Building an Analytics Infrastructure

Authors Note: All frameworks have moved to a new home at Strategy Umwelt. Please join me at this new platform for a revised list of mental models, strategy frameworks and principles including an updated version of the 

Why is it that so many digital marketing initiatives fail? There are always excuses ("the banner ad wasn't creative enough", "the people involved were idiots", "the MD killed it before it could get off the ground"), but the root cause may in fact be a lack of an effective measurement plan to allow the identification of campaign or platform success.

In an attempt to make this process simple, Google have developed a framework to try and simplify the process of developing a solid digital analytics infrastructure in any business.

Why are Analytics important?

There are two trends driving the need for digital analytics to be a high priority.

The first is access; the Internet has now made the world's information available to nearly everyone at the click of a button. To add to this, mobile has created the ability to connect to everyone 24/7.

With this access, the consumer journey has forever changed. People are now empowered with information - they can read product reviews, get expert opinion, connect with friends for recommendations, check store inventory, or look up competitive pricing instantly from the device in their pocket.

The second trend is the invention of cloud computing, creating cheap, nearly infinite computing power. This trend has now empowered organizations to collect and analyze more business data than ever before.

Technology innovation is also increasingly breaking down the traditional data walls. Often when we talk analytics, people simply think about Google Analytics measuring a website. In the last couple of years however, this has expanded to a much wider net, including:

  1. Mobile Devices
  2. Cloud Connected POS systems
  3. CRM Systems
  4. Video Game Consoles
  5. Home Appliances

Just think about a company like Nest, and the possibilities of capturing data from a Thermostat or Smoke Alarm in the future.

The end result of all this is that the traditional linear funnel model is essentially dead. Customers are now at the centre of the universe, so our job becomes finding the points they engage with us, measuring the effectiveness of these engagements, and driving improvements in the ways we communicate.

Building an Analytics Infrastructure

For analytics to be useful in any way, it needs to be tailored to a businesses needs. To often, what's generated is affectionately known as "data puke" - top line numbers without context that is essentially useless.

Our goal therefore is to create an infrastructure that delivers real business value.

The first thing we need are the right people. You will need:

  1. Someone who understands the business objectives and strategies (key stakeholders and the marketing department)
  2. Someone who understands analytics (hopefully the agency)
  3. Someone with technical skills to handle the implementation (agency and development team)

Together, we apply the following model, which we will go into detail:

Building an Analytics Infrastructure

Defining a Measurement Plan

The measurement plan for analytics is a critical step. Again, plenty of other models exist to simplify this process, but by far the best one I have seen is the Digital Marketing & Measurement Model developed by Avinash Kaushik.

This is another 5 step process outlined as the following:

  1. Document Business Objectives
  2. Identify Key Strategies, Goals & Tactics
  3. Choose KPIs
  4. Choose Segments
  5. Choose Targets

Document Business Objectives

Defining a Business Objective can be hard. Often this requires organizational alignment across multiple departments and the buy in from senior stakeholders, which often means the marketing department struggle to get consensus.

What definitely should not be the Business Objective is the a statement like "get more sales". Everyones goal ultimately is to get more sales! We need to try and narrow down this a bit more to develop an appropriate strategy.

A good technique to help with this process is to ask why. Why does this [website] / [campaign] / [business] exist? This may at least generate some healthy discussion (potentially consider other techniques like the Golden Circle).

Identify Key Goals & Tactics

The definition of a goal is a specific strategy leveraged to accomplish the business objective.

What we want to do here is outline both our Macro and Micros goals. Our Macro conversions may look something like this, depending on the type of website we are running:

Macro Analytics Objectives

Micro conversions are more of our assisting activities, so think things like signing up to our email database - this allows us to continue the conversation, so has the chance to lead to a conversion later on.

Choose KPIs

The definition of a Key Performance Indicator is a metric that helps you to understand how you are doing against your objectives.

Make sure you are not just scratching the surface with metrics like clicks and transactions; use smart KPIs that go a little deeper.

Choose Segments

The definition of a segment is basically a group of people, their sources, onsite behavior, and their outcomes.

While aggregating data helps you to see trends over time, segments allow you to determine why these trends have occurred, so are incredibly powerful. 

Data is useless without segmentation, so make sure this is properly defined.

Common segment examples include:

  • Date and Time
  • Device
  • Marketing Channel
  • Geographic Location 
  • Repeat vs Returning Customer

Choose Targets

The definition of Targets are numerical values you have predetermined as indicators of success or failure. 

Basically these are the numbers you are defining to allow you to benchmark KPIs. So if you get 1 million views of a video on YouTube, a target will help determine if this is fact is a good thing.

These should be driven from things like historical performance. If you have none of this, put in a holding set of numbers until you can first start driving in some form of data and then adjust. Never skip this step and not add them in!

Following this model, you should have something that looks like this (using a fictitious brand as an example):

Defining a Measurement Plan

Document Technical Infrastructure

Going back to our next step in developing our Analytics Infrastructure, the next step involves documenting the technical infrastructure.

It is important to accurately document and plan ahead for the following:

  • Changing Server Technologies
  • Server Redirects
  • Query String Parameters
  • Flash & AJAX events
  • Multiple Sub-Domains
  • Existing Mobile Sites
  • Responsive Web Design

Create Implementation Plan

This step will differ depending on what tool you are using. With something like Google Analytics, this means defining the code snippets and campaign URL structure you will need locked down for future activity.


Pretty simple, have the development & mobile teams execute the plan.

Maintain and Refine

One of the most important things to remember in digital is nothing is static - your business requirements and your technical environment need to change over time to maintain a competitive advantage.

In order to ensure this is accounted for, we use a model called the Analytics Improvement Cycle, combined with ensuring the agency and development teams are on hand to continue to optimise analytics going forward.

The Analytics Improvement Cycle

This model is broken down as the following:


Basically just collecting the data on hand, through the analytics software.


Packaging the data to collate and send to decision makers. This is often filtered through a reporting template or dashboard.


Analysing what trends are showing us through deeper segmentation. This can potentially involve things like competitive analysis. We create a hypothesis, and then analyse why the numbers do or don't show this.


This involves applying different solutions or tests identified in the analysis phase, and can include things like A/B testing strategies to determine change.

The goal here is to take opinion out of the decision making process - we have the tools to validate, so until we properly test it, it remains the opinion of one person.


The end result of our tests are we make any required changes to the platform based on our test results, and then start the process all over again.

The Improvement Cycle


Setting up digital analytics does require an investment in time, people, process and technology, but with good data creating the foundation for smart business decisions, following this process should become the cornerstone of any digital strategy.

This post continues my series on Mental Models.