Comments on: Comparison of Search Share, September and October 2009

By: How Succinctly Can I Explain Why Pie Charts Are Evil? | Gilligan on Data by Tim Wilson

Wed, 02 Dec 2009 14:02:06 +0000

[...] catching up on some blog reading yesterday morning and seeing an excellent example of pie chart alternatives from Jon Peltier, and then watching seven presentations yesterday, six of which used the same basic presentation [...]

By: Ed Ferrero

Ed Ferrero — Tue, 01 Dec 2009 05:21:37 +0000

The numbers could also mean that Google and Bing users are clicking the search button more often because they can’t find what they want the first time.

Whereas Yahoo users…

By: Jon Peltier

Jon Peltier — Tue, 01 Dec 2009 04:04:30 +0000

So as search volume increases, Google’s and Bing’s shares increase. So people who are making more and more searches (are they more savvy?) are using Google and Bing. The people who are not increasing their searches are using Yahoo and (he-he) AOL. Or am I making too many assumptions.

I would have tried this panel chart as a line chart as well. No need to mess with shades of colors….

By: Ed Ferrero

Ed Ferrero — Tue, 01 Dec 2009 00:04:07 +0000

So, we are into analysis now. For the year to date data I would show the cumulative change in percent share from a reference date. In the chart below, I have picked Dec 08 as the reference date, so I am showing the change in share in Jan, Feb, etc.

http://www.edferrero.com/content/chart002.png

As it is the colours do not add anything to the chart, except maybe some confusion. I could have changed the colours from Excel’s defaults, but I am too lazy. Perhaps I will write some code to do this easily – Excel 07s nasty GUI makes it too tedious.

As Mike points out, the size of the search market has also grown during the period (Oct 09 total searches are 18.5% greater than Dec 08). We should probably show this on a separate chart to give a true picture. Again, I am lazy.

By: Jon Peltier

Jon Peltier — Mon, 30 Nov 2009 13:34:57 +0000

Mike - Thanks for providing the data for the whole year.

In fact, shortening the time axis helps show the trends more clearly:

By: Jon Peltier

Jon Peltier — Mon, 30 Nov 2009 13:21:50 +0000

Ed, Mike -

Thanks for the two good points you’ve raised: what’s the expected (or typical) variability, and how do we present this variability.

By: Mike Woodhouse

Mike Woodhouse — Mon, 30 Nov 2009 10:14:50 +0000

OK, ten minutes’ Googling and I have Oct-08 to Oct-09:

‘Month’,'Google’,'Yahoo!’,'Microsoft’,'AOL’,'Ask’
‘01 Oct 2008′,63.1,20.5,8.5,3.7,4.2
‘01 Nov 2008′,63.5,20.4,8.3,3.8,4.0
‘01 Dec 2008′,63.5,20.5,8.3,3.8,3.9
‘01 Jan 2009′,63.5,21.0,8.5,3.9,3.7
‘01 Feb 2009′,63.3,20.6,8.2,3.9,4.1
‘01 Mar 2009′,63.7,20.5,8.3,3.7,3.8
‘01 Apr 2009′,64.2,20.4,8.2,3.4,3.8
‘01 May 2009′,65.0,20.1,8.0,3.1,3.9
‘01 Jun 2009′,65.0,19.6,8.4,3.1,3.9
‘01 Jul 2009′,64.7,19.3,8.9,3.1,3.9
‘01 Aug 2009′,64.6,19.3,9.3,3.0,3.9
‘01 Sep 2009′,64.9,18.8,9.4,3.0,3.9
‘01 Oct 2009′,65.4,18.0,9.9,3.0,3.9

…which shows that Google have shown a steady increase, Microsoft Bing (and possibly the release of IE8) has increased their share and Yahoo! are suffering. All over the last year only. Even more interesting would be the total size of the search market over that period: is anyone gaining/losing in total number of eyeballs?

Lazy reporting/blogging, worse charting.

By: Mike Woodhouse

Mike Woodhouse — Mon, 30 Nov 2009 09:53:16 +0000

I wonder what we’d see if we looked at twelve months’ data? Are the Sept-Oct changes significant or are they within the variance for a longer timescale. absent any other information, I have to assume the latter, especially considering the marketing source of the “data”.

Must have been a slow news day.

By: Ed Ferrero

Ed Ferrero — Mon, 30 Nov 2009 07:48:00 +0000

I would have said that Google and Microsoft both increased their share by half a percent, and Yahoo dropped by 0.8%.

So something like this chart

http://www.edferrero.com/content/chart001.png
would be more appropriate rather than your last chart. As a rule, I am wary of charting ‘acceleration’ measures, i.e. the percent change in a percent measure.

By: Daniel

Daniel — Mon, 30 Nov 2009 07:38:59 +0000

…the last column chart with the relative changes in percentages is rather misleading, absolute changes in percentage points would be much more useful!