Man vs Machine: Which is more effective in consumer marketing?

Hype around the power of machine learning continues to grow, from medical diagnostics to autonomous driving, and including of course consumer marketing. But can machines truly be more effective than creative human marketers?

The answer is a definitive yes, but…

In order to understand where machine learning really excels, and in what areas a human touch is still required, lets analyze a typical promotional email sent by a supermarket chain on a regular basis to its loyalty club members.

In today’s retail environment, where data driven marketing has become the standard, most promotional emails are personalized to some extent. In its simplest form, that might merely include referring to the customer by name, and maybe giving her a breakdown of club related activity such as her points balance or progress towards certain rewards. A more advanced version might target some of the promotional content to certain customer lifestyle segments based on shopping histories, demographics, geography etc., or even offer items related to previous purchase history, similar to Amazon’s “ Recommendations for You”.

But more sophisticated retailers have been taking this personalization to the extreme, creating “segments of one” in which no two customers receive the same set of offers, even when there are literally tends of million shoppers included in the mailing. Taking into account the number of potential offers available, the number of offer mix permutations is simply mind boggling, and becomes even more complex when introducing multiple objectives for each email campaign. Such objectives might include for example:

– Stretching a shopper’s spend by tempting with cross or up sell offers
– Developing shopper loyalty by offering discounts on items the shopper will purchase in any event
– Encouraging the shopper to purchase in categories she has been avoiding or underspending in
– Optimizing the campaign on a chain-wide basis for top-line sales or gross profit

In most cases, retailers will try to meet some or all of the above objectives by creating a blend of offer types to address each goal, but what’s the optimal mix per shopper?

All of this requires an ongoing and in depth analysis of the shoppers ever developing transaction history, itself a moving target as shoppers continue to purchase multiple times per week via a range of online and offline channels.

What’s more, such emails might be delivered multiple times per week, month after month and so care must be taken to avoid being repetitive, and also needs to take into account seasonality, purchase frequencies per product and much more.

And then there’s the issue of budget control – what’s the optimal spend per shopper or campaign to ensure the highest possible return on the marketing investment?

You get the point.

It’s in this type of high frequency, high complexity consumer environment that highly automated machine learning is imperative. Purpose built applications are designed to find the optimal combinations on both a shopper and chain level, running billions of iterations at breakneck speed to eventually settle on the most profitable campaign outcome. In addition, algorithms are trained to continuously monitor shopper reactions in real time to each and every decision point, learning from them in an ongoing loop in order improve the outcome in future campaigns. In reality, machine learning will beat human marketers every time in situations described above, achieving higher ROI’s, faster, and at lower cost.

So where is a human touch still required?

Emotional and aesthetic elements are best created by humans at this point in time – writing an intriguing email subject line and soft copy and visual design including colors, fonts and layout are all still currently the domain of talented copywriters and designers. There are still some things that a computer just cannot do as well as humans.

But even that might be changing. Armed with a number of different options created by humans, machines can now test for visual and emotional appeal by monitoring response rates and can optimize accordingly. Will we see even copywriting and graphic design become the work of computers in the near future?



Book A Demo