Attribution – which way do you do it?

Share this article

Attribution is definitely a hot topic at the moment and at the Adobe Marketing Summit there were a number of really good sessions on it (there were also a number of insane equations flying around too that I won’t get into).

So, what is attribution?

It’s about divvying up the spoils.  It’s about applying credit (usually a dollar value, but can be a number such as leads etc) for a conversion to the thing or things that contributed to the conversion.  For example: let’s assume someone came to your site and spent $1,000.  Using different models, you’d split that $1,000 across the different channels that they interacted with.

Most of the world still seems to continue to use Last Touch attribution – that is, they say the most important thing that drove someone to convert was the last thing they touched – which might be a Paid Search ad, or a Display Ad, or an Email etc.

But in reality, 60% of visitors interact with our sites through many different touch points*, and so giving credit to the last one is neither fair, nor realistic.  So, we use attribution to figure out what channels they interacted with and give those channels a different percentage of the conversion.

(*Forrester)

There are many different types of attribution and they tend to be fairly specific to each company:

  • First Touch – assign all credit from the conversion to the first channel they interacted with.
  • Last Touch – assign all credit from the conversion to the last channel they interacted with.
  • Equal Weight – assigns an equal credit to all of the channels that drove conversion.
  • Mid Weight – also called Starter, Player Closer.  This assigns an amount to the first and the last channels, and then an equal split across the channels in the middle.
  • Adjacency – gives a weighted score based on the position to the conversion, for example, the highest score is given to the channel closest to conversion, a lower score to those further away from conversion.
  • Latency – same concept as adjacency but uses a time scale to help determine the scores.
  • And then we get into more complex models, such as multi-conversion even attribution, where multiple purchases are considered as well.

So, taking our $1,000 in revenue, and looking at the different channels, let’s imagine our visitor came to our site via Paid Search, then Social, Display and then finally Email before they purchased.

Using the different models, we would credit the different channel interactions as follows:

attribution

So which one is right?

They’re all correct.  The model you choose depends on the type of marketing you do, the customer lifecycle, and the type of purchase being made.  In this respect, attribution tends to be specific to companies.  The percentages that you use for Mid Weight and Adjacency are entirely up to you.  But, you should remain consistent with that calculation over time.

How do we go about doing this then?

Ok, so, you’ve decided you want to step into the world of attribution.

  1. The first thing to figure out is what it is you want to divvy up.  Is it revenue?  Is it leads?  Is it orders?
  2. Then you figure out the way you want to divvy it up.  Do you want to use just Last Touch, or would an Even model be better?
  3. Lastly you figure out what you want to assign credit to.  Is it channels? Is it campaigns?  Keywords etc?

As you can see, it’s really not that complicated once you make a few key decisions.  Don’t get caught up in the model – the important stuff is number 1 and 3.

What’s needed to do this?

There are a few things that you need to ensure that you do to enable this.

  • You need to capture campaign code at the visitor level
  • You need to capture channel at the visitor level
  • You need to capture the order of channels that the visitor came through

Once you have those three things in your data (at the visitor level), you can then apply the different models to them.  And to do that, you can use Excel to export your data at the visitor level, then you can calculate the attribution across those visitor conversions, depending upon which model you’re using.  The math involved is not that difficult if you’re used to using Excel.  You need to figure out the number of channels each visitor used, the conversion amount, and then the percentages that you’ll divvy it up with.  Once you’ve done that, you can figure out attribution.

And, if you’ve licensed ReportBuilder, you can use that to extract the data and automate the reporting over time.

While attribution is not media mix modelling, it can help to show channel value, and hence it can begin to help align marketing spend. Caution here, though – there are many algorithms that are much better at media mix modelling.

If you’d like to find out more download our Guide to Attribution.


Share this article