, , , , , ,

With any research project, results are an extremely important part. They are the lifeblood of why the study is held, this isn’t to say that they are always conclusive, but one way or another they always have meaning.

In the end everything has a purpose.

Results: the path ahead

Data was collected for analysis from twelve websites that graciously allowed us access to their Google Analytics. These sites ranged from extremely small (not many user sessions) to significant sizes (millions of user sessions) based on the 3-month collection period that was used, with most of them deciding to remain anonymous.

Obviously the size of the website would end up influencing the results as the bigger the site the more could be collected through Alexa Internet. This doesn’t mean that the smaller sites would end up being useless, as there was always a conclusion to be made no matter how small the data sample. Each website would end up having at least one comparable and valuable metric.

Results: data conveys a thousand words

Hundreds of pages of data were collected in the forms of Excel Sheets, CSVs, PDFs and images. All these documents were then used to create a master file, through the means of Excel, in which each metric that could be used was added.

After consolidation, it was easier to see the bigger picture and understand what data could be used to determine the deviations between Alexa Internet and Google Analytics.

Results: the nitty gritty

With the use of the master file, the next step was to transform the data into a visual source. Depending on the size of the website some metrics were or were not available for comparison.

In the case of common metrics such as: bounce rate, page views and time on site, they were available for most.

As a whole, ten of the twelve websites returned results that could be analyzed, two websites had to be removed from the sample, and this was due to Alexa redirecting results to the domain instead of subdomain. Thus making the data collected with both tools incompatible. Some data was spotty as not every site had data available for all metrics, simply because not enough data had been collected by the Alexa service for certain smaller websites.

Each graph below portrays the intermediate results, in all three cases the data did not tend to be the same when comparing the metrics. Although it is difficult that this stage to determine which results are the most accurate, it can be said that Alexa and Google Analytics overestimate or underestimate results when compared with each other.

For the bounce rate %, which is the rate at which a visitor will leave the website without interacting, six websites could be compared. In most cases, Alexa tends to overestimate the percentage slightly in contrast to Google Analytics. In two cases Alexa the divergence is greater and in one case Google Analytics is the one that displays an overestimated bounce rate compared to Alexa.

For the number of page views per user, which is as indicated the amount of pages the user will view during the session, nine websites out of ten returned results. Google Analytics appears to somewhat overestimate the numbers compared to Alexa, with two results being quite substantially amplified. On the other hand Alexa inflates the views for one website in the sample in contrast to Google Analytics.

For the time on site, the data was converted into seconds to make the comparison easier, seven out ten websites returned comparable results. Alexa tends to amplify the amount of time the user spends on site, with one website being severely overestimating by about 500 seconds, while Google Analytics overstated the results for one website.

The next step of the data analysis would be to set out on the route of finding a correlation between Alexa and Google Analytics, but in the meantime some conclusions can definitely be made. This also means a greater understanding of the results for each tool would need to be achieved, especially when it comes to comprehending just what each disparity means and the real weight of the discrepancy.

Results: between a rock and a hard place

Although interesting data was retrieved from the analysis, some of the results were hard to come by.

This was ultimately due to the limitations encountered with the tools that were used. The method of the toolbar does influence the results that are collected especially when it comes to smaller sites. That is simply the nature of the toolbar, not everyone has it, so when it comes to websites that don’t have as much traffic as the bigger ones, well, that’s where the problems start.

Unlike JavaScript code that is integrated into the website specifically to pick up user data, the toolbar doesn’t have the same degree of coverage.

Unfortunately, it all boils down to not having enough data collected through the toolbar method for all sites in order to render the data useful and feasible.

Alexa Internet is indeed a very nifty tool when it comes to checking your position and that of competitors, but as a whole it lacks the depth that would allow a webmaster to have a global vision of their website.

by Megan Fuss and Sophie Johner