30 . 9 . 14

Attribution Modelling – Part 1

It has long been a problem for many companies, establishing the exact effect of each channel and their involvement in conversions. 

Attribution modelling aims to given a better understanding of the entire customer journey and each interaction they make with the brand, what forms of media they use and when they are used. This is Part 1 of a series of posts where we discuss the rationale, set up and example scenarios of attribution modelling.

Editing the Standard Google Analytics Grouping

Before we look to evaluate each channel’s role, we alter the standard Google Analytics grouping path to match the clients’ utilised channels. This granularity allows attribution to be viewed at a customised level, particularly useful for campaign or even keyword category level.
We can use this information against conversion data to attribute a value to each source involved with the respective conversions. So how much do we attribute to each source?


The above screenshot shows a typical attribution view from Google Analytics, although we have customised the channel names to better understand attribution from the clients’ perspective (more on this to follow). This data can be presented against the number of conversions and conversion value to better understand the value of each source.

Let’s say the journey represented by line 6 in the table above created £2,000 of revenue, does this mean we attribute £1,000 to both ‘paid – non brand’ and ‘affiliate’? Should the entire value be attributed to ‘paid non-brand’ with the assumption these users may not have visited through the affiliate channel, had they not discovered the brand through paid search? Should it perhaps be 60:40, 30:70 or a 90:10 split?

The example discussed above obviously becomes more and more complex for users having 3, 4, 5+ interactions before converting.

Which Model to Use

The model we use to attribute value to each interaction point depends upon the client and the data, with no one model suiting all. For clients with longer conversion journeys, a time decay model may be more appropriate, with the later steps of the conversion receiving a slightly higher attributed value. For sites with short conversion journeys, perhaps models with slightly more weight towards the earlier steps could work better initially.

Using Excel

Once we have modelled out the attribution data we can more accurately assign a value to each step within each journey before adding this up against all conversion funnels to calculate how much each channel is worth to the business.

Downloading all the data into Microsoft Excel, we are able to factor in, for example, what a 20% reduction in paid search would look like or how an increase in an email marketing campaign would have affected the customer journeys and ultimately conversions and revenue.

Finally when comparing these figures against spends we can look to adjust budgets to create a more cost effective marketing solution, implementing this, and then repeating the process.


Stay tuned for Part 2 where we will discuss example attribution modelling scenarios!

Other Helpful Resources


Maximize business opportunity with data-driven decision making. Find out more about our data team.

Phil Gawrylo is Senior Data & Insights Analyst in the data team at twentysix. He enjoys the challenge of unearthing actionable insights from data analysis to help inform marketing strategy.

Share This: