Marketing on the Web is a complex and difficult activity--compounded, one expert says, by too many analysts and marketers who follow what he calls the theory of MCU: "make crap up."
That colorful piece of advice was offered to Strata attendees Tuesday by Avinash Kaushik, Chief Education Officer and Founder of the Market Motive consulting firm. This would not, by any means, be the last bit of off-color humor that Kaushik would deliver in his discussion of the many common errors take when they approach big data as a solution to web marketing.
Specifically, Kaushik aimed his MCU comment at analysts who often apply their own personal agendas and standards to the many dimensions involved in figuring out how a customer goes from not buying a product to actually completing the transaction. In the past, the last click a customer made would be the one that was given all of the importance. Then, Kaushik explained, it was the first click which took all.
"This is like me giving my first girlfriend all the credit for me marrying my wife," Kaushik joked to the crowded session.
Eventually, all parts of the transaction were given equal weight when measured and considered, at which Kaushik also scoffed. Then, the importance of each dimension would be decreased over time, to which Kaushik grudgingly gave some approval.
"Maybe this is half-decent," he admitted.
But then came what seemed to be the worst idea if all: dimensions of a transaction would be arbitrarily personalized to a business by experts who think they know what they are doing. This practice, Kaushik declared loudly, was really these experts exercising their knowledge of the MCU theory.
While Kaushik was harsh on such practices within the web analytics and marketing sector, he was not above admitting that the sheer volume of web analytics data was a difficult obstacle for any marketer to face. And, he offered a solution.
"Is there something better than rear view mirror attribution analysis" he asked? "Yes! Media mix modeling!"
By performing controlled experiments on ad and media buys in different venues or geographic locations, Kaushik argued, you can get much better results than by trying to read the tea leaves to figure out what data means.
"Data is huge, fast, and free," he said. "It is cheap to have and display prettily."
As such, it means businesses should be prepare to change the way they arrive at decisions with big data.
Take decision making. Under the "old" methods, workers would find data, not be able to locate much data, hire an expert to find data, and then have the boss reject anything found and implement a decision based on his or her gut feeling.
Now, Kaushik continued, here's the new way: imagine all the data in the world, collect all the data, deal with processing that data, implement Hadoop, then Hadoop nodes 2-9, do some data recon, e-mail the boss, do something, then something... and finally do an action--maybe.
This is the standard approach to big data marketing and it needs to change. Bigger is not always best.
"Real time data without the capacity to take real time action is useless. The thing we need to seek is right time data," Kaushik said. If you really have to have real time, he added, then replace any humans in the process with an intelligent algorithm.
"That's the only way to have true real time," Kaushik said.
Kaushik's discussion was full of frantic, near manic, energy, but he raised some excellent points about how big data cannot be treated as the be-all solution, and that the right people were definitely needed to affect real solutions for marketing with big data.
"We need people to have this," Kaushik concluded: "People with skills in scientific method, the design of experiments, and knowledge of statistical analysis."