Gold Mining in Analytics: Identifying Relevant Metrics and Filtering out Irrelevant Data

When decision makers need to make actionable choices, they sometimes rely on data from Web analytics. For the analytics’ data to be reliable, it must be relevant and filtered.

The process of providing reliable data is similar to a gold hunt for gold. First, a gold miner will find a prospective location. Then, he/she will take a sample from the location and pan (filter out) all the rocks and gravel from that sample, leaving behind only the gold. This process helps the gold miner determine why this could be a good location to continue searching or not. This is the same process one must use in gathering data from Web analytics for decision making.

The purpose of analytics is to help us better make informed decisions. In order for analytics to achieve that, one must know the story it tells and be able to reason from its information.

Identifying Metrics

To determine the story of the analytics we need to identify the “5 Ws” – who, what, when, where and why. This will help the decision maker understand the reasoning.

  • Who’s coming to our site?
  • What did they do on the site?
  • When did they do something?
  • Where did they come from?
  • Why did they interact the way they did?

Inside of an analytics program there are plenty of metrics that one can use to gather information for a story.

Various types of metrics include:

  • Visitor locations
  • Number of visits
  • Pages per visit
  • Visits per day
  • Average time spent on site per page
  • Mobile or desktop usage
  • Browser usage
  • Content views
  • Traffic sources
  • Paid vs. organic search
  • Search terms
  • Number of conversions

To identify the correct metrics to use, we need to understand what decision is being made and what is needed to make it.

For Example:

At a small B2B services company, the owner wants to know if their in-house SEO program is working or not. The lead SEO expert needs to evaluate the program for success and assess the reasons why it is working or not. The SEO expert determines these metric points are needed in the report:

  • Number of visitors arriving from search (The Who)
  • Average page views from search visitors (The What)
  • Average time spent on site from search visitors (The What)
  • Bounce rate from search visitors
  • Number of search terms (The Where)

In this case, the when and why are not relevant data for the reasoning.

The lead SEO expert might report the SEO program is working because search visitors are up, visitors are staying longer on the website and they are viewing more pages. In addition, visitors are finding the site with more search terms.

Filtering Data

But this might not be the correct story.  To tell the correct story one must filter out irrelevant and misleading data from the metrics.

In this example, the SEO expert needs to filter out:

  • Searches from keywords containing brand or company name.  A visitor typing in your name already knows who you are and isn’t really searching for your products or services. This is basically a direct referral, not a true organic searcher.
  • Everybody who’s not a new visitor. This depends on the company. But any small- to medium-sized company should filter these out because repeat visitors can skew data.
  • Visitors from other countries. Unless you are a global company, there is no reason to include them as they are not real prospects.
  • Search terms with 100-percent bounce rates. These are probably bots, not humans, and they don’t provide any value as they are bouncing.

With these filters in place the story could change dramatically. In fact, the SEO lead might report that the SEO program is not working because search prospects/visitors are declining and they are viewing fewer pages.  Also, they are finding the website with fewer non-branded/company search terms.

This is the correct gold sample our decision maker is looking for. This data should help him/her make a decision as to how the SEO program should or should not be adjusted for improvement or continued success based on the story it tells.


Filtering data is the most important step in understanding how to truly use Web analytics.  It can turn a story from positive to negative or vice versa. Not filtering is like working out a complicated math program and then forgetting to carry the negative sign. When that happens, the solution to the problem turns out all wrong.

Site Footer