Last Wednesday, Stefan Tornquist (VP Research, US at Econsultancy) moderated a lively discussion on Measurement, Analytics, and Attribution that quickly maxed out the attendance capacity of the
I was joined by digital analytics gurus Jim Sterne (founder of eMetrics Summit & Digital Analytics Association) and Tom Cunniff (founder of Cunniff Consulting) to discuss what marketers need to do beyond gathering information, and how to apply measurement and analytics to strategy across the business.
A lot of important points on the current state of attribution and what’s in store for the future were discussed during the first half of the hangout, which are summarized here…
ST: Where are we now as an industry?
We’re at a point in digital measurement, analytics, and attribution, where the difference between being able to measure everything and being able to use it is paramount.
The trouble is that useful metrics vary brand by brand. Without Best Practices, marketers must decide which metrics matter to their business and which metrics don’t.
ST: How is performance data being counted, how is success being defined?
Leading marketers are using both ‘top down’ and ‘bottom up’ attribution strategies to evaluate and adjust their media mix either quarterly or yearly.
With ‘top down’, marketers look at their spend buckets, measure the results, and adjust spend. ‘Bottom up’ describes full-on digital attribution, where marketers place value on clicks and assists toassess performance on a more granular level.
Did the click create awareness, persuade the consumer, or build loyalty? Ultimately, how much did the click contribute to sales?
The vast majority of marketers use the last-click attribution model, awarding total credit to the last channel the consumer interacted with before purchase. The problem with the last-click model is that doesn’t represent the full view of the consumer’s path to purchase and is often misleading.
Tom Cunniff dismissed the idea that consumers are Pavlovian, where “we can give a single stimulus, get a response and measure it.” If only it were that simple!
Sure – the ‘last click’ is often what enticed consumers to purchase, but the last click may also represent a coupon redeemed from a customer who was brought to the site by another channel or a branded search made to relocate a product the consumer was already intent on purchasing.
Attribution is fuzzy, correlation can be made, but causation is often unclear.
ST: If most marketers who are doing attribution use last-click and 75% of purchases only include one click, are we too hard on this metric?
Last-click is a very useful metric, but it’s often the only metric used to measure attribution. The singular view is what introduces a ‘lower-funnel bias.’ Last-click is more useful as a place to start measuring campaign performance.
Smart attribution tools offer a wealth of models (first-click, weighted, best-click, and more) for marketers to play with and gain a more encompassing picture of what’s really working and how.
When it comes to choosing the best attribution model, Jim Sterne summed it up in a George E.P. Box quote. “All models are wrong, some are useful.” Jim reasoned “there is no single source of truth, rather a lot of data that is directionally useful”.
Different attribution models allow you to look at data in different ways, deduct insights and make meaningful decisions.
ST: How do you address the gap between online and offline, and between devices? Does the future of attribution involve bridging these gaps?
We’re looking at a more connected future of attribution, where mobile, tablet, and CRM data is integrated with unified digital data to offer a more complete view of the “customer experience” across devices, online and offline.
Brands understand that the customer experience is unique to channel, their media consumption and interaction differing on mobile, on tablet, on web, and in-store.
Marketers are collecting cross-channel information to map out the customer experience from point of awareness to persuasion in order to build user segments and personas based on individual behavior. The next step is applying unique attribution models to these segments and personas.
Getting ‘one to one’ with media mix modeling may be closer to reality than you think.
See the full discussion on Econsultancy’s Youtube page.
The second half of the discussion featured a debate on rearranging organizational structure to optimize media effectiveness, a deeper dive into one-to-one digital marketing, and Social Media analysis. We’ll have the summary of part two up shortly.
Econsultancy’s next Hangout will be on the subject of channel integration. This takes place tomorrow.