• by Dave Carpenter , June 24, 2011

"We know our digital advertising works," a client recently said, "because when we stop running it, our site traffic and revenue drop."   His company spends over $100 million each year on digital advertising; they are sophisticated marketers who want to use data to improve results.  If you ask him what the stacks of reports on site traffic, ad impressions, etc. he gets actually tell him, he'll tell you, "not much."  Why?  Because he and most digital marketers are accountable for getting results, they want specific, actionable information -- what to run with and what to change -- about creative, site and page placements, position, size, frequency, and other factors they can control.

Measuring web sites and online marketing campaigns has kept marketers, agencies, and others busy for years.  Web analytics report visits, page views, visits and other counts of site activity.  Ad server analytics report impressions served, clickthroughs and other silo-specific measures.   Once upon a time, each delivered the "hot metric."  But, as soon as we begin to understand what each "hot metric" tells us, we begin to understand where it falls short, and we begin again to search for the "Golden Metric" that we can confidently use.

The latest Golden Metric candidate is attribution.  Attribution recognizes that people are touched by advertising and content in many ways as they move from first impression to action.  It brings together ad serving, site and other data to identify sequences that represent the full multi-channel path-to-action and enables you to assign credit to the ads, search words, content and other factors that influence people to take action.  Because it can help you understand what contributes to online success, attribution is gaining acceptance even as it evolves.           

Attribution's promise is that it is moving us closer to the Golden Metric that will tell us which actions will improve performance.  Its focus on paths, enables you to see which sequences result in the most conversions, assign economic value to sequences, discover where viewers derail, determine how to shorten the path, and more.  And, perhaps most importantly, it enables you to know how to shift marketing resources to generate more of the most effective ones. 

Today some attribution providers offer "U" attribution, which gives weight to first, last, and intermediate views and clicks along the path.  "U" attribution is a significant step forward, but it isn't the final stage in attribution's evolution, especially when the weights are assigned "intuitively" rather than from patterns in the data.  If we knew before we started what the "right" weights were, then there'd be no reason to do any analytic work.  But, marketing analytics repeatedly reveal that our intuition for which campaigns, creative, placements, etc. will work best is often wrong when we see the data.

The rigor of determining the model weights from the data gives us more confidence in our ability to act on the results.  The best modeling approaches use campaign goals as input to model, recognizing that information gathering and buying processes differ for different products, services and ideas.  Our tactics and measures of success vary based on whether we looking to increase awareness or targeting highly qualified prospects, so our model should vary as well. And, they provide analysis of all the factors we can act on-breaking results down by placements, page position, creative, size, etc. 

It's not enough to know that our digital ads help drive business.  We want to know exactly what to do to improve results.  Getting reports by the ton doesn't help.  Moving beyond simplistic measures that don't accurately reflect complex interactions online is hard work.  Doing it right is far tougher than most consultants and analytics vendors are willing to admit.  It's only when we recognize the failings of our current approaches that we are willing to do that.  But, if we want a Golden Metric, we must demand data-driven, actionable information, not just the next convenient measure.