Why Micro-Segmentation Matters in a Post-COVID World

Author: Ranjith Ramachandran

A food chain cooks up conversion with personalized offers that sizzle


This article is based on a recent TDWI webinar, “Microsegmentation: What Is It, And Why It Is Important in a Post-Covid World,” sponsored by Talend and Databricks (June 3, 2020).


If you are a brand marketer or member of a customer analytics team, you might already use micro-segmentation as a way to reach and engage your best customers efficiently. If you don’t, let’s explore why micro-segmentation matters, especially now – a time when bringing back consumers is crucial to your company’s rebound and future growth. 


What is micro-segmentation?


Micro-segmentation is the process used to create hundreds or thousands of specific groups of customers, enabling companies to interact with them in more customized and personalized ways. It breaks up large customer clusters created from traditional macro-segmentation techniques, which group customers based on geographic, demographic, behavioral, or psychographic attributes. Usually, the goal is to uncover which type of customer spends the most money, so targeted communications and offers can be made directly to them.


Micro-segmentation is more granular than macro-segmentation. It combines attributes to pinpoint smaller subgroups of even higher value. Using a hybrid of factors is the best way to identify who your best customers are, how they behave, and what matters most to them. This involves bringing disparate data sources together, then applying sophisticated analytics to identify not only who spends the most now, but who likely will in the future, too. This is an example of how predictive analytics and machine learning are playing more important roles within companies.


Predictive analytics and machine learning go mainstream


Most companies seek to improve their decision-making with better insight. Now new research highlights the relevance of deeper analytics: according to TDWI, a majority of corporate analytics departments say they want to increase their predictive analytics and machine learning capabilities.


Among 109 respondents of TDWI’s 2020 Data and Analytics Survey, a resounding 71% say demand for deeper analytics has increased internally, up from 64% last year. Putting this demand into action, about one-third of respondents say they are actively building predictive analytics and machine learning models within their organizations, with at least some projects underway. Common use cases for advanced analytics include customer churn, behavior and sentiment analysis, next-best action, recommendation engines, targeted marketing/advertising, personalization, customer experience, and more.


COVID-19 changes everything, including analytics


The business slowdown resulting from Coronavirus has given corporate analytics teams a chance to begin taking action on strategic data initiatives. But Fern Halper, TDWI’s Vice President and Senior Research Director for Advanced Analytics, cautions that while teams may be excited to dig in, the business landscape is no longer the same. “Teams are being asked to help organizations tune their analytics to fit the changing circumstances,” says Halpern.


To Halpern’s point, more than 50% of respondents say they are being tasked with answering new kinds of questions based on COVID’s economic impact on their company. TDWI’s survey also finds that about 30% need to update their models due to changing customer behaviors, and another 30% are required to add new data sources.


“We’re hearing that it’s more important than ever to understand the behavioral changes of your customers,” says Halpern. “Machine learning is used for grouping customers into specific segments based on common characteristics, enabling targeted communications and personalization. So even if you do segmentation now, it’s a good time to revisit this.”


What micro-segmentation looks like for a major restaurant brand


Recently, Wavicle Data Solutions helped a restaurant chain successfully segment its customer data into sub-groups that share behavioral characteristics with other attributes. First, we broke the total customer universe down into two groups:  


  1. “Known” customers, for whom some level of personal data exists
  2. “Unknown” customers, who are anonymous – no personal data exists


Focusing on known customers, the biggest group is people who had ordered food using the restaurant’s mobile app. We then segmented this group into additional clusters such as:


  1. Customers who began ordering food online recently, and then bought again
  2. Customers who ordered food multiple times in the recent past
  3. Customers in a continued relationship with the restaurant chain, who have not purchased recently



When looking at these groups in combination with other attributes, we found sub-groups of higher value customers – say, for example, mobile app users between the ages of 25 to 40 years, located in Dallas, Texas, who bought meals for kids at least twice in the last 30 to 60 days. We also looked at food item combinations, seasonality, and more to explore customer groups who are more likely to purchase in the future based on past behavior.

Details about each customer’s digital experience, such as POS transactions, mobile app orders, digital analytics, clickstream and social data, and customer loyalty incentives were ingested from batch and real-time data sources using Talend RealTime, then stored in a customer data platform built on Databricks Unified Platform.

Here predictive analytics and machine learning created both macro and micro-segments of customers, with matching customized offers for each audience. At this point, the campaign management team was ready to execute by sending personalized offers via the restaurant chain’s mobile app. Opens, coupon redemptions, survey feedback, and other responses were measured, then fed back into the customer data platform for further analysis and customer segmentation.

The illustration below represents our four-phased approach to testing personalized vs. mass offers based on finely-tuned customer clusters. In phase one (“Crawl”), we started initial testing with 1,000 known customers. By phases two and three (“Walk” and “Run”), personalized offer testing increased from one million to five million active digital customers. Finally, in phase four, our client was ready to implement customized offers in all markets at scale based on their micro-segmentation process (“Fly”).



Dynamic content and personalization beats “one size fits all” marketing


Sending coupons to everyone in your database is a time-consuming, inefficient, and expensive strategy. While mass marketing tactics may generate some results, “one-size-fits-all” marketing is outmoded given the rise and accessibility of digital engagement, dynamic data sources, and advanced analytics. When trying to reach your target consumer, would you rather crawl or fly? As your company’s analytical capabilities sophisticate (or if you decide to work with a trusted partner), you’ll evolve your mode of marketing for better competitive positioning.


But here’s the key: identifying subgroups through micro-segmentation isn’t enough by itself. Marketers must tailor their communications and offers to match the behavioral preferences of high-value groups. By doing so, you can expect results including:


  • Improved campaign performance
  • Increased upsell and cross-sell 
  • Higher campaign conversion rates and ROI
  • Stronger customer loyalty 


Micro-segmentation is the next step for businesses looking to execute more intelligent marketing campaigns – especially in a world that’s slowed to a crawl. Once you get started, you’ll see how dynamic content and personalization can make a measurable impact on your marketing outcomes and ability to soar in the future.


Watch our Micro-Segmentation webinar with TDWI and Talend to learn more.

Watch now