Marketers (and many publishers) are wrestling with the problem of cross-channel attribution: understanding what each channel adds to the entire process.
Producing a breadcrumb trail of user paths is too simplistic. The real key is understanding the incremental effect of each unit of media.
“Why do birds suddenly appear,” mused songstress Karen Carpenter, “every time you are near?” Her hypothesis: that they, like her, wanted to be close to you, is a pretty decent description of the way most online marketing is tracked.
Last click tracking, for a publisher or a media manager, means having your channel as close as possible to the final conversion. There have been winners in this method, notably search and some affiliates, and losers, such as display.
It’s easy to see why last click is considered to give a skewed view of the online world. Consumers don’t just pitch up to a search engine and decide to buy a new laptop, or take out insurance.
Marketing activity aims to target consumers at all stages of the buying cycle, whether through brand building TV ads, coupons in magazines or paid search ads.
Marketers (and many publishers) are wrestling with the problem of cross-channel attribution, understanding what each channel adds to the entire process. At one level that’s a simple process: a universal tracking technology that can record impressions, clicks and conversions for each user.
But producing a breadcrumb trail of user paths is too simplistic. The real key is understanding the incremental effect of each unit of media.
Take, for example, re-targeting people who have come to your site but not converted with display ads. This is often considered ‘low-hanging fruit’ and can yield very attractive CPAs or ROI, sometimes better than search.
However, there’s a dark secret to re-targeting (whisper it) some of those people would have come back anyway without you prompting them. You are in fact paying for some conversions you would have got anyway.
As you recoil in horror at this discovery, let me reassure you I’m not saying there’s anything wrong with re-targeting. We run successful re-targeting campaigns for many of our clients.
The essential part is in knowing the incremental effect, and thus what your marginal cost is to acquire those conversions. Then you can really value the worth of the media channel.
That’s relatively easy to do with re-targeting by splitting cookie pools into control and target groups, and, rest easy, our experiments show that it’s still a worthwhile activity.
As this year develops, the focus on cross-channel attribution is only going to increase as marketers try to find the right values of the media they are buying.
Only then can we tackle the second part of the problem: optimisation. As Ms Carpenter might have added, we’ve only just begun.
Jonathan Beeston is Global Marketing Director at Efficient Frontier and Context Optional a guest blogger on Econsultancy. He can be found on Twitter and LinkedIn.
Recommended reports
Online Measurement and Strategy Report 2011
An analysis of the online measurement industry based on a survey of digital marketers. The 67-page report and 29-slide summary presentation contain valuable insights into the use of paid-for and free web analytics tools. Beyond web analytics, the report also looks at a range of data sources and information requirements, and the extent to which companies are able to develop a coherent measurement strategy.
Alternatively, see our Media Growth Trends or FUNNEL - Engage presentations.
Senior PPC Analyst at Epiphany Solutions
2:45PM on 18th February 2011
Great post Jonathan.
"Marketers (and many publishers) are wrestling with the problem of cross-channel attribution, understanding what each channel adds to the entire process."
That is so true - I often encounter issues when running client paid search campaigns where we see a particular platform under perform in terms of hitting the required maximum CPA, yet you can be certain that it is responsible in aiding conversions further down the funnel.
As a result, I could be running a Facebook or display campaign that is bringing in conversions at a CPA of £100 compared to Google's £10 CPA. As a result, the common sense approach would be to divert those budgets all into Google where conversions are cheaper. But it's what's to say that the exposure both Facebook and Display are offering isn't contributing towards the excellent conversion rates on the search campaign, particularly on brand terms or even direct/organic traffic to the site?
6:52PM on 20th February 2011
hi jonathan, please could you explain what you mean by "splitting cookie pools into control and target groups"? I don't understand what you mean.
Global Marketing Director at Efficient Frontier, an Adobe company
7:41PM on 20th February 2011
@Peter - thanks for the comment. Glad I'm not the only one with these challenges!
@Pete - apologies, I lapsed into a bit of jargon there. So when people visit a site you can drop a cookie on each person. That collection of cookies (the pool) can then be split, randomly, in two. The first half can be the control group. You retarget those people with public service ads (ie non-relevant) and see what rate they convert at. Let's say it's 10%. Even though you haven't prompted those people, some have naturally come back and converted. The second group you do show proper ads to, to convince them to convert. Let's say they convert at 20%. So now you know your retargeting generates an additional conversion rate of 10%, and thus you can evaluate its real ROI based on what it costs you.
6:22AM on 21st February 2011
very nice I like it thank you for your sharing
Online Marketing Engagement Manager at Adecco management & consulting
1:48PM on 21st February 2011
Probably the best solution I've worked with is regression analysis to assign a positive or negative 'multiplier' to each media source.
It's not easy (or indeed cheap) to setup and it does work less well with constantly active media like generic paid search (v.s. intermittent campaigns like TV) but it does overcome the issues with tracking offline to online response and some of the subjectivity as to what credit to assign different ad formats etc.
Jonathan - I like your idea of a setting up a control group using public service ads to measure the incremental value of retargetting, very clever.
Head of Tracking Technology at OMD
3:38PM on 21st February 2011
Good article Jonathan. I agree that It is important to determine the incremental uplift each channel provides, as otherwise the advertiser risks running a highly subjective attribution model.
However we have experienced issues with splitting cookie pools; specifically with those users who regularly delete cookies or have multiple devices/browsers. Cookies are not users, and hence an A/B split base on cookies can result in high levels of contamination within the results.
We have seen evidence that in just 30 days, more than 40% of users have been counted as 2 or more cookies.
7:34PM on 21st February 2011
very interesting article ... analytics are getting increasingly important as is the complexity of the solutions you need to make sense of the data they collect.
9:54PM on 23rd February 2011
hi
taking kwik fit as an example for the control group would you serve a generic ad (e.g. kwik fit are great) and the target group a specific ad (e.g you looked at michelin tyres, you can buy these now)?
Global Marketing Director at Efficient Frontier, an Adobe company
3:19PM on 24th February 2011
@Peggy - for the control group you would use a public service ad, e.g. for a charity, that is completely unconnected with Kwik Fit.
3:40PM on 9th May 2011
Hi, I know Google is planning on implementing some features into Analytics to help with attribution moedelling, but until then does anyone know of a tool or program that would suit an SME? TagMan looks great but I'm pretty sure our budgets wouldn't stretch that far.
Thanks in advance.