What Attribution model to use for Analytics


It is been a question that has been haunting me for some time now. To which channel(s) to attribute conversions and how to evaluate this in a daily/weekly/monthly basis.

To start with I am known do come up with nutshell versions of indepth articles so that the "busy" marketer can just try applying this knowledge immediately. However, attribution is def. a big data/modeling problem and I would advise to spend the adequate time reading more and developing your own "solutions" for your business.

How to evaluate if you have an attribution problem?

The quickest way to figure this out is going to your GA account Conversions>Multi-Channel Funnels>Path Length and cross path length versus conversions and top paths. If you see most of your conversions coming from more than 1/2 paths you probably have an attribution issue.

path length in interations and conversions

Example of multi-path funnel:

different paths showing until conversion

What to do if I don't have an attribution model

Do a little dance, make a little love and be happy that you can probably use a Last interaction model (excluding direct) or a U-shaped/Position model where first click and last click normally get a 40% weight each of attribution and the middle channels get 20%.

What if I do have an attribution model problem?

You are like most of us and you need to define a good way of attributing conversions and budgeting accordingly to different channels.
Google already provides a tool that helps you compare the default (last click model) with others, under Conversions>Attribution>Model Comparison. This will give you both conversions and CPA comparison of different models. Amazing right?
Now comes the tricky part. Which model to use?

  1. Last click model: In this case the last interaction will receive all credit, in most cases the Direct channel. (not the most amazing model).

Where to use it: If your product has 0 consideration cycle and people want to buy it there and then).

  1. Last Non-Direct: More interesting than the previous with the Last Non-Direct Click attribution model, the direct channel visits are ignored and all the credit goes to the last channel that develired the conversion.

Where to use it: Direct users were already acquired from some other channel so this model provides a useful benchmark specially if you are not doing awareness campaigns.

  1. First Click: The First click attribution model is the opposite of last click. Here the first interaction will receive 100% of the credit of the conversion.

Where to use it: In case you are doing awareness campaigns to promote your services/products or not known brand.

  1. Linear model - In this case each channel in the conversion path receives equal credit for the conversion.

Where to use it: When your objective is to keep the customer engaged throughout the purchase cycle. All the channels will have equal importance.

  1. Time Decay model - the touchpoints that are closest in time to the sale or conversion will get most of the credit.

Where to use it: The default decay is 7 days so it is useful for short-purchase cycle services/products

  1. Position Based model - 40% credit is given to the first and last interaction respectively and the remaining 20% credit is distributed to the middle channels.

Where to use it: When you want to both give value to the channels that introduced the client to the brand and the ones that generated the sale.

You might also want to read:
  1. Google Attribution Models
  2. Multi-channel attribution modeling