrome usage was almost perfectly equal to the slight drop in Firefox usage.”
But that was buried in text, inadvertently making your point about visualizing this data even stronger.
Comment from Tony Rose
Time: Thursday, April 1, 2010, 7:59 am
I think that using the lines makes it a little bit easier to make out the slope. My eyes naturally try to scan across one segment to see if it’s going up or down. Using a line makes it that much easier, in my opinion.
I’m surprised that Firefox is down after 2 years. The lost market share between IE and FF appears to have gone directly to Chrome.
Only other helpful piece of information would be to show the % change from 08 to 10 maybe to the right of the chart in a table or besides the last data point.
Comment from Jon Peltier
Time: Thursday, April 1, 2010, 8:38 am
Ed -
Thanks for checking in. Sorry I missed your comment, but as you said, it was buried in text, and like most web users, I like looking at all the pretty pictures.
I’m going to follow this up with my own browser stats, showing every month to try to capture the variability in the data.
Comment from Nicholas Hebb
Time: Friday, April 2, 2010, 2:02 am
The thing that stood out the most to me wasn’t the charts, it was the dates.
I think we’ve entered an era where two-digit years can lead to ambiguity. I hadn’t thought of it until seeing the “March 10″ label above the pie chart. Without the context of the neighboring “Sept 08″ and “March 09″, I’ll bet most people in the US would see March 10 and assume the 10th of March.
It’s just something to keep in mind when picking a date format from now on.
Comment from Jeff Weir
Time: Friday, April 2, 2010, 2:18 am
Could be my mispent youth, but I find that reporting the table figures to two decimal places distracts me from metally juggling the numbers in my head. Or as you put it, they get in the way in the conversion of text to data.
I’d prefer one or no decimal places. Granted, with no decimal places it’s harder to see some of the movement in the smaller series, but really compared to the lager browsers there is no movement to speak of.
In fact, I think this does a better job than the graphs:
Comment from Jon Peltier
Time: Friday, April 2, 2010, 5:59 am
Jeff -
Good point. I used the same two decimal digits that Google uses to present the data. Removing the percent sign also cleans it up.
Whether the simpler table beats the line chart might be a matter of taste, but both are effective.
Comment from Jon Peltier
Time: Friday, April 2, 2010, 6:11 am
Nick -
Good point. We use two digit years to save space, at the potential cost of confusion. I’ll try to remember at least to use an apostrophe, as in March ’10. At least until 2032.
Pingback from Visualizing trends in browser usage
Time: Friday, April 2, 2010, 11:40 am
[...] Peltier takes the data I published last week on browser trends and gives it a visual makeover: Where do you think a “time line” got its name? Use a line chart. At a glance you see all of [...]
Write a comment
I welcome comments from my readers. If you have an opinion on this post, if you have a question or if there is anything to add, I want to hear from you. Whether you agree or disagree, please join the discussion.
If you want to include an image in your comment, post it on your own site or on one of the many free image sharing sites, and include a link in your comment. I'll download your image and insert the necessary html to display the image inline.
Read the PTS Blog Comment Policy.