Demystifying predictive analytics: evaluate your infrastructure

Saken Kulkarni

Saken Kulkarni

Technology’s changing by the minute, and so is your customer’s behavior. How do you find out where your customers are and how they’re making purchase decisions, and then swiftly react when that all changes? Customer-centric, predictive analytics are an important piece of the solution, and a key theme at NRF’s 2014 Big Show (and what this blog series is all about). A couple weeks ago, I talked about how to start building a customer-centric organization by analyzing the status quo. The next step is evaluating your infrastructure to lay a solid foundation for your analytics program.

I live in New York City, where owning a car is an unnecessary luxury. When I lived in Washington, DC I had an old Toyota Prius that I used for weekend trips. My Prius was a great car, dependable and energy-efficient, but it had one minor issue. Every time I revved the engine past 55 miles per hour, I noticed that the vehicle would shake. It simply was not designed for high performance. Similarly, all customer analytics databases are not created equally. Your organization may not need an Aston Martin when an Accord will do.

You may think of a database as your customer data “library.” Promotions, returns, demographics, and transactions are filed and organized, retrievable through searches (often through Structured Query Language, commonly referred to as SQL). Like a library, some “books” are checked out on a regular basis, while others gather dust. In my experience, I’ve found that an organization’s choice of database management system is heavily dependent on its data maturity level.

Customer analytics database maturity model

Analytics database maturity model_1

I’ve noticed that the distribution of retailers within these groups can be described as a bell curve, with “developing” retailers as the majority and “basic” and “advanced” retailers as the two tails.

Database maturity model_3

Level 1: Basic

Don’t despair if your customer data needs fall into this category. It means that you have a clean slate to work with in the journey toward advanced analytics. Level 1 organizations may benefit from designing, developing, and implementing an independent customer analytics data mart for their analytics needs. An independent data mart connects directly from operational systems (like a point-of-sale system) and usually focuses on a specific department (like marketing). Think of your independent data mart as the non-fiction or biography section of your library.

Basic structure of an independent data mart

Independent data mart

For example, a luxury goods retailer may design a data mart that contains customer information, promotions, and augmented demographic data for its direct marketing department. These department-specific data marts can be designed and built in weeks and can cost significantly less than a full-scale Enterprise Data Warehouse (EDW). However, retailers must be careful to design the data mart to scale as needed. Your data needs will grow as your organization grows.

Level 2: Developing

Most organizations find themselves in this group. Their data needs are more advanced than what can be provided through independent, departmental data marts. They also require data and the interaction of data from different departments (marketing, supply chain, HR, and finance). For instance, a regional grocer creating a customer loyalty program may need customer information from marketing as well as transactional information from finance. This grocer may benefit from the implementation of a full-scale EDW.

An EDW is a central location for integrated data from various operational sources. Think of an EDW as your library that contains a variety of sections (non-fiction, biographies, children’s books, etc.). In its most elementary form, data from operational systems are loaded into a staging area, a quasi-landing zone for data transformation, and then into the EDW itself through a process called Extract Transform and Load (ETL). ETL takes the data from the source system, transforms it, and loads it into the EDW (see the diagram below).

Basic structure of an enterprise data warehouse

Enterprise data warehouse

The main advantage of an EDW is that it provides a central location for IT management. This ease of management can reduce redundancies and create efficient methods for managing permissions, ensuring data integrity, and maximizing database performance.

The implementation of an EDW isn’t without its challenges. An EDW consists of data from several departments. These departments must work together—not in silos—for customer analytics to be conducted efficiently. Use your Analytics Center of Excellence (ACoE) that you created when you analyzed your status quo to encourage collaboration across the enterprise, or appoint a steering committee that meets on a regular basis.

Level 3: Advanced

Level 3 is the gold standard for advanced customer analytics. Imagine a world where a retailer can analyze cross-sell affinity by analyzing terabytes of real-time transactional data, while simultaneously analyzing text data from blogs, customer reviews, and wish lists, and then matching those customers with social media profiles. This rich analysis is what we expect when we talk about the power of Big Data customer analytics.

As you may imagine, this type of analysis requires the Aston Martin of data storage and processing. Big Data products like Apache Hadoop, an open source framework, can be extremely useful in the processing of large amounts of data quickly and accurately because they allow for storing and processing of large data sets across distributed clusters of servers for more efficient processing.

The implementation of Big Data infrastructure does not mean that you have to discard your EDW or independent data marts. Rather, these products can work together to provide rich sets of customer analysis to help you target and segment your customers. Keep in mind that products like Hadoop require technical talent with a strong background in open source languages. Keep your customer analytics team up to date with the latest versions and patches, and don’t be afraid to use Massive Online Open Courses (MOOCs) to train your employees.

Select the proper horsepower for your advanced analytics needs

Evaluating your database infrastructure is an important second step in the creation of a data-driven culture. Be honest with yourself and your organization in order to design, develop, and implement a data storage system that is scalable, efficient, and cost effective in both the short and long term.

Stay tuned for my next post on how to leverage these databases in the design of your customer analytic methods.

About sakenkul
Saken is a Business Analytics consultant at Slalom Consulting. He focuses on the intersection of customer engagement, analytics, and data visualization

3 Responses to Demystifying predictive analytics: evaluate your infrastructure

  1. silvonsoftware says:

    “These departments must work together—not in silos—for customer analytics to be conducted efficiently. ”

    Silos are going to be the end of us, especially when it comes to better data management. Different departments still need to use the same data so no matter who accesses the data it needs to be up-to-date. If the data is in silos you can’t keep it updated.

  2. Pingback: Demystifying predictive analytics: design your analysis | The Slalom Blog

  3. Pingback: Demystifying predictive analytics: train and drive adoption | The Slalom Blog

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: