It’s hard to explain what’s quite so appealing about an advertisement for toothpaste the day after you’ve squeezed the life out of your current tube. But if we have learned anything from the Instagram’s current and continued success, it’s that data targeting works. And it works well.
With that in mind, in part one of a two-part series, we asked five Chicago companies how they’re using data in ways we might not expect, and what results they’ve seen. They told us about initiatives ranging from detailed wardrobe customization to criminal recidivism.
Dating apps are so successful, in large part, because we as people enjoy voicing our opinions despite minimal information. In short, it’s human nature to take pleasure in casual judgment.
Personalized mid- to high-end men’s and women’s clothing service Trunk Club has recently applied this idea to fashion, an industry where it feels, perhaps, much more practical. Engineering Manager Matt Frey gave us the details regarding their new rating feature and how they use data to convert customers, below.
What’s one way your team uses data that people may not expect?
We’ve recently started leveraging anomaly detection for critical customer actions on our site. This allows us to detect deviations from historical customer patterns, such as fewer than expected number of people signing up for Trunk Club on any given day of the week. We’ve been able to use this as an early warning system for any issues that may impact our customers, even if there are not increased errors.
In the past month alone, we’ve had more than 11 million user product ratings.’’
What’s a recent data-driven project your team worked on that created success for either the company or its customers?
We’ve launched a feature called “Style Swipes” that allows customers to quickly view and provide thumbs up/thumbs down feedback on hundreds of clothing items. This feature has been very successful not only in terms of customer engagement, but also in providing our stylists and data scientists with tons of data on customer likes and dislikes.
Using this data, we are able to more accurately provide clothing suggestions to customers and pack trunks with items they are more likely to keep. In the past month alone, we’ve had more than 11 million user product ratings.
While having a closet like that of Cher’s in “Clueless” (digitally run, seamlessly integrated) might be #wardrobegoals for both fashionistas and the “ensembly challenged” among us alike, data-driven solutions extend far beyond the big screen. Enter KAR Global, a business dedicated to helping manufacturers, financial institutions and insurance companies buy, sell and trade used vehicles via dedicated algorithms. Vice President of Inventory Solutions David Rymarz told us about how their data models work, below.
What’s one way your team uses data that people may not expect?
In the used-vehicle industry, data is everywhere you look; from physical and digital auctions to dealership retail data and much more. Many people would be surprised to learn that DRIVIN, KAR’s data science engine, seamlessly gathers billions of data points across all of these environments and stitches them together to generate an exclusive set of data science models.
But we don’t stop there. We use these data points to provide personalized, real-time vehicle recommendations for our customers. Combining hundreds of vehicle inventory data points with unique dealer retail data, our algorithms are so tailored that no two dealers will receive the same recommendations.
We use these data points to provide personalized, real-time vehicle recommendations...’’
What’s a recent data-driven project your team worked on that created success for either the company or its customers?
One example of how we’ve been able to contextualize our unique data set is with our sales optimization tool. The tool is designed for our buyer solutions team representatives, who help thousands of used car dealers find the best inventory for their lots. The tool generates tailored, real-time vehicle rankings available for purchase through an online auction, including vehicle details about estimated retail price, mileage and more. The optimization tool has already helped our representatives have more meaningful conversations with dealers about inventory to source, while providing the evidence dealers need to feel confident in their buying decisions. It’s a win-win.
Have you ever thought of your gym membership as more of a recurring charitable donation than a practical investment? Employee benefits can quickly become disadvantageous to those they are supposed to support in the same way. But unlike fitness centers, companies don’t benefit from ghost accounts. Evive uses big data to try to change that pattern among employers, encouraging transparency and increasing access to health and wellness plans.
Christina Sung, manager of outcomes and engagement analytics, explains how data plays a part.
What’s one way your team uses data that people may not expect?
The outcomes team analyzes data from multiple sources to understand which benefits are most important to our clients’ employees. For example, we assess which benefit topics best resonate with their employees based on engagement within Evive’s platform and the personalized communication (e.g. email or text message) that is delivered to them.
We direct Evive’s users to the most relevant vendor programs based on certain traits...’’
What’s a recent data-driven project your team worked on that created success for either the company or its customers?
One of Evive’s goals is to help maximize the use of benefits by our clients’ employees. To help achieve this goal, we direct Evive’s users to the most relevant vendor programs based on certain traits, such as demographics, biometrics and their past claims history. We recently measured results with a program partner focused on managing hypertension and saw that the registered program users’ medical claims costs were much lower compared to those who did not register.
As social media users, we rarely consider what goes into social media content moderation. The process requires much more than deleting posts, and the ethics of the job runs deep. We recently spoke with Jacob Zweig, co-founder and principal data scientist at Strong Analytics, about a project that gave him a greater appreciation for such challenges.
What’s one way your team uses data that people may not expect?
When people think of machine learning and artificial intelligence, ideas of completely automated systems and scary sci-fi robots often come to mind. In our work at Strong, we instead focus on building systems that leverage machine learning to augment human decision making. For example, in a recent project, we worked with our client to re-imagine last-mile logistics. Rather than thinking of autonomous delivery vehicles, we partnered to build a solution that uses advanced computer vision and deep learning to augment drivers’ preferred workflows, ultimately increasing delivery efficiency.
...we focus on building systems that leverage machine learning to augment human decision making.’’
What’s a recent data-driven project your team worked on that created success for either the company or its customers?
As a data science company, we develop data-driven solutions to problems in many different industries. In a recent project, we were tasked with the difficult challenge of content moderation on social media. What made this especially challenging was that we were tasked with specifically identifying abusive content that was not computer-generated (CGI). For example, it’s one thing to recognize an image of a weapon. But with the quality of today’s video games and 3D rendering, it’s quite another to confidently identify real weapons versus CGI weapons. To tackle this challenge, we had to go beyond the standard image classification approaches being used today and think about how to identify the visual properties of CGI independently of the image content.
Analytics8 provides tangible insights to power logistical results for companies like Anheuser-Busch and GE Aviation by using historical data collection and predictive models. That CV doesn’t only sound impressive. Central Region Vice President Brian Yaremych told us how their tools are making a social justice impact with economic consequences.
What’s one way your team uses data that people may not expect?
We helped one of our customers create a platform that analyzes a variety of employee and personnel data that helps predict when an employee is at risk for leaving their job. The solution, through the use of automated intelligence and machine learning, analyzes several datasets, such as employee performance reviews, compensation and tenure, to highlight factors that drive employee retention and morale. Management has been using these insights to tailor their personnel practices for a more desirable work environment.
The solution also uncovered some unexpected insights, such as inconsistencies in how performance reviews were being conducted.
With this data solution, the county is equipped to help reduce crime and improve reentry planning...’’
What’s a recent data-driven project your team worked on that created success for either the company or its customers?
For one of our recent projects, we implemented a data and analytics solution for Crow Wing County, Minn. Their major objective was to reduce criminal recidivism, or the tendency for a convicted criminal to re-offend. Discharge plans were created for offenders, but the one-size-fits-all approach wasn’t optimal.
We first took inventory of all their data sources and cleansed and standardized the data for easy consumption by business intelligence tools. We then developed an analytics platform that gives the county an easy way to review all offender data in a single location to create informed discharge plans on an individual basis.
With this data solution, the county is equipped to help reduce crime and improve reentry planning, resulting in a safer community, better lives for ex-cons, and better allocation of county funds.
