Interaction Optimization vs. AI: A Three-part Series
Part 1 of 3
Why Experimentation Is the Past and the Future of Marketing
I’m always amused by the hype that surrounds the introduction of any new technology. Regardless of what it is or does, inevitably, someone will claim that it will invalidate everything that came before it. Or even better yet, that it will eliminate the need for human intervention altogether.
Even The New Yorker is lamenting the triumph of math and machines over the advertising men of Madison Ave.
The truth, however, is always more complicated – and interesting.
What’s got me smiling recently is all the talk coming out of martech about artificial intelligence (AI) and machine learning (ML) and how it’s going to revolutionize marketing as we know it. While AI and ML are certainly not new, their application to the classic marketing problem of engagement and optimization is NOT. And frankly, it’s something we’re all, really, just starting to get our heads and arms around.
What AI and ML can do is bring a specific set of fine tools and automation to a universe of problems marketers heretofore tried to solve with manual, blunt-force instruments.
Experimentation v. Individualization Shootout
As someone with over 20 years experience with marketing automation and predictive analytics, I know too well the value of interaction optimization tools like heat maps, A/B tests, and site statistics. What I’ve come to realize is that while, yes, many of the tools I’ve used for years are now are too slow and manual or cumbersome to keep pace with the breakneck speed of today’s commerce, the strategy and techniques behind them are still just as valid today as they’ve always been.
Looking back, the insights we gained through interaction optimization may seem quaint and underwhelming today, but at the time they were quite extraordinary. I still recall my first heat map and being able to see what people were looking at and how they were moving across the page and site. It felt like a quantum leap, where we were finally homing in on the individual, and going far beyond the generalizations we could intuit from just A/B tests and site analytics.
Given where we came from, and how fast the technology evolved, we could be forgiven for the exuberance, and maybe even the thought that we’d come as far as we could. But of course, it was really just the beginning – the starting point on a continuum that leads directly to where we are today with AI and ML.
If you acknowledge interaction optimization and AI as both being on the same line, as instruments of experimentation, the false dichotomy of choosing one versus the other falls away and what you’re left with is how do you leverage each to get the most out of both.
Man Versus And Machine
So, instead of man v. machine, you have man AND machine. On one side there is interaction optimization and these tools of insight that are still very valuable, but also very manual and, by nature, imprecise. And on the other, there is the automation and fine-grained precision represented by artificial and machine intelligence.
Back in the day, we were always on the hunt for anything that could move the needle. We played with everything. Navigation. Layout. Buttons. Bigger buttons. Blocks of content. We ran A/B tests on it all. Eventually, it produced the insights we needed, but it was time-consuming and inefficient, and, ultimately, the results were ephemeral.
What we can do today with self-driving AI and ML is light years beyond anything possible with manual tools. Instead of testing static treatments or big blocks of content, we can now break it all up, mix and match text and images, with unlimited variants of each, and start to exploit the winning results – all in a fraction of the time it took to run a manual test.
But the machine has its limitations, too. Technology has come a long way, but it still can’t match our capacity to create or derive insight from experience or even simply lay out text and images. Absent a takeover by Skynet, machines still operate in a human-run world, where decisions about strategy and what to prioritize or even which segment to target are made by people with real-world business and brand imperatives – and problems.
Which brings us to the crux of the issue. What we’re essentially doing with AI and ML is asking a machine to solve fundamentally human problems. In the next blog, we will identify the three biggest problems facing digital marketers and how man and machine can work together to overcome them.