The very mention of machine learning provokes anxiety-generating questions such as, “Will it replace humans?” and “How do I trust the results?” and “Will I need to get a degree in statistics?”
We use machine-learning devices increasingly in our everyday lives. When you ask a virtual assistant, such as Siri, a question and receive an incorrect answer, you move on and don’t worry about how it works. However, in marketing, machine learning is used for high-impact decisions worth millions of dollars, and we need to have a lot of confidence in it before we adopt. Let’s open up the black box of machine learning to help reach that level of confidence.
Will This Replace Humans?
Humans provide the three Cs of machine learning: creativity, common sense, and context. A few years ago, two competing booksellers on Amazon were using algorithms to price their books, each looking only at what the other was doing. The price of one text spiraled upward of $23 million! There was no human guiding the machine’s decisions and no common sense regulating how much customers would reasonably pay.
Humans will always be involved in creating, guiding, and learning from the algorithms that assist machine learning. It is not human decisions that are automated, but rather, the process of generating the information that guides those decisions.
Where Do I Start?
Someone might not want to automate decisions based on customer behavior until they can combine all customer data and all Web-analytic data and all mobile data. It is a good thing babies don’t use this all-or-nothing logic when they are learning how to walk. Imagine a baby observing people dance the salsa and thinking, “Oooh, that looks hard. I think I’m good being carried.”
Don’t wait for perfection. Start with the dataset for which you have the greatest critical mass and the most confidence. Often, that might be Web analytics or CRM data. Use data you believe is the cleanest, the most consistent, and provides you with the best view of your customer.
Then, as you layer in more data — and as the algorithms grow smarter — you will see improvements in the overall confidence, precision, and accuracy of your predictive models. You will have insights that could influence which datasets should be brought in next or how to change your current strategy given the ever-evolving ecosystem of customer information.
Starting small instills trust and confidence and helps move you forward toward more advanced machine learning. And, in the process, you are going to set up your organization to be smarter, more sophisticated, and more effective — and that’s how you’re going to compete on analytics.
Rest assured that humans will always be the main players guiding marketing. Employing machine learning in the process instills greater confidence in highly targeted marketing campaigns and helps guarantee ROI while ensuring relevant, consistent experiences that delight your customers and prospects.
The post Demystifying the Black Box: Adding Transparency to Machine Learning appeared first on Adobe Blog.
