Who knows what is best for your customers? You or a computer algorithm?
That is the question that big data has raised for many marketers. And like all great questions, the answer is, “It depends.”
I’ve wrestled with the question for a while, because I appreciate both sides. There’s no doubt that tremendous value can be unlocked from all the data that customers emit these days. Digital pheromones that drive marketers wild, if you will.
Computers are capable of analyzing such voluminous mountains of data and, if we let them, acting on it within milliseconds. How could we humans ever expect to compete with that?
On the other hand, customer experience is hard to get right. In principle, it’s simple: give people what they expect, when they expect it. But the “what” and the “when” vary tremendously depending on context — who the person is, what they were doing at the time, the thoughts and emotions that color their perceptions.
It’s a whole that is often greater than the sum of its parts.
User experience (UX) professionals hone their craft for years to design and execute such compelling customer experiences. They are guided by assumptions, perceptions, and intuitions that are hard to bake into a quantified algorithm. Although admittedly, the good intuitions are called “instincts” and the bad ones are called “biases.”
So which is better, the computer or the human brain? Well, it depends.
But I think I’ve finally figured out what it depends on.
Expected Experiences Vs. Serendipitous Content
The world is divided into two kinds of marketing touchpoints: those that customers expect and those that are serendipitous. It’s this division that forms the basis for the following algorithmic marketing matrix:
Expected experiences happen when a customer consciously interacts with you. They click on an ad and go to a landing page. They browse through your website. They make an e-commerce purchase. This applies to offline interactions too: they call you on the phone, visit your retail store, or consume the service you offer.
Serendipitous content, in contrast, means pleasant surprises. It’s an ad that someone happens to notice that triggers their interest. It’s a recommendation that appears in an e-commerce store or the right rail of a publisher’s site. It’s an offer in email that “coincidentally” resonates with someone at just the right moment.
When it comes to expected experiences, customers are not particularly forgiving. If you don’t fulfill their expectations, they will feel let down, and they will hold that against your brand.
In those high stakes scenarios, a human marketer’s supervision over the user experience is greatly preferred.
With serendipitous content, however, the balance shifts. There are effectively unlimited opportunities for ads, recommendations, and cross-sells to pepper the digital landscape. But the stakes involved in any one such touchpoint — to the marketer and the recipient — are relatively low.
In those scenarios, automated algorithms are more cost effective than hand-tuned human direction. You can roll the dice, because you have little to lose and potentially much to gain.
Deterministic Marketing Vs. Probabilistic Marketing
Another way of thinking about this is to consider whether a marketing touchpoint is deterministic or probabilistic.
A deterministic experience will always be delivered in a consistent, predictable way — exactly as it was designed by the marketer. However, when we use machine learning and predictive analytics in automated algorithmic marketing, we’re exposing the customer to a probabilistic experience.
The following graph, from an article I wrote on marketing automation and experience design, helps illustrate this relationship:
Letting machine learning algorithms probabilistically hone in on the “best” content (best in a statistical sense) to show people at serendipitous marketing touchpoints makes a lot of sense. But when a customer expects a specific experience, a more deterministic approach is better.
Mixing Expectations & Serendipity
Back to the question of which is better in marketing, the computer or the human brain, the real answer is: both. Each applied in the context where it is most effective. And in many cases, they can be applied together.
For instance, data-powered algorithmic capabilities, such as machine learning and predictive analytics, can certainly contribute to delivering better expected experiences. But they’re best used behind the scenes, to help marketers identify new segments and scenarios. When it’s time to deliver an experience based on that insight, however, the marketer retains control.
In addition, in a well-designed customer experience, there are often many opportunities to expose people to relevant serendipitous content. As long as expectations are being met with the primary content of the experience, incorporating secondary content — ads, sidebars, recommendations, cross-sells, etc. — can leverage one touchpoint into several.
Serendipity is great — as long as it is never at the expense of fulfilling the promise of an expected experience.
Opinions expressed in the article are those of the guest author and not necessarily Marketing Land.