Reading Time: 3 minutes

From an online shopping site to product emails, if your offers aren’t personalized, you’re leaving an opportunity on the table for one of your competitors.

This week, I’ll outline a few ways with which businesses can offer simple personalization, using methods that scale and are easily implemented. By “simple” personalization, I mean methods that only require counting past events – there’s no matrix factorization or neural networks required! To achieve such personalization, there are three categories to look at:

  1. Onboarding
  2. Browsing/Shopping
  3. Marketing and advertising reach-outs 

Onboarding is the process of collecting data from your customers before they shop or browse your site. It’s hugely important because user preferences can and should dictate how you sell to each of your customers. And onboarding can be easy to implement with all the customizable survey tools out there these days.

To assure you’re getting the most out of your onboarding process, ask good questions that help you polarize your audience. Polarize? Yes, if everyone (or no one) chooses the same answer to a question, it doesn’t help you at all. Aim to ask questions that split your customers in half. Gender is a great place to start. Then aim for other areas where you feel your customers are really divided.

When you have a large number of users who have answered these questions, and the questions do a good job of separating your customers into groups, it’s trivial to build personas and use much more tact when marketing or advertising to them.

When a customer is browsing your site, that person expects a customized experience. Maybe not in the very beginning, but after browsing a few pages, users want a tailored experience. This is why many sites use your browsing behavior to offer more-personalized products.

The simplest way to do it? Market Basket Analysis.

Market Basket Analysis involves looking at large collections of browsing (or purchasing) data and answering the following question: for a given collection of items, what ELSE was generally browsed with these? As I understand it, this type of analysis came from looking at baskets of goods purchased together at supermarkets. Think about Market Basket Analysis like this: Buying cereal and bananas? Great. Market Basket Analysis helps us ask the appropriate question: Got milk?

As a more relevant example, let’s say Sarah browsed a red cashmere sweater by Vince, a pair of black distressed jeans by G-Star Raw and a grey tee by Ksubi. To provide personalized recommendations for Sarah using Market Basket Analysis, we:

  1. Collect all the times those 3 items were also browsed together, and
  2. Report on what else was browsed, and then
  3. Offer those additional products to Sarah

We’ve seen this technique increase engagement by up to 30%. This is significant. 

Moreover, Market Basket Analysis is a very straightforward method to understand and implement because it’s based on counts of previous events. We highly recommend it.

When sending marketing emails, please don’t send the same message to everyone! You’re certain to really resonate with just about no one.

Reach-outs are a great time to revisit the onboarding data and the Market Basket Analysis from customer browsing or purchasing behavior. Once revisited and analyzed, use that information to personalize your offers at scale.

While these are rather simple examples of personalization, our clients see double-digit increases in engagement from their implementation.

We love talking about personalization and if one of these areas is particularly interesting to you, contact us and let’s talk about how some of these tools may be helpful to your business!

All the best,
Zank and the Bennett Data Science team

Of Interest

3 Must-have JupyterLab 2.0 Extensions
JupyterLab has matured into a very useful tool. Here are some big-time extensions you or your team should be aware of:

Springer Released 65 Machine Learning and Data books for Free
Springer has released hundreds of free books on a wide range of topics to the general public. The list, which includes a total of 408 books, covers a wide range of scientific and technological topics. This post lists all the books (65 in number) that are relevant to Machine Learning.

How Satellites and Machine Learning are Being Used to Detect Plastic in the Ocean
Plastic is terrible for marine life and detecting plastic pollution in the ocean is notoriously challenging. Plastics come in many colors, break down to microscopic sizes, and are made of a variety of chemicals. Adding to the problem is the vast size of the ocean, to which millions of tons of plastic are added each year. To effectively target cleanup and pollution prevention efforts, it is essential to identify which parts of the ocean collect the most plastic, and satellites bolstered with machine learning are showing promise in tracking plastic pollution.