Blog post

When Machine Learning Drives Marketers Bananas…

By Matthew Wakeman | October 14, 2021 | 0 Comments

…What’s a marketer to do?

Machine Learning and AI are taking off within marketing – the majority of marketers we surveyed are piloting or using ML/AI in production.

But most of these marketers (73%) don’t trust ML or AI with important decisions.

What’s driving that lack of trust? In most cases…

  • Marketers don’t understand how the model works.
  • They couldn’t explain it to someone else if they tried.
  • Data scientists’ explanations demand an understanding of how different algorithms work.
  • Data scientists use technical language that changes as frequently as the algorithms do.

Without a 9-month machine learning bootcamp (and 3 years of hands-on experience to learn the different algorithms), how can marketers build trust in machine learning and AI?

Marketers need a method for understanding and using ML/AI that they can use, without the math and complex language (Gartner subscription required).

Learn from Chris, the digital produce marketer in our research.

Chris needed to run a summer promotion for pineapples, and wanted to target physical BOGO (buy-one-get-one) coupons to customers that were likely to purchase (high propensity). She considered writing a set of business rules to identify those customers, but the data science team jumped in and mentioned a produce promotion model that they had used in the past.

Chris didn’t need to learn the ins and outs of model methods (regression, CHAID, random forests, k-nearest neighbors or deep learning, for example) in order to figure out if the data science team’s machine learning model for produce promotion would help her succeed in her promotional goals. She assessed the model in a two-part structured discussion with the model-building team, and identified key metrics for success of the model. That conversation covered four main themes:

  1. Marketing outcomes
  2. Data usage (and model validation)
  3. Model deployment (process and ease of getting it “to production”)
  4. Monitoring and Measurement (once launched, what would indicate you should reject it? Revise it? Rejoice about it?

She used the discussion guide for marketing conversations with a data science team (Gartner subscription required) to:

  • Better understand the model developed by the data science team
  • Easily and fairly assess whether the model will improve marketing outcomes
  • Gain consensus for model usage and improvement

The full story of Chris’ journey to math-free machine learning model mastery can be found here (Gartner subscription required): The Math-Free Method for Marketers to Build Trust in Machine Learning and AI Solutions.

So what ended up driving “Chris the pineapple marketer” bananas in this promotion? Cannibalization of other produce – bananas!

 

Leave a Comment