Multi-Channel Attribution Modelling – don’t write off the default models

Avinash Kaushik is a great measurement thought provoker (up there with the likes of Tufte imho), all-round nice guy, and friend of mine. I always come away from his posts feeling challenged and stimulated – quite a feat to achieve for your peers in a niche industry. The following post from him –  Multi-Channel Attribution Modeling: The Good, Bad and Ugly Models – is a great reference read, though I disagree on a couple of items – Essentially the default models of Google Analytics do offer value given the right circumstances and if properly understood…. Digest Avinash’s thoughts first, or dive straight into my input below…

multi-channel-funnels

Section 2. Last Non-Direct Click Attribution Model
http://www.kaushik.net/avinash/multi-channel-attribution-modeling-good-bad-ugly-models/#lastnondirect

Avinash hates this, though for me, the model has always made perfect sense in a non-multi channel world (for a moment lets put to one side the multi-channel funnel (MCF) section in GA).

For example, a big ask of web analytics is to optimise marketing campaigns. From my experience, unless a website sells a low price item (say less than $50), a campaign rarely results in the conversion. People go away, think, consult with family, friends etc. and then come back to the site directly to convert. It therefore does not make sense to give credit to the Direct channel when a campaign has been involved. That is vastly over-rating the impact of Direct (brand awareness).

Of course now that we have MCF reports, there is an argument to say that the standard reports should align with it. However there are two BIG caveats with doing that:

a) It is a huge software development task to make all 100+ of the standard reports in GA reflect the MCF model.

b) A very large part of GA’s success is its intuitive use and visual simplicity (information architecture). Essentially, you can teach yourself a great deal and that is exactly how I got started. However, there is only so far you can go down that road as the amount of information grows. I would hate to see GA take a backwards step to the days when web analytics users required a week long training course from their vendor before they could even use basic reports…!

Section 7. Position Based Attribution Model
http://www.kaushik.net/avinash/multi-channel-attribution-modeling-good-bad-ugly-models/#positionbased

Its written off by Avinash, but don’t under-estimate the value of the default 40/20/40 split. If your company has a weak brand recognition (with strong product recognition), the first touch-point is very important to you.

Avinash’s comments are relevant to a strong brand recognition company (with weak product recognition). For example, a visitor trusts the Sony brand but is searching for a specific product that the brand may or may not have. In that scenario, the first touch is not so important.

In Summary
The more you think about attribution modelling, the more complicated it becomes. Bear in mind, that as an analysts, what AM brings to the table is INSIGHTS – its not a magic bullet. You are looking to get smarter at your marketing, not searching for the pot of gold at the end of a rainbow – as so many an agency pitch presentation promises…

Looking for a keynote speaker, or wish to hire Brian…?

If you are an organisation wishing to hire me and my team, please view the Contact page. I am based in Sweden and advise organisations in Europe as well as North America.

You May Also Like…

Sayonara Universal Analytics

Sayonara Universal Analytics

My first Google Analytics data point was 15th May 2005 for UA-20024. If you are of a certain age, that may sound off...

1 Comment

  1. Sofie Westlake

    Thanks for this post Brian, I agree with your comments. It’s possible we need to go back to the drawing board to rethink how The attribution modelling can be applied. I don’t think many companies have cracked this yet. But everyone would like to. Just hope people don’t get disappointed.

    Reply

Submit a Comment

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

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Share This