Predictive analytics are increasingly becoming a necessary capability for organizations that wish to better achieve business goals through the use of data. They provide a systematic mechanism through which a myriad of data can be transformed into value. Depending on the organization, however, getting the right data warehousing / business intelligence system in place can be a daunting task. Budgets, IT infrastructure, collaboration of cross-functional stakeholders, and senior-level sponsorship are critical factors that must be aligned to enable success.
As Senior Director of Measurement Science at National Consumer Panel (NCP), a joint venture of Nielsen and IRi, I have the advantage of working for an organization whose primary purpose is to supply quality market research data. The heart of our business are the data provided by our panelists and, combined, the data are a critical element of the consumer insights Nielsen and IRi provide to many of their respective clients. To maintain the largest national representative, the longitudinal consumer panel, available in the marketplace, NCP has a deeply rooted history in data and measurement science. Thus, using data about our data is critical for ensuring that our organization provides the best quality data possible.
NCP has established Key Performance Indicators to measure how well the panel is performing, which provides a natural place to look when considering where to start. How well our panelists participate and how long they stay on to participate are two critical KPIs for NCP. Panelist participation includes sending in their purchase data through their scanners and answering product preference and other surveys. The longer a panelist is an actively participating member, the better for our business and our clients’ business. Thus, recruiting the best potential panelists to replace those who leave the panel for one reason or another provides an opportunity to leverage predictive analytics.
A predictive model using a select set of data inputs has been implemented at NCP to better assess the value of potential panelists before they are invited to join our panel. The use of a standard test vs. a control group evaluation of the model later confirmed that the attrition rates of panelists who had been selected had lower predictive model scores than the panelists who had not been scored.
Demonstrated success with this predictive model has helped to lay the groundwork for building a more comprehensive database, which holds the promise of generating additional efficiencies in business processes. First, building the database has required automating the aggregation of relevant data to feed into the database. This alone has created efficiencies to enable faster turnaround on a variety of analyses that previously required substantial effort to simply pull the necessary data together.
In addition, having an integrated database enables a richer variety of data (e.g., behavioral, attitudinal, and demographic variables) to mine and more fully address a range of business questions. One of the next steps will be to evaluate variations on the established predictive model considering a wider array of KPIs. Additional implications include improving panelist retention and participation, as well as cost savings on panelist recruitment and scanner replacements.
We’ll be discussing the importance of early “wins” in predictive anlaytics at the upcoming IBF’s Predictive Business Analytics Forecasting & Planning Conference, April 22-24 in Atlanta. I look forward to meeting you in person.
Senior Director, Measurement Science
National Consumer Panel