Attribution Modeling – Giving Credit Where Credit Is Due

Attribution Modeling – Giving Credit Where Credit Is Due
Wednesday, July 20, 2016
DAC

Even as a digital marketer I can’t pretend to operate in isolation. As a brand, you have multiple channels to choose to invest in including digital, OOH, TV, radio and print. All of these have the potential to exert influence on a consumer’s path to purchase and all of them are vying for your brands precious marketing dollars. So isn’t it important to know which channel(s) are actually driving sales?

 

What Is This Thing You Speak Of?

Attribution is pretty straightforward. It is understanding how various tactics are interconnected and the part they play in reaching a desired outcome. For businesses this could be generating leads, increasing conversions, building brand awareness or improving customer sentiment.

 

Different Strokes For Different Folks

There are many ‘out-of-the-box’ models to choose from. Some are simple and narrow in focus. Others are more complex and holistic. (I personally don’t find a lot of utility in the first few models but that doesn’t mean they won’t offer insights to your organization):

Last Interaction – this is an oversimplified concept that isn’t necessarily realistic. It states that the last touch point prior to the conversion receives all of the credit. So even if a consumer encountered email marketing, display ads, and a television commercial before visiting the website the site would get 100% of the credit. Show an example

Last Non-Direct Click – this is similar to the last interaction model but takes one step back. The philosophy is to give 100% credit to the second last touch point because the assumption is that that is what drove you to take an action. For example, a compelling native advertisement created the interest and the decision to buy was already made before interacting with the direct channel.

First Interaction – I like to think of this as the ‘love-at-first-sight’ model. It states that the first channel was so compelling the consumer was immediately smitten and was blind to all of the touch points that followed. Show an example

Linear – we are told at an early age that life is not always fair but the linear model firmly disagrees. Think of this as the ‘participation ribbon’ model – whether right or wrong each channel gets an equal slice of the credit pie. So if there were 4 distinct channels in play each would receive 25% credit. Show an example

Time Decay – who knew marketing could share similarities with radioactive isotopes (and I don’t mean that marketing is toxic!). This model applies a half-life formula (of your choosing) to each channel. The last touch point gets the most credit and the preceding channels get increasingly less credit after a pre-defined period. So, if a path to purchase takes 20 days and you had 3 different interactions the model could look like 60%/25%/15% respectively.

Position Based – I call this the sandwich model. The ‘bread’ (first and last channel) each get 40% of the credit and the remaining 20% gets split evenly between the remaining ‘ingredients’.

Custom Models – if you, or someone in your organization, is an actuarial guru you can also develop custom models that address the specific needs of your organization or that fill in the gaps left by other models.

There is no silver bullet for selecting one that fits your needs. Make an educated decision on which to apply that can be justified and test it out.

 

All Things Considered

Attribution modeling can be applied to both offline and online activities and can also extend to things that are outside of your control.

Types of online attribution include your website, paid search and display campaigns, organic rankings in SERP’s, social media and content. Offline attribution encompasses traditional marketing mechanisms such as out-of-home (billboards, transit, street furniture and other), print (magazines, newspapers, flyers and pamphlets), radio and television.
nintendo and john stret

But even the most perfectly crafted and symbiotic marketing strategy does not guarantee results. Brands need to remain acutely aware of non-marketing factors that could simultaneously be at play eroding the potential of your marketing initiatives. Variables such as the economy, politics, cultural tastes and biases, geography and even weather can all disrupt marketing campaigns. A failure to keep a pulse on these factors can render any marketing efforts moot even before they are deployed.

For example, you have this great new product that you want to promote. You work with your agencies to develop the right messaging, to the right audience, in the right places and at the right times. But after the first month, there has been no traction and eyebrows are starting to raise. What could have possibly gone wrong?

Upon further examination, you come to the realization that the state with the highest concentration of your target audience recently experienced a huge downturn in the economy. Regardless of the value your product offers pockets are tight and disposable income is at a minimum. Until the economy improves, or unless you start giving your product away for free, the conditions just aren’t there to achieve the projected results.

 

The Final Word

Attribution modeling can seem scary at first but ignoring what data can tell you is even more dangerous. Start experimenting with different models and use the insights they produce to tweak your marketing investments to ensure the highest ROI. I promise you that if you believe in the power of attribution modeling you will gain an advantage over your competition.