Is Your Attribution Model Appropriate?

Recently I have spent an awful lot of time thinking about and talking about data accuracy issues in the field of web analytics. The widespread use of cookies as a tracking mechanism and the underlying assumption that “one cookie = one visitor” is a big part of the problem, but cookies are not the only problem. Another problem, one that I actually believe to be more substantial than cookies and visitors, is the challenge of campaign attribution.

Challenge? What’s hard about campaign attribution? You tag campaigns and web analytics tells you what works, right? You get pretty ROI graphs and click-reports and all that fun stuff? Campaign analytics is easy!

Wrong.

One of the best-kept secrets in online marketing is that most campaign attribution data is completely wrong and the models used to evaluate campaign performance are wholly inappropriate. The relative nascence of digital marketing practices, combined with conflicting measurement systems and poorly understood interaction between online marketing channels, likely means that hundreds of millions of dollars are wasted annually on marketing efforts that don’t produce their intended results.

Companies are increasingly responding to this observation by re-examining their marketing measurement systems. Even the most cursory analysis yields a great deal of information about the “campaign attribution problem.” Popularized recently by Microsoft with their “Engagement Mapping” efforts as well as analysis published by Forrester Research and others, it is clear that the most widely used online campaign attribution model is inherently flawed.

To correct these flaws and begin to improve both the accuracy of measurement and the general understanding of how marketing really works online, Web Analytics Demystified recommends a new approach to campaign analysis. Dubbed “Appropriate Attribution”, the approach leverages widely available but infrequently used data to triangulate towards the true value of online marketing efforts.

Given that the majority of online advertisers have direct response goals, and that most marketers are still generally unsatisfied with the campaign measurement tools at their disposal, Web Analytics Demystified believes that Appropriate Attribution is the first step towards improving companies’ collective understanding of their digital marketing efforts.

Eventually marketers will have access to robust warehouses of data detailing consumer interaction with online media and advertising, but the adage “you must walk before you can run” is as true in digital marketing as it is in life. Before business owners and marketers become fully equipped to benefit from complex marketing mix analysis of online and offline channels, they are well advised to address the campaign attribution problem to increase the return on their valuable dollars spent for online marketing efforts.

Thanks to the fine folks at Coremetrics you can read all about Appropriate Attribution and learn how you can start to get a better understanding of your online marketing efforts today.

Download your copy of the Appropriate Attribution paper from Coremetrics today.

Posted Monday, April 20th, 2009 | 8 responses | Add a Comment | Share, Save or Email

Joshua Dreller

It amazes me that Attribution isn’t literally tearing through the industry like a wildfire. From the moment I heard “death of the last ad clicked” a few years ago, I was immediately converted.

If you’re not doing attribution, you’re literally mismanaging your advertising campaigns. If you’re okay with that, then I don’t know what to say to you.

Rob McLaughlin

To echo Joshua really…it is so easy to organise your analytics such that you can see multiple touch points I simply can’t understand why it is not done as standard.

Following from one of Eric’s posts on the economic downturn – marketers surley can not continue to believe unreliable and siloed ROI reports.

Great post, thanks.

Christopher Berry

Well said Peterson.

One of the reasons I don’t think it hasn’t exploded is because Search (”Direct”) people and Analytics people aren’t always the same people. Frequently, they don’t talk to each other. That’s started to change. (Web Analytics Wednesday’s in Toronto, at least, attract large contingents of search folk.)

I’ve had several attempts at reconciling WA data with all the varieties of tracking data on a 1:1 basis. The literal ‘best that I can do’ for attribution is to take extract time series from all the different systems and try to arrive at a multi-channel view. It’s imperfect.

Part of the reason, so the legend goes, is that one tool tries to correct for fraud – while the other tool does not, and the myriad of other technical issues.

I’ve got to ask: what percentage error are we, as marketers, fine with?

eric

Josh and Rob: I think the (unfortunate) reality is that many people simply aren’t aware of the scope of the problem they may have.

While I was writing the paper I quickly reviewed the major web analytics applications and their UI for campaign analytics and with one exception there was really ** nothing ** that called out the attribution model in place. So even though many applications have flexibility in this regard, the knowledge that campaigns goals are being assigned based on “90 day backward-looking last-click” or whatever is hidden from everyone except the analytics administrator (who is often not the marketer who needs the data.)

My goal for the white paper is just to get people to start asking the questions “What is our attribution model” and “Is that model appropriate?” Nothing more. The Appropriate Attribution calculation forces people to determine the model they’re using and to examine the (likely) converse to that model (e.g., FIRST for most companies.)

Christopher: Thanks. Your comment about Search and Analytics being different is a good one. It makes me think — and sometimes I think we forget this — about the relationship between Analytics and Search (and really the rest of the business.) This comes up in the big “Unique Visitors” debate as well … I think a lot of people forget that Analytics is supposed to be a service organization for the rest of the company, not an entity in and of itself.

This is obviously a longer conversation but the only answer to your question “what percentage error” is to ask the internal data recipients and to engage them in a conversation about the relative costs associated with getting better/different/more accurate data. It’s not to say the inevitable answer of “we want 100% accuracy” is good or even reasonable, but when I talk to companies about attribution models I wonder whether these internal conversations are even happening.

Thanks to all of your for your comments!

John Squire

Thanks for shedding some much-needed light on attribution. You say in your post that the lack of a solid campaign attribution model “likely means that hundreds of millions of dollars are wasted annually on marketing efforts that don’t produce their intended results.” In any other corporate function, that’s tantamount to gross negligence. Unfortunately in Marketing, the 50% is wasted axciom seems to be implicitly expected as rule of thumb with each campaign.

We know that impressions and clicks just don’t cut it – it’s kind of like trying to analyze or predict the weather in California based on what you know about the weather in Oregon. How many CMOs, CEOs and CFOs would fall out of their chairs if they knew their online marketing teams are investing dollars and resources based on guesstimates and gut feel? I know from speaking with leading marketers that have embraced multi-touch attribution that “there’s no going back to last click only attribution”. In fact, I’ve witnessed more than one Executive or Industry Expert take the position, while speaking at the same conferences we all attend, that Marketers should be racing to analyze and demonstrate how campaigns work in concert (first, last, and all clicks in between) across the Internet to influence website visitor acquisition, conversion and retention. Anything short of that is just not worth considering. More importantly, I think we all realize this isn’t going away, no matter how hard some marketers want to ignore applying quantitative instruments to their craft.

Getting the conversation started via your model of Appropriate Attribution should be an easy starting point for any marketer WITH just about any analytics solution.

Good post.

Matt Lillig

I couldn’t agree more. Especially when it’s been proven over the past 3-4 years that a combination of running display and search campaigns yields the highest return on investment. The problem though is that most advertisers don’t have the necessary tools to discover this ROI potential for themselves. Are my display and search campaigns receiving credit beyond last-click?

Without proper attribution metrics, advertisers base the success of their online campaigns (display, search, email, etc) on how well they convert. If their display ad converts well, they assume it’s working great and they keep throwing money at it. If their display ad converts at a low level, they assume it’s a poor performer and they pull back on the budget.

But just because a display ad does not convert well on its own, it doesn’t mean it’s a poor performer. In many cases, the poor coverting display ad is very valuable at driving conversions for an advertiser’s other ads such as search (as the studies have shown and data I have collected). But because advertisers don’t have proper attribution tools to give display ads credit (beyond a last-click direct response), they’re faced with having to make assumption based budgeting decisions.

I’ve discussed in the past how advertisers can save themselves money by using tools like Yahoo’s Assist attribution metric in their Yahoo! Search Marketing reports.

http://mattlillig.blogspot.com/2008/12/yahoo-assistsput-money-back-into-your.html

It’s really a no brainer once you understand the consequences of not using tools that provide attribution metrics. This is a hot hot hot topic in the industry right now and I’m glad to see that it’s getting more and more coverage.

Huayin Wang

Eric, thank you for your great practical thinking around attribution model! To josh’s point, attribution model should raise fire storm. There is no single reason why we are not seen this, but one reason I think is the lack of awareness, especially for decision makers.

On a related point, I do not think it is an easy problem. It is not difficult to understand the issue, but extremely difficult to figure out the right solution – Fermat’s Last Theorem for analytics? I also do not believe that WA tool vendors are able to provide complete solutions but certainly an important part of the solution. I have made my point elsewhere do not want to bore all here.

Web Analytics has been centered on metrics for so many years to the point that in many people’s mind, data analytics is all about metrics, or right metrics, or key metrics. This is sad.

The supply-side of Econsultancy’s Online Measurement & Strategy report | Aurelie Pols at Web Analytics Demystified

[...] another side of the crystal ball with their campaigns i.e. traffic acquisition and I’ll refer to Eric’s post about attribution as issues are far from being resolved in this space! I’ve witnessed it often: discussions about [...]

Add a Comment