g2o animated logo transition
Learn how we’re helping our clients through the COVID-19 disruption. Click here to learn more.
Notification bar close button
Unfortunately, Microsoft is no longer supporting new web standards for Internet Explorer and recommends that users upgrade their browser experience.

On the bright side, there are some awesome modern browsers like Microsoft Edge, Google Chrome, Firefox, and Safari that can enhance all your web experiences. What’s your next...browser?

Predicting customer value and retention with data

With predictive models, we helped a credit association better understand who their high-value customers were and which customers were most likely to leave.

Research & Analytics
Data Operations
Back button
Photo of a mature man on his smart phone


A credit association wanted analytical models to better understand their customers so that they could grow revenue. They wanted to know who to contact and when with their marketing efforts. But data that was siloed and incomplete was getting in the way.

They came to g2o for help getting their data into shape because of expertise with customer data analytics. They had some specific goals in mind.

First, they needed to convert low-value customers into high-value customers, and secondly prevent those high-value customers from leaving. These are simple goals, but ones that require a deep understanding of the company’s customers:

  • Who are my high-value customers, and why?
  • What is my retention rate, and who is likely to leave us?

With these insights in place, it would be possible to make predictions around lifetime value and future revenues, but the credit association lacked a firm grasp on this information.

A handful of environmental factors further complicated this initiative. Another strategic partner set the course for these revenue goals but failed to execute on the strategy. As a result, g2o was under pressure to deliver and demonstrate our ability to partner with the client’s on-site team. This project would prove to be a test of flexibility, partnership, and performance.

Photo of a mature man on his smart phone


Our team worked with the credit association to build analytical models that predicted conversion factors of low-value customers to high-value customers, and the likelihood of high-value customers to leave.

We accomplished this in two principal stages:

  • Using the right data
  • Applying the right science

While predictive models typically get the glory, much of the heavy lifting rested in engineering the appropriate data. The credit association maintained records in siloed product systems: checking accounts, life insurance, healthcare, etc., hampering their intelligence. For example, the checking team knew that Chris Miller was a customer, and the HSA team knew that Chris and Claire Miller were both customers, but no one knew that Chris had both HSA and checking accounts, or that he lived with Claire. To be successful, our client needed to market themselves effectively, but they didn’t have the visibility to their audience to do so.

g2o was able to collect these datasets, enrich them with missing information, and aggregate them into a quality dataset suitable for analytical models. Once we had a single view of the customer, we aligned that data to their appropriate address, so that the client could market to a specific household efficiently.

This process takes a rich understanding of data governance and careful curation of third-party insights. To inform our predictive models, we added location intelligence, life stage, wealth, and numerous other demographics and attributes to our client’s existing data. That meant that customers wouldn’t just become lost business anymore when they changed their name or address.

Once we armed the client with a clear view of their customers and how to reach them, we uncovered who needed their attention. To achieve this, we developed predictive models to test relevant variables until we formed the most accurate algorithm.

Photo of a mature man on his smart phone


The project provided the client with an iron-clad understanding of their customers, precisely who was in danger of leaving, and which customers were ripe for upselling, by name.

With the hand-off of our predictive models, this financial institution was empowered to serve their customers better, and their bottom line will see the difference. These methods yield smarter marketing, more effective targeting, and more efficient use of resources.

Changing and inconsistent customer information, poor data entry and other data quality issues can lead to a lack of context and understanding. A single view of the customer removes ambiguity and supports better decision making.

Related work