Realizing the dream:
how to achieve accountability in internet advertising

By: Amir Yarkoni, Seperia (easynet), October 2009

The internet brought with it promise of a new kind of media for advertisers: one that would be transparent, traceable, and accountable for all advertising expenditure. Theoretically, ad dollars would be translated to clicks, clicks would then be tracked down to conversions, sales , and even subsequent dollar value of acquired clients -- closing the spend-return loop to achieve a full ROI picture.

However, 15 years past the dawn of the internet, during which time huge technological advancements have taken place, we are still far away from fully realizing this vision. Why?

In this article, I will outline some of the main complications, and propose directions advertisers may consider to get closer to the original promise, and to advertising optimization.

Main challenges

1. Attribution. Which traffic source is to be credited for a conversion? Oftentimes, a user is converted after arriving multiple times at the website via diverse sources. Should the "last click" be fully credited as the source, disregarding the prior ones? Why not the first click? Or split the credit between them in this way or other? How far back should clicks be considered as part of the "click stream to conversion"?
   
   These are not easy questions to answer, and they strongly depend on specific business characteristics. Add to the equation a multitude of technical variables related to the tracking mechanisms themselves. The core issue of attributing credit to click sources is all but straightforward.
    
2. Granularity. As new advertising formats continually roll out and are refined (especially by Google) which is the "granular unit" to be attributed a success or failure and optimized accordingly? Is it a keyword, a text ad or a landing page? Is it an ad placement or banner creative?
   
   The answer is, of course, their combinations. However, analysis of this multi-dimensional environment creates a massive dataset that requires a far more complex decision-making process, and a highly powerful analysis tool and staff.
    
3. Untracked influences. Offline and online advertising do not live separately. People exposed to brands offline will interact more with the brand online, and vise versa. But the available tracking tools tell us a partial and disparate story: an online transaction generated via a certain tracked click, or a transaction at the brick and mortar cashier that just occurred.
   
   Moreover, online display advertising and social media may influence brand perception and attraction, which would not translate directly into incoming clicks, and therefore are not tracked. However, they could have influence on increased brand related searches, increased CTRs for the brand ads, and ultimately more conversions. We have seen multiple cases during periods of intense brand-related advertising activity where conversions have increased significantly. When brand advertising slows, products return to lower interest and conversion levels for branded clicks.
    
4. Lifetime client value considerations. Not all clients were born equal, as every salesperson knows. However, most direct response advertisers work by a "cost per client acquisition" (CPA) target. This means that they are willing to pay up to a certain threshold for acquiring a single client.
   
   CPA value is usually derived by predictions of lifetime client value. However, that client lifetime value may differ greatly from theme to theme, country to country, and a host of other parameters. Erroneous campaign management decisions could be made by relying too much on across-the-board CPA targets.
   
   Another challenge, lifetime value is dynamic and changes over time, more data is accumulated, and prediction models are updated and refined.
    

Results

At the moment, many advertisers, even some of those spending millions of dollars on internet advertising, do not have the technological tools nor business rules and workflows required to successfully meet these challenges. Thus, they run partially tracked and often under-optimized campaigns.

So, how to optimize?

Prior to outlining some of the necessary steps required to meet the challenges described above, there is an important mapping pre-step. Before diving into technicalities, define the characteristics of your online business, prospect characteristics (or personas), projected touch points with your brand on-site and off-site, estimated length of decision cycle, main sources of traffic that get them to your site, and your brand position, properties, and activities.

After defining the nature of your ecosystem:

1. Define business rules and workflows for setting targets, determining ROI, and assigning credit to conversions. Don't be too rigid, you will need to constantly refine them. But it is important to have a methodology in place.
    
2. Select the technological tools that can meet your requirements. The baseline will be a solid web analytics system, and perhaps other systems for more specialized tasks such as SEM analysis/management and testing. Make sure those systems can meet your selected attribution model. Another thing to look for is system ability to analyze a multi-dimensional environment at deep granularity level.
    
3. Implementation and adoption of the tools are critical because these tools will not help you so much right out of the box. They will need to be adjusted and customized to your needs, and adopted within the working team. It is also important to integrate your backoffice or CRM data into the analysis, because generic conversion tracking is often not enough to make educated campaign decisions.
    
4. When designing advertising campaigns, make sure to design and enforce a workflow for assigning unique tracker IDs in a pre-defined granularity level -- granularity that is different in each advertising format.
    
5. The art of segmenting is at the heart of setting targets. Try to achieve segmentation of clients in a way that will be most helpful for optimization. Over-segmenting will not give you sufficient statistical data for calculations and will overload campaign management. If you under-segment, you risk seeing too rough a picture of activity and will be unable to optimize particular segments. This results in and overall mediocre performance.
    
6. One of the most challenging tasks that advertisers face is predicting client lifetime value in a way most useful for optimizing bottom line. This is not a purely mathematical or academic subject. It does require the math, but together with some holistic assumptions and thorough understanding of your online business environment.
    

Summary

Honestly, you should not hope for 100% optimization of internet advertising -- the number of variables is just too high, and behavior too dynamic. However, the higher your ability to combine a viable advertising model, and a holistic marketing approach, together with robust tracking and analysis - the closer you get to optimizing your advertising, and achieving higher advertising accountability and ROI.