Attribution: Busting The Myths

As one of the hottest topics in marketing, attribution is often presented as a panacea for marketers’ dilemmas, allowing you to understand how different advertisements in a purchase funnel work together. The typical description first shows how conversions attributable to various channels change when going from last-click to even-distribution or U-distribution (pick your favorite open-shaped letter!).

The line of reasoning then claims that multi-click attribution is needed to properly optimize your search campaign. Some attribution vendors even claim that attribution can solve the media mix problem, i.e., finding the right budget allocation to maximize the overall impact of a marketing campaign.

The truth, however, is far more nuanced than a broad brushed statement claiming the superiority of multi-event attribution over a last-click approach.

Last Click Is Not Always Bad

The typical marketing mix of an omni-channel marketer is about 50-70% offline (TV, Print, PR), 10-15% search, and about the same on display. For this marketer, if an attribution technology is only deployed for search marketing , the net impact of the deployment would be far less than the value of figuring out the right media mix and the media flighting plans for the advertiser.

We note that when looking at multi-channel data, attribution can show big synergies — but when looking at search alone, the impact is small. This is best seen with a couple of examples.

SEL_Aug1

In this dataset, I took several search advertisers and analyzed the search-only purchase funnels. Funnels were categorized into three buckets: single click, self-assist (i.e., those searches where the same word was typed multiple times), and true multi-click or assist funnels. When looking at search only, 90% of all conversions either happen with one click or with the same keyword typed. Multi-click search conversions happen 10% of the time.

In other words, last-click would capture 90% of the funnels perfectly, and even if another multi-click approach had an impact, it would only impact 10% of the funnels.

However, when looking at multiple channels, the picture changes (literally — see below!). Here, you can see that even looking at online marketing channels only, 27% of funnels involve multiple channels. This number goes up further when you include email marketing as part of the mix.

SEL_Aug2

In short, if you are doing search primarily, changing attribution rules usually doesn’t change much. You typically see a 10-20% assist funnels and 5-10% non-brand to brand type funnels. For the most part, you are fine working with the last click. However, when looking at multi-channel data, when search forms a small component of the overall picture, multi-event attribution can matter.

Attribution Is Not Media Mix

A hypothetical multi-channel advertiser has deployed a multi-event attribution system, and the chart below shows the conversions attributable to each channel by last- and even-distribution methodologies. It seems to show that display and social have a positive assist effect on search.

SEL_Aug3

This begs the question: how does this inform how the budgets across the three channels should change ? It doesn’t. At best, you may come up with a heuristic to shift budgets (like increase display budgets by 5%) but there is no guarantee that performance will improve. This is because:

  1. Attribution is a look back in time. It might tell you what channels caused conversions to happen in the past, but it doesn’t tell you the same will happen in the future.
  2. You cannot assume causality. A channel might seem to do well based on a heuristic attribution method, but there is no guarantee this will happen if you change the mix. A classic case is when branded paid search seems to do well compared to TV. So you move money from TV into paid search. Paid search would collapse, because TV and paid search are synergistic.

There are ways to overcome the causality problem, but they require experimentation and algorithmic attribution approaches. The typical attribution flavors (first, last, even, U) simply don’t do the trick.

Is Multi-Event Attribution Useless?

Not at all. When deployed correctly, i.e., across all online channels with econometric models accounting for offline marketing effects, attribution forms the backbone of bottom-up media mix models that do predictively tell you where to spend the media.

Secondly, multi-event attribution is of relevance to a multi-channel marketer with substantial budgets in two or more media (search and display, for instance). Even typical heuristic attribution can provide some insight as to channel interaction.

However, it is important to keep in mind that the value the attribution can provide you is strongly dependent on what technology was deployed. A channel tagging+rules based attribution system will give you some insightful reports, and that’s where it ends. An algorithmic attribution platform with advanced econometric models will provide you with bottoms-up media mix models that predict different marketing outcomes with a high degree of accuracy.

Conclusion

While attribution is vital to many large advertisers today, it is often oversold and misrepresented in what it can do accurately. Rules-based attributions are limited in their capabilities as they are descriptive and not predictive in their capabilities. For a small marketer with limited budgets, last click-attribution might be just fine. The same applies to marketers solely focused on search as the primary marketing channel.

While every marketer might see different cross channel interactions (and hence, attribution impacts), it is important to understand the capabilities and limitations of attribution technologies before one invests in them. When done right, it could provide you with valuable business insights and recommendations; when done wrong, it could be a colossal waste of time and money.

Opinions expressed in the article are those of the guest author and not necessarily Search Engine Land.

Related Topics: Channel: SEM | Paid Search Column

Sponsored


About The Author: is Director, Business Analytics at Adobe. He leads a global team that manages the performance of over $2 BN dollars of ad spend on search, social and display media at Adobe.

Connect with the author via: Email | Twitter | Google+ | LinkedIn



SearchCap:

Get all the top search stories emailed daily!  

Share

Other ways to share:

Read before commenting! We welcome constructive comments and allow any that meet our common sense criteria. This means being respectful and polite to others. It means providing helpful information that contributes to a story or discussion. It means leaving links only that substantially add further to a discussion. Comments using foul language, being disrespectful to others or otherwise violating what we believe are common sense standards of discussion will be deleted. Comments may also be removed if they are posted from anonymous accounts. You can read more about our comments policy here.
  • http://www.rimmkaufman.com/ George Michie

    Good stuff as always, Sid. As I pointed out a few weeks back (http://www.rimmkaufman.com/blog/attribution-myths-vs-reality-part-1-statistical-limits/09072013/) even high rent statistics struggles with the distinction between correlation and causation absent a controlled testing environment. Moreover, these systems generally fail to understand demand elasticity and the cost of incremental traffic.

  • http://www.rimmkaufman.com/ George Michie

    Good stuff as always, Sid. As I pointed out a few weeks back (http://www.rimmkaufman.com/blog/attribution-myths-vs-reality-part-1-statistical-limits/09072013/) even high rent statistics struggles with the distinction between correlation and causation absent a controlled testing environment. Moreover, these systems generally fail to understand demand elasticity and the cost of incremental traffic.

  • Raphael Calgaro

    The fact that most Analytics implementation are still in the visits level and not visitors level tend to make these stats inaccurate. The email i received on my phone in the morning triggers a search on my business computer and concludes with a transaction on my personal computer. Those are generally considered as 3 visits and three visitors whereas I am still only one person. The 3 touch point are not linked and no value is attributed to the prior visits.

  • Sid Shah

    @Raphael: I agree that analytics tracking cant capture cross device or online-offline sales funnels. You can over come some of these issues with econometric methods. However, my point here is that the statistics that you would get with any bottom up attribution method for the example you describe would still be inaccurate (first click, last click, “algorithmic” etc).

 

Get Our News, Everywhere!

Daily Email:

Follow Search Engine Land on Twitter @sengineland Like Search Engine Land on Facebook Follow Search Engine Land on Google+ Get the Search Engine Land Feed Connect with Search Engine Land on LinkedIn Check out our Tumblr! See us on Pinterest

 
 

Click to watch SMX conference video

Join us at one of our SMX or MarTech events:

United States

Europe

Australia & China

Learn more about: SMX | MarTech


Free Daily Search News Recap!

SearchCap is a once-per-day newsletter update - sign up below and get the news delivered to you!

 


 

Search Engine Land Periodic Table of SEO Success Factors

Get Your Copy
Read The Full SEO Guide