alphanumeric symbols forming a cycle

The Analytics Cycle

Thanks to all analytics software companies for providing the simplest way to deploy analytics, by simply copying and pasting a piece of JavaScript into all pages then done. However, this makes most analytics implementations start without consideration and planning on what to measure and how to measure. It is hard to say that the act-first implementation approach is wrong but it would easily make the analytics stop at the traffic measurement without going into the analytics, where the end goal is driving business.

The approach is an analytics cycle. It is nothing new but just a plan-do-check-act cycle for analytics.

AnalyticsCycle

Define/Refine

We should know why we exist. Every digital property should have (at least) a conversion goal, which is tied back to the business objective. Conversion goals could be as simple as the number of visits (yet this should be strongly discouraged), visitor engagement based on time spent or page view per visit, lead generation by submitting some forms or subscribing to the newsletter, or even sales. They can all be translated into KPIs to guide us in the analytics planning and future analysis.

We will come back to the define/refine step in the analytics cycle to refine what we want to measure. Business objectives and conversion goals can be changed, such as moving from traffic generation to engagement to lead generation and then finally to sales conversion for an e-commerce website. Moreover, changes could arise due to technical environment changes such as a website revamp from page-oriented to AJAX-oriented.

Plan

This is the most important step in the analytics cycle to get measurement right and we can get some useful reports at the end. We already know the measurement objectives and translate them into KPIs, then we need to identify how to achieve them. Whether we are measuring a website using JavaScript tracking code, or a native app with iOS or Andriod library? Do we want to implement tracking code directly, or do we want to use Tag management tools to load tags? We should have goals but do we have other custom variables to be defined and how to get their values? How to name pages or screens of a website or app, events, custom variables, and other possible values such that we can have consistent and understandable reports?

We should think through all those possible questions and come up with a flexible plan. Analytics is a database out of our control, once data is recorded, there is no way to amend it. However, since it is a cyclic approach, we can start with a rough idea and test it with a short cycle, then come back to revise.

Implement

Yet many Tag Management tools want to enable business users to deploy tag and tracking code by themselves using Tag Management solutions, the implementation is still a rather technical task requiring a good amount of technical knowledge mainly on the web front end, such as HTML, CSS, and JavaScript.

Nowadays tracking implementation should be done through tag management tools by putting the tag loader snippet into all target web pages to load the tracking code and other tags, instead of manually putting the entire tracking code on all pages. This could significantly reduce the dependence of website developers on tracking code maintenance and speed up subsequent tracking updates as there is a high chance that no change on the website should be done. However, there is still the possibility of requiring help from website developers on tracking code updates, when there are additional custom variables to be tracked but cannot be accessed with CSS selector or not even on page yet. In this case, website developers could be involved in populating required information in the data layer object or DOM.

Report

The final bit of the analytics cycle should be connected back to the first step of the cycle. Dashboards should be created to quickly present KPIs as they are the key measurement of success. Dashboard should be consumed daily or at least weekly. Regular reporting is trying to get a bit more details on the digital property performance, covering:

  • trends and ranking on numerous metrics
  • sources of traffic, visitor profiles, and consumed contents
  • breakdown/pivot of dimensions for correlations
  • download data and combine it with offline data for insights beyond online-only information

Whenever unexpected traffic is observed, such as a sudden spike or drop, in the dashboard or regular reporting, ad hoc analysis by drilling down into data is required by multiple levels of data breakdown and filtering.

Moreover, when there were some significant changes made to the digital property, we could expect changes in analytics results as well. However, the change in digital property won’t change the overall traffic but only resulting a shift of data in detail level, such as introducing responsive design to the website, in this case, multiple levels of data breakdown and filtering are also required.

Looping back

As a cycle, it loops back to the first step. Business objectives changed, then we need new KPIs. The technology changed, and then we need to revise the approach to measurement. New features are added, then we need new tracking. Some problems or questions cannot be answered due to missing information, so we need new custom variables. The analytics is a continuous and self-improving journey that results should be getting better and better on each iteration.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *