The promise of Internet marketing has always been about measurement.
“It’s the most measurable medium ever!”
“We can make minute measurements of individual behavior!”
“We will finally know which half of our advertising dollar is being wasted!”
But we got it wrong right from the start. That first statement is not true in the least. The Internet may be the most measured, but not the most measurable.
Direct mail is much more measurable.
Direct sales are much more measurable.
Telemarketing is much more measurable.
Infomercials are much more measurable.
Yes, we have more data than ever before, but we’re measuring whatever we can rather than what we should. Click-throughs, page views, and conversion rates are the results of working with the data we have rather than what is important to the enterprise.
We found this wonderful, shiny data shovel and we’ve been digging ourselves deeper and deeper into a hole of improper expectation. We succeeded in getting our online budgets by giving the impression that online advertising and marketing are exact sciences.
We have fervently embraced every new technology and every new data set and hailed them as being so much better than sliced bread.
Advertising clicks delineate impression impact.
Search keywords reveal actual intent.
On-site behavior exposes individual desires.
Social media sentiment delivers exact brand awareness and sentiment.
It just ain’t so.
A last week’s eMetrics Summit in Dusseldorf, Rory Sutherland, vice chairman, Ogilvy Group U.K. brought the point home in his inimitable fashion.
He said, advertising and marketing are not exact sciences. There is not a single answer that is demonstrably the right answer to a given advertising or marketing question. There is no best creative, best list, best medium, or best offer. Humans are not rational. There are too many variables and too many permutations. You can’t tabulate what a person might do with anything that even approaches precision.
And as David Weinberger likes to say, “The universe is analog, messy, complex and subject to many interpretations.”
And while we do not have the luxury of working with constants and laws and certitude, we can still find a great deal of value.
We just have to set the proper level of expectation and get the budget keepers to understand what Bob Page, VP of analytics platforms and delivery at eBay means when he says we are not accountants; we are statisticians. We do not count beans; we deal in probabilities and likelihoods.
If you are comfortable with uncertainty, then you can take the metrics to their logical conclusion. This was the topic of the eMetrics Summit’s sister conference, Predictive Analytics World. That’s where data scientists and higher math types rub shoulders with those who are trying to apply said equations to business problems.
Machine learning and big data heavy lifting do not depend on certainty, but on approximations. They are in the business of building mathematical models of our world in order to make a better guess at who is most likely to buy, who might commit credit card fraud, and what’s the best price spread for insurance plans.
They do not create a documentary photograph of the world – but a model. And as George Box quotably wrote, “All models are wrong, but some are useful. “
Whether you are trying to get more people to engage with your brand, make a bigger splash with the best possible tweet, or increase sales on your website, the miraculous, modern, digital computing machine can extrapolate ad infinitum. But it cannot create the model out of thin air.
Humans are required to decide what ingredients should go into the recipe before the computer oven can cook the results for us. Knowledge, cognition, ingenuity, creativity, and inspiration will always be in demand.
Senior managers will value those traits more than huge data sets once they follow Sutherland’s imperative to use data as an aid to thinking, not a substitute for thinking.