Predictive Analytics Innovation Summit, Chicago – Day 2 Highlights

Highlights from the presentations by Predictive Analytics leaders from Time Warner Cable, AT&T and Verizon on day 2 of Predictive Analytics Innovation Summit 2014 in Chicago.

By Anmol Rajpurohit, @hey_anmol, Jan 2015.

ieanalyticsPredictive Analytics Innovation Summit was held by Innovation Enterprise in Chicago during Nov 12-13, 2014. It provided a platform for leading executives to share interesting insights into the innovations that are driving success in the world's most successful organizations. Data scientists as well as decision makers from a number of companies came together to learn practical predictive analytics from top companies like Amazon, Twitter, Verizon, Microsoft, etc. Industry leading experts shared case studies and examples to illustrate how they are using analytics to innovate in their organization.

Highlights from Day 1.

Here are highlights from Day 2 (Thursday, Nov 13):

Tanya ZyabkinaTanya Zyabkina, Director, Marketing Analytics, Time Warner Cable discussed how analytics can be used to optimize subscription price change. Price changes due to promotion expiration can be described as “natural experiment” i.e. TWC cannot perform a true test due to legal and operational restrictions. Also, the customer groups that get price changes are pre-defined by the promotion code on the account and generally cannot be changed by the company.

She described what approaches TWC used while estimating incremental disconnects, upgrades, and downgrades attributed to the price change due to promotion expirations. The disconnects, when analyzed, show a substantial increase around the timing of a common price change point. However, the increase is largely explained by confounding factor, which is dominated by customer moves.

Segmentation helps TWC understand which segments are more sensitive to price change and based on this information, structure future promotional offers to minimize customer attrition. She concluded by sharing the following recommendations from her experience of analyzing natural experiments:
  • Focus on incremental impact
  • Look for cofounding variables
  • Create well matched synthetic controls
  • Use impact segmentation to optimize change

Nadeem Syed
Nadeem Syed
, Director, Marketing Analytics, AT&T talked about "Predictive Analytics in B2B Marketing". Digitization is changing marketing, especially B2B Marketing. By 2020, about 450 billion transactions would take place on the web with companies having more than 1000 employees storing 200 TB of data on average. For business buyers, 90% of decision making takes places before first contact. Outbound marketing has to be more focused as the buyers are getting more informed and each transaction now involves more stakeholders. Marketing needs to evolve in the following three ways:
  • Shift in skills and mindset
  • Data to drive prioritization
  • Predictive Analytics for results

Segment using clustering (k-means, Gaussian mixture models, hierarchical clustering) in order to utilize rich data set from multiple sources to discover new and niche segments. The optimization of Conversion Funnel through prediction helps data driven prioritization, timely alerts and maximization of value. Methods such as logistic regression, SVM, Boosted Decision Trees can be used here. Sentiment index can be used for proactive support. This can be done by tracking sentiment in near real-time, followed by anticipating & responding to needs. This process must also involve close-loop analysis with surveys. The secret sauce for B2B marketing is:
  1. Be intentional
  2. Talent Makes the Difference
  3. Culture Matters

Among the above three, intentionality is most critical. It involves aligning analytics priorities with business strategy, identifying the most relevant business questions, starting small, testing and adjusting. Talent with strong business acumen and strategic perspective is key. In regard with culture, “data-first” mindset within marketing helps a lot.

Ksenija DraskovicKsenija Draskovic, Head of Predictive Analytics & Data Scientist Group, Verizon delivered a talk on "Improving Customer Insights with Predictive Analytics: The Current State and Big Data Trends". She started her talk describing the huge volume of data Verizon holds. Talking about the current state of predictive analytics at Verizon, she emphasized that Verizon has a comprehensive view of customer as the analytical data set foundation. Verizon has an automated process consisting of: data collection, data preparation & enrichment, scoring and model performance monitoring. By integrating predictive analytics with business processes, her team handcrafted models to support strategic, tactical and event driven business decisions. They are constantly monitoring overall model and business results keeping ROI in mind.

She argued that the key to success is picking up the problem that everybody in the organization is interested in solving and that is high priority for your organization. With regards to Big Data trends, she mentioned that integrating unstructured data analytics with current data and processes is critical. Text data is heavily used for business insights and customer sentiment. Current processes are creating an immense need for unified contextual view and centralized analytics. She mentioned and briefly explained the following challenges in predictive analytics:
  1. Not all business units aware of the benefits or possible use
  2. Lack of Data Scientists: Modeling expertise and in-depth business as well as data understanding
  3. Evaluation of Model and Delivering insights