Adam Gartenberg's Blog

Business Analytics and Optimization, IBM and Social Marketing

IOD 2013 Day 2 General Session

The focus for the IOD Day 2 general session was on highlighting practical ways people are drawing new insights from data to make better decisions.  Below are my notes from today's general session.  (And in case you missed it, today's session ended with the announcement that going forward this conference will be known as IBM Insight, held next year Oct 26-31.)

Mike Rhodin, SVP IBM Software Solutions Group

Competing today requires infusing intelligence into the fabric of your enterprise.
It's not just social data on the rise - we're seeing the convergence of social, mobile and cloud, adding in instrumented data fro 19B sensors around the world, along with the growth in enterprise data.  

The question is - what's your strategy.  Will you be one of the many washed overboard by the next big wave, or do you have a strategy to navigate this sea of data?

We must think of analytics not as a thing that we do, but as a component of everything that we do.  It must be infused into every process in the organization, guiding every decision.

Today we're looking at how to democratize analytics, putting it into the hands of people that need it the most, at the time that they need them, in the way they want to use them, regardless of technical or analytics experience.

We are rethinking how we deliver capabilities.  Cloud has become a preferred delivery vehicle.  True value lies in the ability to accelerate transformation, and particularly true with analytics-driven organizations.  

Today we're delivering more than 100+ solutions delivered via IBM SmartCloud

Solutions that are transforming the analytics experience:

IBM Concert - new user experience that blends analytics with social, mobile and cloud

  • New, visual user experience
  • Personalized, prioritized task lists
  • Collaboration and mobile-enabled
  • Compatibility with Cognos TM1 on cloud
  • Available on cloud December 2013.

Project NEO - Next generation of data discovery
  • Doesn't require user to understand language of analytics - ask it questions in natural language and Neo does the rest
  • Semantically answers user's question to select appropriate data set and uses analytics to identify the most relevant
  • Beta in January 2014

Beth Smith - VP, Business Analytics Products and Solutions
Demonstration of Predictive Maintenance and Quality (PMQ) Solution.
Transform what once were operational challenges into opportunities.  For example, a textile company reduced their defect rate by over 50%, and an industrial manufacturer is detecting fraud before it happen.

Step 1 - Capture (at rest, in motion)
Step 2 - Predict by analyzing and doing real-time scoring to predict future outcomes
Stpe 3 - Act on that analysis

Example - A car's telemetry data show that engine is running hot and the timing belt is not working properly.  The dealership knows that based on historical data and patterns, it means it's likely that the water pump is about to go.  The dealership can alert the owner and order the necessary parts before he even gets there.  This is far better for the car's owner than having it break down on the side of the road, and the manufacturer can better forecast inventory needs, minimizing warranty claims.

Real world examples:
  • A city gov't decreased the number of poor quality roads by over 50%.  They are engaged in proactive predictive maintenance, leading to increased productivity of field workers.
  • Mining industry - A mining company in Australia is looking at many indicators to see which equipment needs preventative maintenance, allowing for higher utilization
  • Israel Electric - Provides 90% of country's energy.  They are now alerted 30 hours before a failure, as opposed to what was just 30 min.  They can prevent power outages before they occur.

David Becker, Director, Election Initiatives, Pew Charitable Trusts
Jeff Jonas - IBM Fellow, Chief Scientist, IBM Identity Analytics

David Becker
Elections in the US have a 48 month business cycle, and yet from a registration standpoint, most of the activity is compressed in a very short timeframe right before the election.  This last month is like black friday and the day after Christmas rolled into one and multiplied by 1000.

Part of the reason this is such a problem are the voter lists.  In the US, you need to get on the list to go vote on election day.  #1 problem of keeping voter lists up to date are moves.  1 in 8 Americans move in a given year, and that figure is 1 out of 5 for younger Americans.  Adding to this is that many of the registrations are coming in paper form, adding to the time needed to process them.
  • 12.7M out of date records at any given point in time, because people have moved since last election
  • 1.8M deceased individuals
  • 2.7 Million registered in multiple states (because moved out of state)
  • 1 in 3 voters think their records update automatically with each move (it doesn't), and over 50% of voters are unaware they can update their info at the DMV
  • 1.2M votes were lost due to registration problems in 2012
  • 51M citizens (1 in 4) are eligible but not registered to vote.

They put together a team of experts to look into this.  They thought that it might not just be people issue, but also a technology issue, and reached out to Jeff Jonas to see if he might be able to help.

Jeff Jonas
Realized they were talking about the same type of problem that hear about from organizations all the time. It's a problem of understanding context.  e.g., You have two records in two states - same name and DOB, but everything else is different.  How can you tell if they are the same person or not?  It turns out that you can connect that person and address with DMV records in each state, and those DMV records have additional information (e.g., SSN) that can be used to confirm that these are the same people after all.

Utilizes the G2 system, including selective anonymization, so while the system knows the SSN number, it's not human-readable..  
(Note:  At this point Jeff described what I think might be the best description I've heard of a "one-way hash" encryption in plain english:  If I give you a pig and a grinder, you can make sausage... but if I give you a sausage and grinder, can you make a pig?)

There are now 7 states working with PEW for Electronic Registration Information Center (ERIC).  Helps states find voters who moved.  Helps reach out to people outside of the peak month right before the election.

Results - since started being used in 7 states in 2012:  
ERIC was able to identify 850,000 registered voters in the 7 states participating, and over 23,000 dead people who were still registered to vote.

Also actively encouraging people not on roles to reach out to get them to vote.  Identified more than 5.7 M eligible but unregistered voters.  300,000 of these registered (pretty high rate since could only reach out via direct mail).  Most of them registered using online registration, moving away from paper-based - cheaper, more accurate.

Paper registration transactions costs $0.83 to process, but online registration costs only $0.03.  Allows public officials to redirect that money to better uses.

G2 is available inside of SPSS Modeler Premium.

Manoj Saxena - GM, IBM Watson Solutions
Watson represents what we consider to be the 3rd generation of computing.
  • First generation - counting, tabulating systems
  • Second generation - calculating - programmable systems
  • Third generation - reason and learn - Cognitive Systems Era

At the heart, Watson does three things
  1. Understand natural language at scale.  (e.g., Watson could consume and understand 200M pages of data a second).  Now Watson can understand a question up to 20 pages long, not just 3 sentences like in Jeopardy
  2. Generates and evaluates evidence-based hypothesis - not just response, but scoring the possible responses for you.
  3. Adapts and learns from interactions of user selections and responses.  (Cognitive systems gain value over time.)

Watson in Healthcare:
  • Training Watson on cancer:  25,000 case scenarios, 2 million pages, 15,000 hours of training
  • It used to take researchers 3 years to take genomic data and create a new cancer therapy, and after Watson, they think they can do that in 10 months.

Expanding Watson beyond healthcare to areas such as call centers
  • Customers are concerned that we're sharing more information, but call centers we call know less about me.
  • Watson Engagement Advisor - services offering within Smarter Commerce.  Tremendous response from customers

Next in 2014 for Watson:  
  • Cognitive cloud and API so you can start writing apps against it.
  • Next big challenge for research:  Today, Watson is operating at the level of a first year medical resident.  Next want it to take US licensing board exam.  They've started working on it, and are in striking distance.  

Fred Balboni , WW Business Analytics and Optimization Leader, IBM Global Business Services
Is there a blueprint that we can use to create value from analytics.
30,000 engagements over past few years.

Found 3 common components that have led to success in analytics programs.
  • Strategy:   Sponsorship, Source of Value, Funding
  • Technology: Expertise, Data, Platform
  • Organization:  Culture, measurement, Trust

Karen Etzkorn, CIO, HSN inc.
  • $3.3B direct to consumer retail portfolio that includes HSN and cornerstone portfolio brands
  • Reaches 95M homes
  • Most important trend in online retail is with data and analytics.  retailers who understand what buyers want, how they want to buy and engage will be successful.  Want to create boundaryless retail (not omni-channel).
  • In midst of shift in what retail even means.  Driven by consumer and redefined by technology, social networks, and mobility.
  • Must invest in capabilities to efficiently and effectively filter the data without being driven by noise.  
  • Transform culture from being data rich and information poor, to one that leverages real time insights to drive lifetime loyalty.
  • Create enterprise wide view of the customer
  • Develop advanced analytic capabilities to understand/anticipate needs
  • Optimize each interaction

New name for the conference going forward will be: IBM Insight.  Oct 26-31, 2014