SearchSEOSEO Diagnostics With Google Analytics

SEO Diagnostics With Google Analytics

Start your investigation without letting assumptions or biases affect your thoroughness.

When things are going well on the SEO front, no one really asks questions. Conversions and traffic are up? Great! It must be because of last month’s optimizations. However, when performance declines or just fails to meet expectations, it’s all hands on deck until an explanation is found.

Once you’ve eliminated the obvious thing like the server being down, a Panda/Penguin hit, or some change in offline advertising, you’re left with digging through the data in your analytics program. If you’re lucky, you’ve got something as easy to use as Google Analytics.

Unlike regular reporting where you’re looking for specific data points to pull and compare to previous time periods, diagnostic efforts entail looking for things that are unexpected such as a spike or dip in a graph. While I have a list of go-to graphs to check whenever an issue arises, I sometimes find the issue just by clicking through all of the options in the sidebar. The key is to start the investigation without letting assumptions or biases affect your thoroughness.

Misclassified Paid Search Traffic

I’m an SEO, so of course I’m going to blame my paid search counterparts when something goes wrong! All kidding aside, one of the first things I check is for misclassified paid search traffic. This is more common than you might think because it’s pretty easy to mess up tracking URLs, and a sufficiently mangled URL is going to be counted as organic traffic by Google Analytics. If you want to be proactive on this front, check for misclassification whenever you hear there’s an account structure change, the account is transferred to another person, or a new platform (e.g., Kenshoo, IgnitionOne, etc.) is introduced.

The good news is that if you’ve got a site with search engine-friendly URLs, there’s a good chance that you can spot misclassified visits by examining landing page URLs and filtering on some element that would exist only for paid search traffic. For one client of mine that means looking for a parameter I know is used to dynamically change content on the landing page.

I navigate to this report via Traffic Sources > Sources > Search > Organic. I then select Landing Page as the primary dimension and specify a filter to isolate the URLs with the parameter of interest. I’ve found that some paid search traffic often leaks through to the organic side so it’s the spikes in such activity that I look for.

Figure 1: A spike in paid search traffic being counted as organic traffic.

Unusual Increase in Direct Visits

I once had to diagnose how it was possible that while organic search traffic was steady, conversions had declined. My initial guess was that the traffic to the site was unqualified, but a quick investigation of traffic-driving keywords proved that wasn’t the case. I also quickly confirmed that the total conversions from all channels were coming in at expectations – i.e., organic search conversions declined while some other channel(s) increased by the same amount. How could that be?

The answer became apparent by cycling through the Traffic Sources > Sources reports – i.e., direct, referrals, organic, and paid. Once I saw the graph below showing a sudden increase in direct traffic I knew something was wrong. It turned out that it’s possible to mangle cookies sufficiently so that Google Analytics will think a new session is started between page views by the same visitor. As a result, visits from organic search results were being properly captured initially, but at some point during the visit the session restarted and the second session, which contained the conversion, was recorded as direct traffic.

Figure 2: An unexpected increase in direct traffic.

Answers That Lead to More Questions

Although I start with Google Analytics’ graphs when diagnosing a problem, sometimes they just don’t reveal anything. In such cases it’s necessary to dig into the details and do some mental math. For example, the following data segments visits by keywords bucketed as branded and unbranded along with not provided data. At a glance you might think that all is normal, but notice the conversion rate for not provided data. If the not provided data is made up of branded and unbranded keywords (there are no other options), how is it possible that the conversion rate is higher than either of the other two segments?

Unfortunately, this is one of those cases where Google Analytics answers one question, but leaves you with another to investigate. Explaining this sort of discrepancy is an article of its own.

Figure 3: Conversion rate for not provided too high.

The three ideas in this piece might make a diagnostic effort look easy, but that’s just because I zeroed in my findings. The way things typically play out is that hours are spent by multiple people digging around and doing sanity checks on any findings before they’re bubbled up to senior management.

Happy digging!

Image on home page via Shutterstock.

Related Articles

Google Analytics: A guide to confusing terms

Analytics Google Analytics: A guide to confusing terms

2y Rebecca Sentance
Are You Optimistic About SEO?

Search Are You Optimistic About SEO?

3y Bryan Eisenberg
7 SEO Tools to Improve Content Marketing

Asia 7 SEO Tools to Improve Content Marketing

3y Navneet Kaushal
5 Tips for Getting the Most Out of UTM Tagging

Asia 5 Tips for Getting the Most Out of UTM Tagging

4y Zac Martin
To Affinity and Beyond

Paid Search To Affinity and Beyond

4y Elizabeth Marsten
Maximizing SEO Results for Hotel Marketing

Asia Maximizing SEO Results for Hotel Marketing

4y Gordon Choi
Bing Product Ads: Early Returns

Paid Search Bing Product Ads: Early Returns

4y Elizabeth Marsten
Breaking Down Silos Key to Converged Media in Marketing

Paid Search Breaking Down Silos Key to Converged Media in Marketing

5y Thom Craver