Discovering Attribution Modeling - Analytics Angels

Discovering Attribution Modeling

What is Attribution Modeling?


As you think about your marketing mix investments, you will realize that your B2B customers are going to interact with multiple channels prior to adding value to business. Someone may Google search on marketing analytics services and find Analytics Angels, then go to Facebook to further research Analytics Angels, then visit Analytics Angels’ website directly, and finally sign-up for services. This is called a “search session.”


It’s great that there are so many touchpoints guiding the B2B customer to the finishline, but which channel interaction gets credit for the conversion? This is both the definition and reason for Attribution. We need Attribution Modeling to find out how much credit each touchpoint gets before conversion. This enables efficient budgeting, to find out what is working and what isn’t, and accurate ROMI calculation.


What Attribution Modeling is NOT is a “set it and forget it” practice. You cannot build a model and walk away because the industry, your data, and your business objectives will evolve. You shouldn’t make things too complex right out of the gate. Its human nature to want to overcomplicate things, but only through active monitoring of data over time will you really be able to experiment with what’s best for your business.


Understanding different model types

Remember when we talked about “search sessions” last week and the customer journey with all those fun clicks before conversion? You might still be scratching your head wondering which one gets credited for the conversion, but we’ll review a few ways this can go down.


First up we have Last-Click. Not to be confused with the order of introduction, but this model gives credit to the last Channel before conversion, First-Click models give credit to the first Channel, and Linear models give equal credit to all Channels. These are default models in many platforms that offer Attribution Modeling (like WebTrends or Google Analytics). Read about more models’ pros/cons here and find out why Time Decay modeling is your friend. Custom Modeling is also great if you have a stellar data measurement strategy!


Some honest thoughts shared with influencers in the space:  Last-Click and First-Click models are kind of lazy and unfair. You cannot derive much value from them. Why? They skimp out on giving credit to your Channel efforts in every other step before conversion. This will skew your ROI reporting and mess up your budgets. We can go on with this topic, but let’s regroup next week.


Setting up for success with a measurement framework


One of the key necessities for Attribution Modeling is successful analytics tracking and correctly tagging all your ads and campaigns. You need to have an appropriate measurement framework for Attribution to work in your favor. Why? It all relies on… wait for it… GOOD DATA!


There are four high-level concepts of analytics tracking that you need to understand before getting started:


  • Campaign Tracking – In which campaigns are your customers interacting with you? You will need to set-up tracking links to collect this data.
  • Value Event – Actions that do not tie directly to a conversion, e.g. someone watching a video. Set-up Goals within your web analytics platform for tracking.
  • Conversion Event – Tied directly to a baseline value-add on a website, i.e. form completion. These are also called micro-conversions and can be tracked by Goals.
  • PII (Personally Identifiable Information) – Try building a framework that collects PII so you can attribute success at an individual level and optimize/personalize content along the journey.

See the diagram for an example of this framework. It is a visual way to see the purpose of every webpage, event, and action within any view. It tells you what value events you want as a result of these campaign investments. These are the building blocks for proper Attribution Modeling.

Finding your Attribution problem


To find out whether or not you have an attribution problem, the first thing you can do is look at your Path Length report in your web analytics platform. If a significant percent of your conversions have greater than one path length (aka more than one channel they pass through during their journey to conversion), you might have an attribution problem.

To dig a little deeper, what you should do next is check out your “assisted conversions” report. In this report, look at the credit a model gives each channel, i.e. Last-Click. If value<1 then Last-Click worked for that channel and if value>1 that channel is not getting much/any credit with that model. Once you do this exercise you will really start to see how blindly picking an attribution model can hurt your company’s marketing budget.


This exercise is simply a diagnostic test on your attribution health. It should be noted that there are simply too many paths a customer takes to actually control the path of every potential customer. If you have time, you should definitely read this blog post by web analytics guru Avinash Kaushik to read more on attribution issues. We will probably reference it a few more times.

Finding the right Attribution model for you 


Attribution models are not “set-it-and-forget-it” practice and take some effort to fine tune. Set-up a process to test and optimize your model to find what is best for your business since the digital ad environment, industry trends, and your business objectives will change over time.

One approach you can take immediately, is to compare different attribution models and see how they alter Channel values. For instance, you might see a shift in how much social media is valued if you compare a Last-Click model to a custom Position-Based model because in the latter, social will get more credit. Play around with this method a few times and get to know each model type.


The “Holy Grail” of models is the kind that is based off of your own data, AKA Custom. If you want to try your hand at this model, make meetings with your business leaders to understand historical performance, current marketing mix, and spend patterns. Here are a few questions to get you started:


  1. What type of user behavior do you value?
  2. What does the repeat purchase behavior look like historically?
  3. Are there any micro-conversions defined with engagement type goals, tied to the economic value?



Share with friends

Sign Up For Our Monthly Newsletter