More Accurate Customer Segmentation with Dentsu Aegis

See how Dentsu Aegis improved their data department’s agility and speed to more quickly recommend specific segments for targeting with a particular advertising campaign.

Content targeted at those more likely to buy is more effective than targeting everyone: that’s a given, and marketers and advertisers have known this for years.

Even back in 2014, Forrester reported that more than 75 percent of digital marketing professionals were already using advanced analytics and clustering techniques to personalize targeting. And 68 percent were already using machine learning algorithms for more personal targeting and interactions.

Hyper-targeting is already the norm in the advertising industry, so how can companies iterate and continue to improve? Generally, the next step is refining models for more accuracy but also making targeting faster, more real-time, and developing the ability to provide targeting recommendations on command. One company was able to do just this, further empowering sales teams to close deals with more advanced targeting.

The Challenge

Dentsu Aegis is a media buying company that allocates advertisers’ budgets on campaigns across various media (TV, digital, search, etc.) using targeted segmentation. When Dentsu Aegis pitches their services to potential customers, the sales staff recommends specific segments that would be the best to target with a particular campaign to maximize return. After they make the sale, the teams need to be able to deliver on those promises and actually maximize return with effective segmentation.

Watch Video

In order to provide this level of specificity, Dentsu Aegis created a data department with deep expertise in segmentation that allows them to provide more business value to their clients. However, the department struggled to quickly provide segmentation recommendations to the sales team – thus slowing the sales cycle because:

  • The teams built a data lake to collect data from multiple sources, but actually using the data meant embarking on the painful process of writing new code (Python, Spark, or SQL) every time.
  • Every time they had a project, team members had to write a query, get the results, analyze those results with another tool, and write more code to reprocess and use the data.
  • Without an easy way to replicate past work, each project required them to start their process from scratch, no matter how similar two prospects’ or customers’ use cases were.
Watch Video

The Solution

Dentsu Aegis chose Dataiku as the all-in-one tool to bring massive efficiency gains to the data department. Now with Dataiku, the department:

  • Quickly prototypes ideas for new ways to segment customers. If the segmentation works well, thanks to Dataiku’s collaborative environment, the whole team can easily reuse the models over and over again.
  • No longer has to write separate queries for similar projects. Instead, the team just prototypes a code-devoted pipeline running directly in Scala on Spark.

In turn, these efficiency gains free up the team to spend more time devising innovative new ways to segment customers and provide value to the company. For example, they used Dataiku to do predictive machine learning to find common features and weak signals. These weak signals enabled them to define very specific audiences that will respond (or not respond) to certain advertisements. For example, they found that users with Internet Explorer tended to be older or that via IP addresses, they can tell who is traveling during the weekend, and they can target accordingly.

 

The Results

Dentsu Aegis’s shift to Dataiku has allowed them to be four times faster in providing segmentation recommendations to the sales team. This dramatic increase in efficiency allows the sales team to close more business.

On top of efficiency gains, the team also has been able to do more prototyping and exploration, allowing them to deliver not only faster segmentation recommendations, but better and more granular ones. Dataiku has allowed everyone in the data department to focus on the strategic core of their job rather than on data preparation tasks.

Measuring ROI for AI Efforts

Calculating ROI on AI projects is often critical to secure resources; however, these calculations can also be notoriously challenging due to the complexity of quantification, number of teams and people involved, and often long delay in outcome.

Read more

Go Further

AutoML and the AI Skills Gap

Get an overview, real life case studies, and a list of essential features to look for in AutoML and augmented analytics tools.

get the white paper

Inclusive vs. Exclusive AI Engineering

Dataiku CEO Florian Douetteau writes in AI Musings about advantages and disadvantages of an inclusive AI approach.

learn more

Privacy-Compliant Data Projects

Move the organization forward safely in getting value out of data despite increased regulation surrounding data privacy.

get the white paper

GE Aviation's Story

From data silos to self-service analytics, see how GE Aviation has democratized data across the organization.

learn more