Time slicing Google Insight queries
Yesterday I wrote about the use of Google Insight to see how widely products had reached in the US. Inspired by a recent post by Erik from Seattle
on the growth of Twitter from month-to-month, I explored the fact that
the Insight product actually has data from as far back as 2004! This
means you can try out timeslices of data on different websites, before
they were successful, to see if there were common growth patterns as they grew in userbase.
For example, you can look for the search term "youtube"
and see what states were active in that search in Nov 2005, Dec 2005,
and so on, until the entire US is saturated.
Interestingly enough, the data actually shows there are many different patterns of growth – I show two below, based on YouTube and Webkinz. Everyone knows what YouTube is, but if you need a refresher on Webkinz, here's the Wikipedia entry.
Before diving into this though, I have to repeat the caveat from my
previous post. First off, this is not traffic from different states,
it's searches. You have to make the assumption that more searches
correlates to more traffic, which sounds reasonable to me but might be
completely wrong. Also, obviously these searches skew toward Google,
which has its own bias as a search engine. And finally, I don't know
how the search index really gets computed, but I'm assuming higher
search index means more searches.
Now that we have that lawyerly disclaimer out of the way, here's the data…
YouTube
Nov 2005: As expected, YouTube starts out in
California and New York, which are two of the more digitally inclined
states, and also the most populous. The site was launched to the public
in late 2005, and as you can see from these next couple pictures, it
grew very very quickly.
Dec 2005: Next, it spreads to other populous states, including
Florida and Texas. One observation would be that because YouTube was
generating lots of traffic from MySpace and search, it'd be logical to
think that they'd have fairly "general" audience growth and follow
whichever states had the most people.
Jan 2006
Feb 2006
Mar 2006: Starting to fill in more to the central states. Note that
it's light in the south even though there's a TON of people there –
perhaps those were the YouTube late adopters?
Apr 2006
Sep 2006: The final month I'm showing is about 2 years ago, where it
was able to achieve widespread adoption throughtout the US. Note that
Hawaii is very dense, along with California. If you look at the Insight
data now, it basically has not changed much since Sep 2006.
WebKinz
Mar 2006: Webkinz, unlike YouTube, started out in Massachusetts. Did
you know that the kid-centric MMOG shared the same birthplace as
Facebook? The Wikipedia entry for Webkinz
lists the starting date for the company as Apr 2005, so it took them a
year to register enough searches to show up on Google's tool. Note that
since the MMOG is tied with a plush doll, much of the expansion is
probably related to actual product distribution and how they rolled the
product out. Similarly, it's been noted that "viral channels" don't
really exist for kids – they primarily learn from each other via word
of mouth, since few are reading blogs or twittering ;-)
Apr 2006: Unlike the YouTube example, Webkinz doesn't immediately
take off in Texas or California. Instead, it makes a slow expansion
into New York and Florida. (In this set of pics I'm skipping more
months since the growth is so much slower than YouTube, btw)
May 2006
Sep 2006: The growth continues westward, with Illinois, Michigan,
Ohio, and a couple East coast states. Note that California finally
shows up with some searches.
Dec 2006
Jun 2007: Much later, you see that the Webkinz phenomenon is still
mostly East Coast focused, and doesn't involve California much even
though it's a very populous state. Strange!
Conclusions
I think there are a couple interesting observations to be taken out of the two, in concert with each other:
- Different products get adopted differently, and follow different patterns
- Horizonal product like YouTube and grow quickly across the US, and hit the most populous states first
- Products like Webkinz, on the other hand, have very distinct signatures and might have to do with the fact that it's
tied to a offline product and/or the fact that the customer base of
kids doesn't have many consolidated online marketing channels - Similarly, the actual underlying growth of some products can be
very fast, with state-by-state growth happening over weeks or single
months, or sometimes the distribution doesn't change much over many
months - Webkinz is also an interesting example of a non-California
centric product that is successful elsewhere. Perhaps MySpace is
non-Silicon Valley, as another observation.
As always, suggestions and comments welcome!
If you end up
doing this analysis for other sites and find other patterns, please
comment and I'll link you to the bottom of this post via UPDATEs.