Attribution modelling – times are a changing

by Peter Young on October 13, 2010 · 2 comments

There used to be a time when as marketeers, us SEO’s sat as a very siloed bunch – but times have changed and continue to change at breakneck speed. SEO techniques are increasingly more sophisticated, and with it the way SEO practioners measure and evaluate campaign success, something we covered during a recent post on whether rankings as a metric for success were dead.

One area which is developing very quickly are analytics. Traditionally many SEOs have used traditional tools such as Google Analytics to evalaute their campaign impact much of the time in isolation to other activity which may be going on within an advertisers marketing ix, however as the influence of SEO has grown the traditional big boys within the paid search campaign management marketplace have started to take notice.

In particular, two tools have started to make interesting manouvres into the organic marketplace, namely Atlas and Doubleclick. Before I go further its possibly worth highlighting why I say interesting. Lets create a scenario:

  • Client x is running three channels namely PPC, SEO and Display, of which that client is monitoring/managing their campaigns via Doubleclick – SEO is monitored for traffic purposes through Google Analytics – a scenario I would suggest many SEO’s are used to.  For simplicity sake we are only going to use PPC and SEO here though.
  • At the end of the month PPCagency A reports on their traffic, leads/sales and lead/sales revenue to the client, using last click attribution from their monitoring systems. At the same time SEOAgency B does the same  – however using Google Analytics as their source for data.
  • Client x gets the report and concludes that PPC leads + SEO leads = overall leads and does not take into account any potential cannabilisation from the above mentioned sources.

The problem with the above scenario in a traditional sense is that we are going to have a fairly significant volume of doublecounted leads in the scenario highlighted above for the following reasons

  • The PPC tracking system in all likelihood has not been configured to monitor organic traffic and as such will set the cookie for last click to PPC.  Say the user then goes away from the site – and then returns via an organic result – the analytics would pick this up – however the paid search monitoring system will not
  • The user then goes on to convert and gets counted as a lead via GA for organic search and as a lead for paid search via the tracking system there. Same lead but attributed to two differenct channels….

Traditionally this has been a very difficult situation to overcome – however increasingly these Paid Search management systems are starting to migrate into the organic search landscape. With Doubleclick for example, one is now able to track organic search like never before, allowing the opportunity to pivot, query and integorate data like never before.

Lets take the scenario outlined above, and incorporate these tools into the framework with a central tracking system in place such as Doubleclick:

  • User enters the site via PPC.  User then goes away from the site and returns via organic search. Depending how my system is configured this would now correctly be attributed as an : organic visitor (if last click), paid search visitor (if first click) or blend of paid/organic (if blended option) has been chosen.
  • The same principle as highlighted above would then be attributed to conversion as well.

Last Click and Multi-touch attribution models both have their own pros and cons. Many argue that the last click is the most beneficial as around 75% of conversions can be attributed to a single click, however there is an increasing appreciation of the role of other channels apart from search within the conversion matrix. Further to this, assigning attribution back to Local or even offline activity is increasingly important in a blended marketing world.

Such a framework provides a far more ‘realistic’ overview of a clients search landscape and a much truer representation of the true commercial impact of their marketing activity. Doubleclick is one of the tools on the market would appear to have the upper hand at the current moment in time, however other tools Atlas shouldn’t be discounted and what these systems now allow organic search practioners to monitor and report on is the true search engine landscape – often taking into account full channel attribution modelling or ‘search path analysis’.

I should add these aren’t the only two systems on the market – however what I would highlight is that I can’t help thinking having run these blended campaigns on a number of accounts that it provides a much more more intelligent framework to work in where paid search and organic search are configured to work completely in synergy with each other. At the end of the day talk a language that is consistent to both effectively allowing you to compare apples with apples – which I can’t help thinking means less work and more time to spend doing what really makes the difference to your clients  – increasing visibility.

Google+ Comments

{ 2 comments… read them below or add one }

Chris Brinkworth - TagMan October 26, 2010 at 2:01 am


Great article. TagMan are the leaders in this area. For more people who want to learn more in this area of attribution, where it includes natural search: I wanted to highlight further research and articles in this area that TagMan have worked on. The posts are at:

In real-time, TagMan allow you to see where Natural Search, Display, Affiliate, Direct to Site etc all fit into the path. This, without the need for complex adserving patches.

Do visit our site for a video and also a Case Study for Boden and Virgin Atlantic at

Tom Cull March 14, 2012 at 7:17 pm

Very valid points – it’s a shame that so many SEO agencies are still taking a siloed approach to results analysis and planning (regardless if they manage other activity of not)

Given that mobile search is 20%+ for most sectors, and there will be very few people who do just one search on just one device before converting, the really exciting stuff is now including multi-device usage into attribution models.

Leave a Comment

Previous post:

Next post: