comScore released a list of Rankings of Top Worldwide Properties  last week, but there’s likely a large segment of internet usage completely missed by their methodologies. I recently blogged about how Domainers Can’t Get No Respect  (a followup piece to my 2nd installment of “Domaining & Subdomaining In The Local Space “), because they haven’t had good independent validation of some of their traffic and conversion rate figures. When I wrote that, I didn’t realize that some of them had apparently attempted to get independent validation, but were thwarted by the methodologies of audience measurment services. Frank Schilling let me know that he’d tried to get audited by comScore  a few years ago, and they’d failed miserably, registering only about one-thirtieth of the US traffic they’re really getting.
Being somewhat familiar with comScore’s data gathering and audience share estimation methods, I can easily see how Domainers’ sites could get drastically under-represented in comScoreâ€™s figures. Read on for details…
Now, lots of other folx have criticised comScore’s methods for arriving at their figures, so this isn’t anything new. But, I can see reasons why domaining companies’ portfolio traffic could be particularly skewed and under-represented compared to most other major company websites.
comScore monitors a sample set (â€œpanelâ€?) of online users, and from that usage bases their audience share projections. (See the comScore methodology description .) They state that they have over 2 million participants in their panel, and that their users are representative of internet user demographics.
Unfortunately, their demographics may not be sufficiently representative to fairly compare traffic across all types of internet sites, and small amounts of incremental traffic to sites could easily end up being zeroed out through their statistical methods.
Here’s some reasons why domaining company traffic could be undercounted by comScore:
- The 2 million-plus member panel sounds impressive, until you divide it down by the continents, countries, provinces/states, and cities in each country — once you break it down, the sample set for each country/state/city, the sample set then becomes quite small, and arguably likely to not represent all demographics sufficiently. According to CIA estimates for this month , the global population is 6,602,224,175 so, the panel would only represent one-three-thousandth of the population, or one one-thousandth of the estimated total worldwide internet users .
- The panel likely is undercounting internet usage from work versus home. It’s improbable that most major companies would be willing to allow comScore’s monitoring software on the desktops of all of their employees, since corporations dislike spyware of this sort. Internet usage from work is a huge component of online traffic, so if this segment is under-represented, it would make sites with more workday traffic show lower amounts of usage than sites with greater nighttime traffic. I just cannot see that very many of the larger, publicly-owned companies would allow this sort of external monitoring of their employees.
- I’ve seen cases in comScore data wherein if a site receives less than a minimum number of visits during a month, comScore will not count it for ranking purposes. Since many domainers have thousands of sites on domain names of varying popularity, the small amount of panel traffic going to a portion of their sites during the sample period may get zeroed out. Domainer company traffic is impressive in the total combined usage from their entire portfolio — the small amounts from many sites can add up to a substantial amount in total.
- The biggest reason that a domainer’s total traffic could be under-counted is that new, wet-behind-the-ears internet users might be far more likely to navigate through typing guessed domain names directly into their browser window, and I suspect that those new internet users would be far less likely to be comScore panel members. The lack of the new-internet-user demographic in their panel could result in significant skewing of the comScore report tracking of domainer sites.
So, what’s the take-away in all of this? comScore would likely need to count even the smallest amount of traffic they get for sites as being statistically significant, instead of “correcting” it out, if they wish to enable fair inclusion of domainer company sites in their reports. Also, they would need to find the way to include more new user traffic, and traffic from inside companies.
Alternatively, domainers would likely be much better represented by companies that use figures from traffic actuals. IPRO  can provide independent auditing of a site’s usage, by monitoring a site’s traffic, or through processing a site’s logfiles.
Also partnering with a company that absorbs data from a number of major ISPs might provide more representative traffic tracking for domainers. There are some metrics companies who get actual usage data from a number of ISPs, and these sources would be far less likely to under-represent the newbie internet user demographic.