For a long time, one key to being a successful digital marketer was sheer doggedness. Across the industry, conference rooms were crammed with exhausted people, impossible-to-read whiteboards and stacks of Microsoft Excel printouts.
More sites, devices, tactics and algorithms in the system just meant there were more things you had to check and double-check.
A generation of "helicopter marketers"
In the process, the industry raised a generation of helicopter marketers. Like helicopter parents, it was their job to hover over every marketing effort to ensure that it was doing the best it could. Initiatives were set up, observed and then obsessively tweaked. The irony is that, in many organizations, program results were being carefully hand-groomed to demonstrate the steady, unrelenting improvement that management demanded to see from an "automated" system.
Somehow, the marketing world became one with teams of people working for computers, when the computers should really have been working for people.
Today, advertising and marketing people can focus on bigger, more satisfying challenges as the enormous workload of optimization shifts over to computers.
Letting machines do the learning
When all a digital marketer's experience tells him or her that the best way to make progress is to helicopter, it's hard for them to let go and let the machine do the work.
It's not about stubbornness or professional pride. It's just that people have ingrained habits that exist for perfectly good reasons. Just as you might feel somewhat wary about not having your hands on the wheel of a Google self-driving car, it's hard for digital marketers to adjust to, well, not adjusting.
But to update a famous quote for the twenty-first century from management consultant Peter Drucker, in a machine-learning system, most of what we call "management" consists of making it difficult for the algorithms to get their work done.
Get the best results from machine learning
The best way to learn is to build systems that don't assume they already know everything. There is surprising power in not just allowing, but enabling machines to learn from data.
Hear from Fortune 500 brands that have been forced to pivot as consumer preferences evolve, as well as entrepreneurs building brands from scratch to meet new consumer needs. This event peels apart the layers of brand building with a carefully crafted roster of top marketing, technology, and creative leaders.Learn more
As marketers have always done, it's smart to start with some theories and test them. However, setting up rigid constraints in an attempt to completely eliminate waste at the outset is actually counterproductive.
It's tempting to force a traditional way of segmentation and to begin with some answers already in place, but it's actually much better to use desired segments as a starting point for learning. By relaxing the constraints, you can let the bidding algorithms find many more bidding opportunities and identify untapped opportunities, which can slash acquisition costs by half in some cases.
Don't game the system
Mark Twain said, "There are three kinds of lies: lies, damned lies, and statistics." In any complex system, anyone can find a way to fudge the numbers so they look like a genius. Marketers who once smoothed and groomed results for management to build confidence in digital are now finding greater success by letting the algorithms do their job. Learning -- real learning -- is by nature surprising and insights-yielding. A "no surprises" once-a-month PDF report is unsatisfying for everyone. The secret to success in this business is now about actually getting smarter, not just looking smarter.
Successful companies understand that optimizing for ROI is very different from optimizing for hitting a set of KPIs or for achieving the lowest possible costs. They also know not to settle for a lowest-common-denominator metric when testing multiple partners.
Instead, the key is to stay focused on what the real goal is.
Toshiba took a longer view to market high-end laptops. In doing so, they not only met their 5:1 return on ad-spend goal, but surpassed it with 8:1.
Getting even more creative
It's only natural to think that, as the expert on your brand, you can pinpoint the creative ad that will perform best, but it pays to stay open to learning. Programs have seen as much as a 20-fold performance difference between the best ad and the worst one. Smart marketers test a wide set of variations. Remember, you can always polish the winning ad later.
Getting stronger creative from computers requires an interesting balance. You need just enough rigor up front to give the algorithms enough to learn from and just enough flexibility to allow the algorithms to discover the optimal outcomes on their own.
Let the robots do their job … and have more fun at yours
Let's face it: as people, we're prone to non-optimizing forces like bias, emotion and human error. Therefore, the natural tendency is for us to overcorrect our programs. Like over-steering a car, this approach will probably end up with you in a ditch.
If you keep switching goals, it will only confuse the underlying algorithms, rendering them completely useless. However, robots are experts at doing just enough to keep the program on its most productive course. Believe in them. A little trust goes a long way.
Like many things in business, setting up programs for success is about keeping your focus on the most important things and getting the fundamentals right. The best part of letting go of the optimization details? Marketers get to focus on the bigger, more human challenges facing their brands. When marketers stop helicoptering, their marketing results can soar.