Friday, August 18, 2006

Alexa as a Research Tool


Most readers are familiar with Amazon's Alexa website. It tracks Web traffic at various sites and offers rankings as to the popularity of those sites. Thus, a site that has a ranking of 100,000 is the 100,000th most visited site on the Web. The Alexa toolbar can be downloaded so that you see all the rankings of sites you visit (as well as user reviews; Alexa rates Websites like Amazon rates books). It also enables your visits to register with the Alexa rankings.

A little while back, Random Roger passed along a post regarding Alexa as a research tool. The idea struck me as interesting. Alexa, to be sure, is not a perfect methodology, as James Altucher recently noted with respect to the Alexaholic site. One site owner passed on to me some strategies for manipulating the rankings by bombing target sites with visits; I also know from the relative rankings of my two sites that the ratings imperfectly correlate with actual logged visits. Still, the growth in visits in my sites has correlated with rises in the Alexa rankings, leading me to believe that the methodology has some merit.

Above, for example, we see the long-term Alexa rankings for the Realtor.com site. There is a cyclical pattern to visits, but not a significant downtrend as one might expect, given the weak residential real estate market. Visits definitely tend to wane in the last quarter of the year, perhaps reflecting the hesitance of families to move after the school year has started and winter is under way. Mid 2003-early 2004 represented a high water mark for visits; late 2005 through 2006 has been relatively weak. One possibility: it's a glut of supply and not a significant waning of demand that is weighing on the housing market.

Dell's site shows a different cyclical pattern on Alexa, with visits spiking toward the end of the year. This, presumably, is attributable to holiday season gift buying. There is an overall positive trend to the visits from 2002-2006, with big spikes at year's end in 2004 and 2005. Thus far, 2006 visits have dipped below the levels seen in 2005. One research application: monitor Alexa carefully as we approach holiday season to see if the expected spike materializes.

Still yet another application: Not only has there been an exponential growth of trading blogs, but the Alexa traffic to those blogs has expanded significantly. Go to Alexa and check out, for instance, the traffic to the Seeking Alpha site. Like many established sites, it went through a parabolic rise into early 2006 before settling back a bit. This may in part be a function of the Alexa methodology, which presumably (like Amazon) weights recent visits more highly than past ones in generating rankings. Nonetheless, the growth across multiple blogs is impressive: this is becoming a major form of publication in the financial world.

Take a look at the list of market blogs included in Ticker Sense's poll. I'm in the process of assembling a composite Alexa index that includes many of these sites, so that we can track overall traffic to the trading blogosphere. Will traffic vary with market conditions? Will jumps and declines in traffic correspond to increased/decreased public participation in the markets? These are a few of the issues that we might investigate with Alexa, as well as with blog search engines such as Technorati.

6 comments:

Michael said...

I just wish there was a better way to measure traffic across so many different sites. Relying on the small group of users who have installed some toolbar has serious drawbacks.

Brett Steenbarger, Ph.D. said...

Hey Mike,

I completely agree. This is a rough approximation at best.

There is a better way, although it would take such effort. It would be for market bloggers to form a "user's group" and openly share with each other hit statistics and other relevant metrics.

This is what happens among trade associations, for instance, when they need to get a handle on industry-wide trends. A "trade association" of market bloggers might be worth pursuing for collegial reasons, as well as informational ones.

Brett

Michael said...

Probably the most feasible would be for everybody to make their SiteMeters public. That's how TruthLaidBear does their rankings -- http://truthlaidbear.com/TrafficRanking.php?start=1

I don't see it happening though...

Sienna said...

I like using Google trends and related keyword research tools to plot the interest levels in different stocks.

For example.. look at the search data results for Microsoft's ticker and compare with it's chart:
http://www.google.com/trends?q=msft

Notice the similarity in both charts for this year. I don't know about anyone else, but I see a clear connection.

I actually want to spend some time plotting and finding the relationship between the search data charts and the real stock charts. This is muddy water however because you have to determine whether that interest is positive or negative. Of course, the events surrounding spikes say alot and may be a good indicator of the true impact of these events.

As for Alexa, it really is quite flawed not just in user base, but in the fact that it is so easily manipulated. With just a click someone can deliver a million 'hits' to their site, which Alexa records, without one single person every visiting.

Brett Steenbarger, Ph.D. said...

Hi Mike,

Thanks for the pointer to TruthLaidBear. I'm not sure everyone would have to make their Sitemeters public; they'd just have to make them available to a designated person or persons for aggregation. Again, very similar to the way companies will share information within a trade group, but keep individual data confidential.

For instance, if a group of bloggers wanted to join forces and go after advertising dollars or publishing deals, there would then be an incentive to document the growth of blog traffic over time. Generally, in the absence of those kinds of financial incentives, data sharing tends not to occur, as you suggest.

Thanks for the perspective--

Brett

Brett Steenbarger, Ph.D. said...

Hi Fugue,

I think you're right; the search engines may be a promising line of research in themselves. It may be the case that Alexa is too flawed to be of value; that's something I hope to put to the test. What I can say re: my experience with Alexa to this point is that it is more accurate in an ordinal way than a cardinal way. That is, it has pretty accurately tracked the growth of traffic to my sites, but has not always gotten the specific magnitudes correct.

Thanks for the note--

Brett