Success of digital marketing, testing, targeting, or personalization critically depends on effective visitor segmentation. Brands are inundated by the sheer volume of data and not able to see the forest from the trees.


The art of online marketing has become the art and data science of online marketing.

The art part has deep roots while data aspects are often misunderstood. Visitor Segmentation is a good example.

Why Visitor Segmentation?

A business can’t win a competitive race with marketing strategy that targets an entire mass market.

This is business marketing 101.

Visitor segmentation provides high and directly measurable value:

  1. Higher marketing ROI: visitor segmentation enables more effective allocation of marketing spend on visitors that are more likely to buy
  2. Higher eCommerce conversions: customer experience optimization in combination with visitor segmentation increases average revenue per visit as well as uncovers strategic customer knowledge

Want to get more revenue from the existing web traffic? Take a few minutes to initialize a customer experience scan of your site and identify your conversion roadblocks.

The availability of high quality web analytics tools makes it possible for businesses of any size to effectively segment its visitors.

The challenge is that even experts do not fully agree about how to use ever increasing volumes of visitor data to make effective visitor segmentation decisions.

We are using the term \’Visitor Segmentation\’ instead of \’Market Segmentation\’ to make it closer to the true nature of the online business.

The term Market Segmentation is more applicable to the pre-online world where companies had to identify physical markets that can be divided into sets of consumers / persons with similar needs and wants.

In the online world the online traffic represents the entire market. The goal of Visitor Segmentation is to use the online visitor attributes to identify those that are likely to become customers.

When All You Have Is A Hammer Everything Looks Like a Nail

Existing web analytics solutions are determining how visitor segmentation is being done.

It is being done by grouping web visitors through the use of one or more data filtering criteria (dimensions) and then by the calculation of associated metric to rank segments in accordance with their performance.

At first look visitor segmentation appears to be a simple concept.

The good news is that almost every aspect of the visitors online activity can be measured.

The bad news is that almost every aspect of the visitors online activity can be measured.

The sheer volume of data can be intimidating – instead of helping,  data can become a source of inaction.

Out-Of-Box Segmentation

Regardless of which web analytics solution you are using you will be able to drill down to filter your data.

The description below will help you  more effectively use of your web analytics solution:

Dimensions: you can view a dimension as an attribute of your visitor. For ease of user purposes the attributes are commonly grouped in logical subsets:

  • Demand sources (referral site, landing page URL, channels,  sources, …)
  • Audience characteristics (cookies, data from your DMP, CRM, IP, location, browser language, …)
  • Technical (device, browser, operating system, …)
  • Behavioral (page viewed, time spent, exit intent, …)
  • Contextual (weather, days, date range, time slot, …)

Measure / Metric: This is a numerical value associated with a segment.

  • Measure is a count – scalar value (number of visits, page views, revenues, …)
  • Metric is a ratio between two measures (bounce rate, conversion rate, …)

Visit Google Analytics for more about dimensions and metrics.

Custom Dimensions And Metrics

Once you venture out of what is normally provided by a typical web analytics tool you have to be able to deal with more complicated concepts:

Scope:  The Scope as a concept represents a  grey area  of the customer segmentation process. It specifies to which data the custom dimension or metric will be applied.

Google Analytics offers the following types of scope:

  • Hit: In eCommerce this is commonly a page view, event or custom variable.
  • Session: Often called a site visit. It is the ensemble of all actions a visitor takes on the site from the moment he lands until he leaves.
  • User: This is a visitor, a live person that can be uniquely identified by a visitor tracking schema
  • Product: This scope is used in eCommerce to measure actions associated with a product.

Here is an example how you can easily get confused.

If you read  great insights  written by Avinash Kaushik, one of the main web analytics experts at Google, you would get the impression that when Google Analytics says User Scope that metrics are calculated relative to the number of unique users.

Read carefully and you will discover that User in scope\’s name represents all sessions that a user had since being assigned to a custom dimension. Therefore, a metric will be calculated by the dividing of some kind of measure (let\’s say revenue) with the number of user sessions. Very confusing!

Attribution: Attribution is  the least understood  aspect of the customer segmentation exercise and web analytics in general. It represents a relationship between Measures/Metrics and Dimensions.

There are many attribution models designed to provide a particular kind of user analysis question:

  • Content: How many times was a particular page viewed?
  • Goals: Which pages URLs contributed to the highest goal conversion rate?
  • eCommerce: How much value did a given page contribute to a transaction?
  • Internal Search: Which internal search terms contributed to a transaction?

Because each attribution model is designed to calculate a known set of metrics, some metrics—such as Pageviews—appear only in certain reports and not in others.

The attribution modeling gets even more complex if you take into the account the concept of the influence. For example, if a visitor participated in a PPC campaign 6 months ago and made his purchase today, does it make sense to attribute today\’s purchase to a campaign from 6 months ago?

Confusing Statistics

Before choosing any visitor segment you also have to consider the statistical aspects of your data:

Randomness: It is like flipping a coin. The results of two series of random coin flips will never be the same. Therefore, by measuring performance of your visitor segment you have to have in mind that such result is a random number.

Segment Size: Different visitor segments will have different sizes and different results. It is common sense to rely more on the results of larger segments than the smaller ones.

Confidence: This is a statistical concept that tells you about the quality of the results – how much you can trust what you measured. Often, the confidence is interpreted in the wrong way. For example, 95% confidence is interpreted as the result is 95% precise and not as there is a 5% chance that the results are completely wrong.

Time Varying: eCommerce is like a stock market. Sentiment, demand, or preferences are always changing. Therefore, you should not assume that your segments and associated metrics will always be the same.

By now you must feel nauseated.

You are getting lost. Visitor segmentation does not feel as simple anymore. It is more like sausage making.

Dealing with so many concepts is very confusing. It is so easy to get lost in the weeds of mechanical steps and low level details.

Common Sense

Before getting lost in the minutiae of the segmentation process let’s pause for a moment and ask ourselves two almost identical questions:

  1. is our mission to identify visitor segments who are likely to buy if they are included in a campaign,or
  2. is our mission to identify visitor segments who are only likely to buy if they are included in a campaign.
The only difference between two of them is the word <b>’only.</b>’

However, this single-word criteria is often the difference between strongly profitable and severely loss-making campaigns (and eCommerce in general).

Here is why. Visitor Segmentation must be about incremental results. It must be about visitor segments who are only likely to buy if included in a campaign.

  • Measuring incremental sales, while essential, is not enough: the goal is to maximize incremental sales
  • Businesses have a strong tendency to direct resources towards customers who would have bought anyway; this often results in comparatively few incremental sales
  • Particular care must be taken when assessing and optimizing the financial impact of incentive- based campaigns. While there may be collateral benefits, offering an incentive to a customer who would have bought anyway has a double cost—the contact cost and the (unnecessary) incentive cost
  • it is essential to use control groups so you can always report the incremental impact (uplift) of initiatives

Four Primary Visitor Segments

By focusing on the incremental sales you will dramatically simplify your visitor segmentation task.

Instead of hundreds of dimensions all you have to do is to follow a simple segmentation logic shown below:

By applying this common sense logic you will finally be able \’to see the forrest from the trees:\’

Persuadables: These are the personas who will only buy if included in a campaign. They represent the untapped potential of your e-Commerce business.

Lost Causes: These are non-customers. There is nothing you can do to move them over into the buying category.

Sure Things: These are the personas who will buy regardless of your campaigns.

Be careful with incentive-based campaigns. By targeting \’Sure Things\’ you will erode your profit margin without gaining incremental sales.

Sleeping Dogs: These are the personas who react negatively to your campaigns.

Example: The most famous example of this kind are the contract renewal notifications done by wireless communication companies. For the majority of their customers this is a reminder that they should shop around and look for another provider who offers a better deal.

Instead of seeing an increase in the contract renewals, a reminder campaign actually increases the churn (it wakes up a \’sleeping dog\’).

Segmentation Modeling

Finding Persuadable segments requires a new breed of technology that enables experimentation and data modeling. That topic deserves its own blog or a white paper.

In general, such data modeling requires use of the Control Group visitors who are not included in the campaign. One can view Control group as a reference point of the segmentation modeling.

Instead of ranking the performance of the entire segment you will be able to compare performance of two subsets of the same segment:

  • Control: performance of visitors belonging to a segment that are not included in the campaign
  • Treatment: a performance of visitors belonging to a segment that are included in the campaign

Want to get more revenue from the existing web traffic? Take a few minutes to initialize a customer experience scan of your site and identify your conversion roadblocks.