Beginner’s Guide to Attribution Modeling | Analytics Advice

Today, let’s look at the beginner’s guide to attribution modeling.  And really, unless you’re going to actually do attribution modeling, this is the only way to go. From a 30,000 foot view, we should be able to see its value, understand different ways to do it, gain insight into the pitfalls we want to avoid, and look at a simple way to get started.

What is Attribution Modeling?

Here’s how Forbes.com defines it:  “this is the science of using advanced analytics to allocate proportional credit to each marketing touch point across all online and offline channels, leading to a desired customer action.”

Think of it this way, your customer goes through many touches from the moment they enter your marketing funnel to the moment they buy.  Some touches may be through digital media while others are offline.  Attribution looks at conversion, or customer action, as a process and identifies the relative value of each touch.  This way, you can do more of what works and less of what doesn’t.

Measuring touches across online and offline marketing media?  Yes, it can get complex.  That’s why when you mention attribution to a marketer enthusiasm will sweep across their face as a bead of sweat forms on their brow.

It can be complex.  It requires a combination of data, measurement, and modeling using tools like Google Analytics, Visual IQ, AOL/Converto and tag management tools like Google Tag Manager to collect data on user activity.  That said, it’s the essential model to understand the success of marketing efforts.  Let’s pause for a moment and look at what it replaced…

”Last Click” is dead, Long Live Attribution

Years ago, the “last click” model ruled.  It gave full credit to the last click that led to the desired customer action.  This mean that if your flow was this:

Google Search –> Client Site –> Social/Display –> Landing Page –> Direct navigation –> SEM –> Conversion

Your SEM, or paid search ad, got all the credit.  Your high ranking organic search?  No, that wasn’t it.  Your awesome landing page?  Nope.  Social?  Nah.  Only the SEM.

So your SEM team would dance and celebrate their wicked success while everyone else wondered why their efforts wasn’t working.  That was until they changed the SEO, social and landing page approaches and suddenly the SEM fell off.

That’s why “last click” died years ago. It still hangs around as it’s easy attribution.  Of course, it’s also not always accurate.

With attribution, the goal is to understand the relative influence of each element to create actionable steps and apply them to improve marketing results.

Attribution Models

There are four primary attribution models.  The model used should reflect an alignment of a company’s marketing and business goals.  Each one looks at different marketing elements and may require different tools.

Online to Offline Model:  Here the goal is to gain an understanding of what digital marketing efforts contribute to a specific offline conversion.

Multi-channel attribution, Across Multiple Screens (AMS):  The goal here seeks to understand the interplay between advertising channels like TV, print and digital.

Multi-channel attribution, Across Digital Channels (ADC):  This model looks specifically at digital touch points like social, search display, and  mobile in the conversion process.

Post-conversion attribution:  This approach looks at phone and other post-lead sales interactions and requires CRM integration.

Potential Pitfalls

There are a couple of ways to derail attribution modeling efforts.  Multiple measurement platforms or methods can confuse the data and lead to inaccuracies.  Except for pure-play eCommerce whose efforts are completely digital, the real world shouldn’t be forgotten.  Phone calls, foot traffic, TV, or even road-side billboards might have an impact and need inclusion.  Other elements considered secondary actions like using a store locator, downloading and reading an eBook or printing coupons (on-site or off-site) also need to be included if they are part of the marketing mix.  Any or all of these can skew data and throw-off your attribution models.

Getting Started

Here’s how to get started and build your attribution model.

First, look at your data.  It might seem counter-intuitive, but before you start you need to know what data you have and what you don’t.  Take a deep look into your data to see what insights it reveals.  Then sketch it out or create a flow of what you have so you have a visual of your funnel.  Depending on the size of your business, this may need to be done for different products or marketing campaigns.

Next, define your key business challenge.  This will help to establish goals and give direction to the attribution reporting.

Ask the questions in line with your challenges and goals.  Are you looking for customer acquisition?  Are you looking for growth or is efficiency more the goal?  Do you need to create a model that measures the best sources for high quality leads?  Questions like these will help to create the model best suited to measure your marketing efforts.