• Register
• Log in

SEO Theory - SEO Theory and Analysis Blog

Algorithm analysis, Web community relationship analysis, SEO practices and techniques, industry news, etc.

• Home
• About SEO Theory
• SEO Glossary

Using free analytics data for competitive analysis

Posted by Michael Martinez on September 17, 2007 in Competitive Analysis, SEO Metrics

Free Web Analytics Providers

1. Alexa
2. AttentionMeter
3. Compete
4. Quantcast
5. Rankings

There have been times in the past when I have spun cautionary tales about using Alexa data for comparative analysis. Many times.

I beat up on Alexa mainly because one marketer likes to bleat about his Alexa rankings. I sometimes wonder if he has made the connection between his proselytizing for Alexa to his customer base and his Alexa rankings.

It is really that easy to influence your Alexa rankings, although I’ve seen other people order their employees to install the toolbar on their PCs (at work and home), write software that simulates the Alexa toolbar, and do other sneaky things in order to improve their Alexa rankings. It’s a metric that some people value and therefore it will be subjected to manipulative tactics.

Any traffic data that comes from downloadable toolbars is suspect, not only because it can be spoofed but because the legitimate data only comes from a small subset of Web users who tend to be both technically savvy and marketing oriented. That’s not very representative of the Web population as a whole.

Some Internet Service Providers sell user click data (hopefully anonymized) to services like Alexa and Compete. They can therefore at least compare their toolbar data to their ISP data and make an effort to normalize their trends. But it’s my understanding that not every Internet Service Provider sells its click data — and what about all those major corporations whose busy little office bees are surfing through company servers and connections?

AttentionMeter aggregates data from Alexa, Compete, and Technorati. I don’t see the connection between Technorati and the other two services but at least it’s a little different.

Quantcast can’t seem to find a Web site smaller than MSN, CNN, or Google. But perhaps I’m asking too much too soon of them.

Rankings will tell you a number but you have to write them down if you want to do competitive analysis.

None of these sites will provide data for sub-domains or sub-directories. They offer only domain-level reporting so you cannot look at the deep Web with these analytic tools.

When you need to do a competitive analysis, you’re pretty much stuck with these five services. You may be offered data from Google Analytics, SiteMeter, or some other third-party hit-counting service. Don’t take it. Not only is that data not useful for competitive analysis (unless you can get it from all the sites you’re comparing), the third-party solutions rely on either Javascript or remotely hosted CGI tools that just don’t capture all the data.

When you’re doing competitive analysis you need to compare similar data sets to each other. If you mix in foreign data structures you skew your results so badly they are useless. If you’re asked to do a competitive analysis for a sub-domain or directory-level account, pass. There is no way to do a competitive analysis for such sites with any measure of accountability.

Accountability is extremely important. In Web analytics, I need to see how the other guy puts his data together. If he is skewing resuts, I want to know. If he missing data, I want to know. If he is leaving something out, I want to know. In short, I wouldn’t trust another SEO’s Web analytics reporting farther than I could throw unless I see two things: a list of sources that I can get to without having to pay anyone money and a clear explanation of how the sources were brought together.

Your competitive analysis has to be bullet-proof. It has to be so simple that a complete moron could understand it. Snapshot analysis is always more informative than “in-depth” analysis. The more in-depth your analysis becomes, the higher I want my boots to rise.

When you’re doing a competitive analysis you have to look at trends and trend groups. I’ve talked a little about trends in the past but I haven’t really said much about trend groups. A trend group is a group of Web sites that you feel match each other well. You didn’t find them algorithmically (at least one analytics service provides algorithmic trend grouping — avoid it like the plague).

Who goes into that group? Whomever you want. They can share topics or not share topics. They can be in the same country, language, or city or not. They can be blogs or static sites. The only limitation to creating a trend group is that you have to define teh criteria you used to put the group together. Simply pulling names out of a hat doesn’t work. You have to use at least one common characteristic, although you can be somewhat flexible.

For the best competitive analytics, you want two or more trend groups. That gives you a stronger frame of reference. You compare the sites within each trend group to each other and you compare the trend groups to each other.

Site-to-site, group-to-group.

Resist the temptation to move sites around. Resist the temptation to compare site performance to group performance. Resist the temptation to do anything other than quick, simple analysis. The quick analysis will tell you more in a heartbeat than the in-depth analysis can tell you in a week.

To chart a trend group you have to define a period of time for which you’re comparing traffic estimates. I like to start with 6 months and push it back to 1-2 years (depending on how much data is available). Sometimes you cannot do that. If you only have one month’s data, however, you cannot do trend analysis. Trends do not occur over the course of a month.

Trend analysis tells you three things: who is gaining traffic, who is losing traffic, and where the traffic is seasonal or episodic. Both seasonal and episodic traffic is driven by events, but episodic traffic is unique. For example, Christmas shopping traffic is seasonal. Traffic following a devastating natural disaster is episodic.

Any sector can have both seasonal and episodic traffic. Any sector can have multiple trends.

It’s okay to compare two different sectors to each other but you won’t gain as much insight into either sector (vertical) as if you compare two trend groups within the same sector. That is probably self-evident but I’m pointing it out anyway. The more diversity you introduce into your criteria, the broader your trends become and the less granular your supportable analytical conclusions can be.

Let’s play with one trend group. I’ll pick five sites and I won’t tell you what their selection criteria are. Most if not all of you should be able to guess at least 1-2 of the criteria without any problem.

1. SEO Theory (seo-theory.com)
2. Ms. Danielle (msdanielle.com)
3. SEO By The Sea (seobythesea.com)
4. Oilman (oilman.ca)
5. Threadwatcher (threadwatchers.com)

Here is the Alexa comparison.

AttentionMeter’s attempt at providing me with an embed blanked this Web site. Bad AttentionMeter! Here is AttentionMeter’s aggregate comparison (BTW — when I returned to the site this data showed up in my browser but the URL was just for the domain name).

You can also look at Compete’s comparison here.

Quantcast has nothing useful to offer and I have no intention of installing their Javascript here.

I’ll type in the numbers from Rankings by hand.

1. SEO By The Sea: 94,428
2. SEO Theory: 300,219
3. Ms. Danielle: 558,974
4. Oilman: 602,336
5. Threadwatcher: NO DATA

Some Trend Group Analysis

Alexa: Alexa indicates that SEO By The Sea has been very dominant in the trend group. Bill Slawski is losing traffic but he still gets a respectable audience. Oilman has all but vanished. And Threadwatcher, Ms. Danielle, and SEO Theory are all relatively new sites. Of the three, it looks like SEO Theory is moving and shaking the most.

Compete: Looking at Compete’s unique visitor counts, it appears that SEO By The Sea is picking up steam (way to go Bill!). But SEO Theory is also moving up, now capturing slightly more than half as many visitors as SEO By The Sea (our actual visit estimates are almost identical). Ms. Danielle does all right, though she seems to have had a slight downturn. And whereas Threadwatcher is losing some ground Oilman seems to be coming back.

AttentionMeter: AttentionMeter will show you data from Alexa, Compete, or Quantcast but the only page that looks unique is this Alexa report summary.

Quantcast: Quantcast still has nothing to offer.

Rankings: The Rankings data shows that Bill Slawski is more popular than I am, but it doesn’t provide any trending data. Without trend data you cannot do a trend analysis. So why did I bother to type in those numbers? Because if anyone wants to see how this trend group performs in Rankings’ data, they can go back and capture data at a later time and compare it to these numbers. You should capture the data once a month and keep track of all data captures.

While we cannot necessarily trust any one of these services completely by itself, together they indicate several things: that both SEO Theory and SEO By The Sea are the most popular sites in the group; that both SEO Theory and Ms. Danielle are aggressively building audiences; and that both SEO By The Sea and SEO Theory have stronger appeal among technically savvy Internet marketers.

That makes sense. Both Bill Slawski and I have focused on very technical niches within the field of Internet marketing.

Ms. Danielle discusses PPC and blogging (and occasionally shares some personal events). Her audience seems to be pretty loyal but she has reached out to other blogs. As one of the sites that links to SEO Theory, I’ve watched Ms. Danielle’s blog pile up Technorati authority points rapidly. She can definitely do the social media optimization and if these trend reports are reliable then her social media optimization appears to be well-targeted.

Threadwatcher and Oilman are struggling but these reports don’t show us why. And we need to keep in mind that their server statistics may totally contradict these trends. So no trend analysis is perfect.

You need to have clear goals when you do a competitive analysis. You cannot just sit down and compare numbers and say this site is beating that site. You have to understand what the trends are that you’re looking at (and I intentionally withheld my criteria in this example) and why each site in a trend group may or may not be struggling or performing well.

You need to look for corroborating data, too. If you’re tracking data through Google Analytics, comparing the trends in GA to these types of trend reports can be both reassuring and frustrating. Don’t expect much agreement on the details, but the trends may still be similar. If you find your Google Analytics, Sitemeter, HitBox, Webtrends, whatever data matches the trending reports from Alexa, Compete, and other services then you can have a little more faith in the feedback you’re getting from your logs.

If, however, you see divergent patterns between your own data and the trend group data, you need to start asking some questions. Why would a trend group you choose chart a different course from your site-level analytics? Have you broken something? Have you left some code off certain pages? Are you filtering out too many sites or file types in your server logs?

You can use your own sites as a baseline to help you evaluate how reliable trend reports from the free services are. You can change out sites in your trend groups until you find a good mix that matches your server trends. When you have a trend group you feel is providing good data, you can start comparing it to other trend groups.

As you build more trend groups and look at more competitive data, you’ll get a feel for how accurate and reliable snapshot analyses can be. When done correctly they are trustworthy far more often than not.

You can never trust these free reporting and third-party reporting tools 100%, but as you measure them against each other, you’ll build a better understanding of your data. As in accounting when you’re reconciling numbers on various reports, you want to find agreement in your Web analytics reports.