Alex Pooley's Blog

Why don't you subscribe to my blog while you're here? I'm a freelance web developer and I blog about Ruby, Rails, and business online.

Go ahead and subscribe to my RSS feed. Thanks for visiting!

Because of their dominance in the search engine business, generating traffic through Google is often a critical part of any SEO effort. Unfortunately, traffic from Google to your web site can make or break your efforts in building an information based web site. Recently, Google turned the steady stream of traffic to one of my sites to a mere drip. This site in particular was left for 13 months as a control. In other words, I put the site online and didn’t touch it or explicitly build links for 13 months. A couple of weeks ago, without any apparent reason, traffic dropped from over a 100 unique visits per day to single digit figures.

This real life experience has highlighted the fickle nature of search traffic. Here’s another graph of the site, from inception to June 17 2008.

When the site hit its stride earlier this year, it was generating over 330 unique visitors per day from Google.

Fortunately I don’t depend on that site to pay the bills.

So what’s the deal? How did the site go from zero to hero, and back to worthless in the span of a year? The truth is that I don’t really care. The fact that anyone or anything (Google, backlinkers, malicious SEO dude), can affect my traffic flow so dramatically, leads me to question the long term viability of the long tail business model.

An important attribute of this particular site is that it did not offer a very strong point of difference, and so it was not very sticky or viral. It supplied a great breadth of basic information, and was very dependent on search engine traffic. Hence, this site existed purely to exploit the long tail.

So here’s the deal. With virtually no marketing effort I was able to take a large (200,000+ pages) generic content site and build it to over 330 unique visits per day with virtually no marketing effort. After a year, the site came round trip and found itself where it started - at single digit hits per day.

My conclusion:

1. Don’t build a long term business around search engine traffic.
2. Do use search engines to test the human interest in your site. As opposed to bot interest.
3. Ensure diversity is built in to the marketing and sales plan.

Is it coincidence that about a month ago Google had dramatically reduced my Adsense income? It’s totally irrelevant whether my trouble with Google is local to my site, or global to Google. Human behaviour has stayed consistent (so I’m led to believe) for thousands of years, while artificial intelligence is always evolving. From today forward, I’ll be targeting human behaviours, and taking the search engines as just another marketing strategy, and never my primary marketing strategy.

My conclusion, and previous experience, leads me to build this rough check list of “things to do” for my next site:

1. Build a site with something of human interest.
2. Build a subscription base with feeds, mailing lists, bookmarks.
3. Organise private advertisements.
4. Build a marketing strategy without organic search as a primary traffic driver.

Time to think again?

(OK, so a handful of people seem to do nicely with it.. or so I read)

Actually, the screen shot above is a really bad example of what I see from day to day. The problem I have with Adsense is how unpredictable the earnings seem to be. I was “this” close to scaling my work after a couple of months of constant income, and then my earnings halved. WTF Google. WTF…

As you can see from the log scaled plot to the left, advertising demand on search terms grows exponentially with popularity. You can see for yourself in “Figure 1″ of this paper from Yahoo! Research Labs.

Using Wordze [aff] as a keyword data source, I plotted 800,000+ search keywords from 760 seed keywords and their corresponding demand to roughly determine if this advertiser demand is warranted. The log scaled plot is below with the horizontal axis as seed keywords ranked by descending popularity, and the vertical scale as the amount of queries that include the seed keyword.

[The reason for the large drop off in the plot is most likely because of my limited data.]

Searchers do appear to follow an exponential model where popular queries are exponentially more frequent than unpopular queries.

This so far does not say anything about the correlation between queries and advertiser demand. Specifically I would be interested to determine if the market efficiently demands popular keywords proportional to query frequency. As a guess, any deviation from a strong correlation would suggest either:

• A relatively good value keyword.
• A poor converting keyword where the market has already factored human behavior in to the model.

I suspect some markets will be more mature and so may already factor human behavior in to the model. This maturity would probably be proportional to the size of the market.

Speculation is fun but almost useless. I have some data laying around where I can test some of these ideas in the future. Stay tuned :)