Matter Solutions Logo (Simple Version)
August 28, 2012

Introduction To Attribution Modelling

Published: 28 August 2012 

Since the addition of Multi-Channel Funnels to the Google Analytics, API companies have been able to create insightful dashboards which show return on investment metrics worthy of a CMO's approval. Mazeberry Express is soon to be a public tool which generates ROI reports based on data from Google Analytics. The most valuable analysis comes from the conversion paths that make up conversions and the time it takes for a user to convert after being introduced to a brand online. This opens up a whole range of opportunities for optimising online marketing campaigns although it is far from perfect.

The idea behind this is attribution modelling which gives business owners and marketers the ability to measure the performance of each media channel in a marketing mix. With this information to hand you can make more informed decisions about where to invest more, what strategies are effective and which channels should be given less. This moves us away from the last-click approach which attributes all of the conversion to the click before a conversion. For example, say somebody clicks on your PPC ad by typing your company's brand name into a search engine. There is no possible way that they would have intentionally searched for your brand without being influenced by something beforehand. With this being the case is it right to attribute all of that conversion to the final touch point? I wouldn't say so, and that is what attribution modelling attempts to sort out by distributing revenue based on reason and an understanding of your buying cycle.

"Half the money I spend on advertising is wasted; the trouble is I don't know which half." John Wanamaker

There are different approaches to attribution modelling and here I will cover the basic ones to get you started:

  • First-Touch Model: the very first introduction to your brand is considered the most valuable in this model. This is a flawed model in most cases and should be used in conjunction with others.
  • Last-Click Model: this is the classic form of attribution which gives all credit to the final click before a conversion.
  • Last-View Model: similar to the Last-Click Model, the final touch point to show an ad is the channel that gains credit. This is also unreliable as although an ad has been shown, it may not have been seen.
  • Linear Model: equal credit is assigned to each channel in the conversion path. If there are four ads shown to a customer before they buy a product or service then each ad is attributed 25% of the sale. Branding campaigns are best suited for this model as there are nearly always marketing channels that have persuaded customers to buy more than others.

There are some general issues with using out of the box models in that next to no marketing campaigns run the same way. Just because somebody bought from your website after clicking on 'channel X' then 'channel Y' it doesn't mean that 'channel X' caused them to click on 'channel Y'. This really comes down to intuition about your business or client and how much weight you can give to different aspects of your campaign. Let's say in the above example somebody typed 'new shoes' into a search engine and clicked on a PPC ad (channel X) then 3 days later searched for your brand name and clicked on an organic search result (channel Y) before purchasing. This would suggest that the PPC ad (channel X) should be given a large proportion more than the branded organic search listing (channel Y). Weighting channels in this way is the first step to creating your own custom attribution model that should be refined over time to reflect the current marketing mix and its influence on sales.
Attribution modelling is far from perfect but it is a step forward in the right direction to making better online marketing decisions and a huge leap away from the traditional methods of years ago. Take a look at the video below for a basic introduction to the subject from Google and Econsultancy:

Ben Maden

Read more posts by Ben

Leave a Reply

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