Optimize your predictive analytics through model ecosystems

Optimize your predictive analytics through model ecosystems
Aug 1, 2019

Friends, I did it. I bought a pair of shorts via an Instagram ad.

It wasn’t hard for them to find me. Summer is on the way. I needed shorts. But now, they haunt me. Wherever I go, internet or app, up pops the same ad that successfully enticed me to buy shorts. Cajoling me. Interrupting me. Perplexing me. I already bought your shorts, I want to scream into the digital void.

In 2019, I’m fairly sure we all face this problem in one form or another: a creepy birthday message from some brand you’ve never heard of; a site you visited once flooding you with email; and, well, I’ll scream it from the rooftops: I already bought your shorts.

Thing is, brands pay a lot of lip service to “lifecycle marketing” and “customer journeys” and the “marketing ecosystem.” You are not just numbers, these buzzwords say, but rather individual human beings engaging in a broad network of moving parts. And yet most companies don’t have the infrastructure in place to deliver on all this talk in the right way; that is, discovering what a given customer wants to hear at specific times, and communicating that message in creative, human ways.

What brands need: a model ecosystem that humanizes data

That’s because while most brands use predictive models to induce specific customer actions—you there, searching Google for sunglasses…reply to this offer…for sunglasses—these models, however many there may be, work alone. This lack of cohesion means companies, no matter their claim to personalization, are still deciding what targeting effort is best for you (and what communication should be sent) rather than letting end-to-end data tell them what and how to serve a customer at each specific moment in her lifecycle.

So: The shorts people, rather than sending me a thank you email or suggesting a great new t-shirt or pair of sandals, might decide I should get the same ad because, say, it’s still summer, and I fit their target demographic, and I like these other brands.

In other words, they won’t have the true personalization that comes from a model ecosystem: an interconnected system or set that links all the disparate predictive and analytical models, empowering brands to utilize the right models at the right time for the right customer.

These ecosystems are crucial for brands looking to follow through on the promise of “lifecycle marketing.” Done well, they’ll produce interactions and experiences that are more personal, human, and relevant, driving engagement, retention, and emotional loyalty: a relationship with the customer built on mutual trust, investment, and empathy.

5 steps towards a model ecosystem

Building a model ecosystem within your company is a complex technological, organizational, and cultural undertaking. That said, here are five essential steps firm leaders should keep top-of-mind when getting started.

1). Define each customer’s journey and key life stages within your company.

You won’t know who to reach with what communication at what time if you don’t know who those customers are and how they’re engaging with your company.

But of course, each customer is unique, which means it’s crucial to leverage data to build distinct customer segments—outlining who they are (media habits, brand affinities, demographics) and how they behave (frequency, recency, channel preferences, product affinity, average order value, share of wallet). This data shouldn’t just come from transactions; brands should use additional data overlays—be they psychographic and/or online behavior-related—to understand not only who an individual is, but what drives and motivates them.

2). Determine what behavior you want to drive at each step in a given customer segment’s lifecycle.

Once you’ve got your customer segments in place, this should be intuitive. You want to acquire a new customer, you want a new member to repeat purchase; then maybe you want to cross-sell to a repeat purchaser and turn your loyalist into a brand advocate.

Remember, too, that each model can have a different goal—and that the lifecycle for each customer segment or even subsegment is not the same. A model ecosystem allows for a flexible, tailored approach that’s both plastic and channel agnostic.

3). Overlay predictive models onto journey maps to target the right customer with the right communications at the right time.

A customer who has just purchased a pair of glasses, for instance, won’t need another pair right away. But four months after acquisition there might be an opportunity to tell a brand story and cross-sell; nine months in might be an opportunity to offer deals on seasonally relevant accessories; eleven months in, you might remind them to get an eye exam.

A modeling ecosystem that continually rescores its models after each interaction—be it a transaction, search, or lack of action—or behavioral change, can enable those targeted communications to happen automatically, in real-time, right when a person logs on. Overlaying these onto comprehensively designed journey maps ensures that these interactions aren’t creepy, and that they’ve got a human touch.

4). Let the creative and experiences bring campaigns to life.

None of this works without dynamic, personalized communications or real-life interactions. To make this a reality, the analytics team must partner with other parts of the organization—customer service, strategy, etc.—at the front end of any process to ensure campaigns have that all-important human element.

5). Don’t stop iterating.

You can only do so much in year one. So build up as you go: the more you learn, the more you can deepen your customer segmentations and predictive models, as well as the channels you use to amplify key life moments.

Get out ahead of coming trends

A model ecosystem won’t just be helpful in the short term. Investing in this infrastructure now—i.e. technologies that allow you to identify individuals across different touchpoints, and that interact with one another—will prepare you and your customers for several emerging trends.

For one, this ecosystem builds trust with customers that will prime them for the data-sharing we expect to explode in coming years. Amazon and Whole Foods are already taking steps in this direction, with the two brands sharing data to better serve their customers.

What’s more, model ecosystems train companies to truly listen to customers throughout their journeys with the brand. This is important because in the future, more and more customers will decide what benefits they want rather than being told what they’re going to get.

Finally, this infrastructure nurtures the emotional loyalty that coming technologies will gauge and act on. In the near future, you won’t just be reaching out to customers based on their transaction interactions, but also at key emotional moments (e.g., think the in-between moments of travel, such as delays, rooms not being ready, upgrades, etc.) or measuring sentiment across touchpoints or brand interactions.

The world is moving towards a humanization of data, and getting started on a model ecosystem is a vital step in catching the wave.

On a personal note? I just don’t want to see another ad for shorts.

ICF’s global marketing services agency focuses on helping your organization find opportunity in disruption.
Go to ICF Next
Meet the author
  1. Emily Merkle, Partner & Line of Business Lead, Analytics and Data Science
File Under

The latest marketing trends, uncovered.

Subscribe to get insights, commentary, and news sent straight to your inbox.