AtData logo

Maggiano’s Little Italy Turns Data Into Engagement

Apr 14, 2022   |   4 min read

Knowledge Center  ❯   Case Studies

A name and an email address … times millions. Maggiano’s Little Italy started collecting this information about guests back when email loyalty programs were in their infancy. Back when the theory was that you didn’t ask people for more than their name and email or you’d never get them to enroll. Back when response to email marketing was measured in double digits.

The Challenge

Maggiano’s committed to developing Direct Marketing as their primary channel for sales lift and new guest acquisition. They used multiple sources of data, targeting analytics, many rounds of targeted direct mail, and careful analysis of results to develop a vigorous and successful feeder program. And yet, the Maggiano’s/Marketing Informatics team knew more about prospective customers than we did about the Loyalty Guests.

It was time to change that, and AtData was the trigger.


Maggiano’s decided to morph their email list ihto a marketing database, and evolve their CRM communication from couponing to engagement. That would be a multi-step process:

Step 1: Append data with AtData Email Intelligence

The problem we’d faced in appending this data before was that we needed postal addresses for the match back to the enhancements and only a very small percentage of the file had postal addresses. That’s a real problem today with email loyalty clubs and traditional enhancement sources.

To solve this problem we turned from those traditional sources to AtData. AtData has the capability of appending enhancements using email addresses as the matching link rather than postal address. This is a major evolution in the information industry. And while some other providers have similar capability, AtData was the most comprehensive source. Plus, the mechanics of the append process were very simple and their customer service was spectacular.

So we appended AtData Email Intelligence fields to the Maggiano’s email data. To increase overall match rate, and to use in our direct mail targeting, we also appended physical address. Matches came in at around 50% of the file -a quantity that assured statistical significance for the modeling.

Step 2: Use Statistical Modeling to Identify Significant Groups of Customers with Similar Traits.

For 2 years, we’d been developing and using customer profiles for Maggiano’s direct mail targeting. These profiles were descriptions of “average” customers at both the national and individual restaurant location levels. And while these are very valuable for acquisition programs, analytics for CRM must be more granular. “Average customer” is not granular enough. For this task we now had to identify specific characteristics of the people who comprise the most significant customer groups or clusters.

To do that we turned to statistics. Beginning with simple summary descriptions, we moved through extensive cross-tabulations and significance trees to explore the data and understand what was there. We finished off with a cluster modeling technique, run separately for men and women. Here’s a chart that shows how the female population naturally grouped:

Cluster Distribution, Females:

% of Included Cases % of Total
Cluster 1 32.4 22.5
2 26.2 18.2
3 13.2 9.2
4 28.2 19.6
Combined 100.0 69.4
Excluded Cases 30.6
Total 100

To translate, 69.4% of all records were “included,” i.e. they clustered statistically into 4 significant groups. 30.6% of all the records were “Excluded Cases” because they fell outside the 4 significant groups. That means that 7 out of 10 of Maggiano’s customers fell into 1 of 4 statistically related groups. But how did we move from this information to actionable intelligence?

Step 3: Convert the Statistical Clusters into Flesh and Blood Descriptions

Here’s where the real fun began. Once we could classify each of the records into a cluster (or flag it as not-in-a-cluster), then we could explore each of the clusters. Here’s an example in which the cluster numbers are cross-tabulated against AtData Homeowner/Renter/Unknown data:

Own or Rent Total
Cluster Distribution Own Rent Unknown
1 % of cases w/in Cluster 1 68.9 28.3 2.7 100
2 % of cases w/in Cluster 2 96.1 3.0 0.8 100
3 % of cases w/in Cluster 3 48.2 6.0 45.8 100
4 % of cases w/in Cluster 4 99.6 0.3 0.1 100
Total % of total cases 82.0 10.9 7.2 100

As you can see, female clusters 2 and 4 are almost 100% homeowners. Are these women married or single? Are there children in the home? What is their family income? Are there any significant interests that they share? What makes them different from one another? As we answer these questions, a solid picture of real people always emerges. Statistics convert to characteristics and a comprehensive view emerges.


Following are samples of two of the many full profiles. In both cases, percentage of loyalty guests and percentage of US households is masked because of the proprietary nature of the information. The first description is of female statistical Cluster #1, the only cluster that was composed of nearly 100% single women:

Compare the description of Elizabeth to another, this one of Female Cluster 2 and Male Cluster 3 which were .statistically related:

From the perspective of profiling the “average” customer for direct marketing among prospects, these three people are exactly the same. They fall into an above average income range. They are. the same age range. Acquisition coupons are usually gender neutral. Treating them as part of the same profile produces solid results in such new customer acquisition programs.

But when you think about engaging these people and building relationships with them, profiling the “average” doesn’t cut it. But cluster segmentation modeling sure does. It opens a world of opportunity. How would you treat Elizabeth differently than you do Laura? With what would you incent each? What is it about the brand that appeals to each? How should you highlight that appeal?

Of course, once this stage is completed, it’s just the beginning. Among the steps that follow are these three simple ones (simple, but not easy):

The information in this case study about Maggiano’s Little Italy and its customers is shared by permission of Steve Provost, Michael Breed and the professionals at this great organization. Thanks to you, Steve and Michael, for the privilege of working with you and for the permission to share a glimpse into the organization that is a leader in your industry.

Talk with an Email Expert