Intelligent machines are becoming a ubiquitous and essential part of business operations. Algorithms trained by faster, smarter data now play a key role in determining customer demand, increasing customer trust and delivering unprecedented productivity and intelligence to the enterprise.
But despite the success of some early adopters, most machine learning projects fail. Regardless of their size, location or industry, businesses have been struggling to fully unlock and realize the value of their information.
In fact, a recent Information Value Index from PwC and Iron Mountain found that 57 percent of businesses are unable to extract significant value from their data; 23 percent are unable to derive any real value at all.
In my view, most large enterprises fail to benefit from potentially rich data sources because they just aren’t equipped to effectively capture and analyze data. Without this ability, they can’t realize the fundamental driver of a data-driven transformation: an impactful data story.
The value of a good story
All good data stories start with a hypothesis that links a business outcome to a data-centric question. The story unfolds as data scientists conduct research designed to answer that question, discovering insights from within data sources. The “happily ever after” takes the form of actions the business can implement to drive true transformation.
CSC has partnered with Microsoft to create and prove the data story approach. We’ve developed effective tools that ensure enterprise data is harnessed correctly. And we have a strong data science method – what we call the modern scientific method – to extract and visualize insights, turning information into actionable goals for the enterprise.
Together, we’ve built solutions for more than 80 data stories across six different industries: banking and capital markets, energy and technology, insurance, manufacturing, healthcare and retail. Here’s a look at a few of our successes:
Banking and Capital Markets
In banking and capital markets, we are building data stories that reduce fraud without disrupting a customer’s banking experience. By combining extracts from external data sources (such as social networks) with data from suspicious transactions, scalable machine learning algorithms are used to reduce the number of false positives flagged by existing fraud surveillance systems.
In retail, we are building data stories that help companies target customers with offers they truly want. We use detailed consumer transaction history to divide the market into micro-segments. Using continuously running campaign experiments to test those micro-segments, we can discover exactly which consumers to target for any given offer.
In manufacturing, we are building data stories that use intelligent digital simulation to find new innovations. Suppose, for example, a manufacturer wants to produce a line of cars that appeals to young professionals living in large cities. These customers often care about fuel efficiency and city mpg ratings. Although they can likely afford the additional expense of a hybrid car, they tend to want a lower-cost vehicle.
To help determine the best options for this vehicle, we deployed into production a digital twin with cognitive computing capability. We asked natural-language questions such as:
“What are the simulation runs with the best predicted city mpg and predicted five-year savings?”
The digital twin predicted the best fuel type, engine displacement, transmission and vehicle class options to optimize both long-term affordability and city miles per gallon. The resulting insights will allow the manufacturer to design a car that matches the needs of this unique market segment.
‘Writing’ a data story
CSC and Microsoft have teamed up to create a simplified and straightforward approach to executing data stories. We call it the Industrial Machine Learning (IML) utility.
IML is a modern take on the scientific method. We start with a hypothesis and collect data that could be useful in evaluating the hypothesis. We then generate a model and use it to explain the data. We evaluate the credibility of the model based on how well it explains the data observed so far, and predict how well it will explain new data collected in the future.
In our IML approach, we build intelligent, enterprise-scale applications in small sprints using the Cortana Intelligence Suite running on the Microsoft Azure cloud. With this system, we can produce reliable, measurable results that have real business impact.
Take, for example, the problem of shortening a patient’s length of stay in a hospital. By reducing recovery time, the patient can have a better health outcome, and the hospital is freed up to treat more patients. Costs are also lowered. But how can a hospital help patients recover faster?
We applied IML to the question, using Microsoft’s Data Factory to pipe batches of patient and medical procedure data into an Azure HDInsight data lake. We then extended the pipeline by streaming patient-generated data (from wearables, social media, etc.) into a Microsoft Event Hub, and we analyzed it using Microsoft’s Stream Analytics Jobs.
After we established a continuous, automated pipeline of data, we ran experiments, searching for leading indicators of lengthy procedure recovery and objectively scoring the performance of those indicators. Using Azure ML experiments, we identified the features that were most helpful in predicting a patient’s length of stay.
We trained a model based on those features, tuned the model parameters and scored the model by cross-validation. Finally, we distributed those insights across the entire network of healthcare providers using operational dashboards that alert hospital staff about future costs and help identify patients who might experience problems in recovery.
We used Azure tools to produce the dashboard. The Azure ML experiments automatically publish their results to a designated Azure Storage account. These insights were fed into Azure PowerBI dashboards to help care providers understand and digest the results.
When it comes to discovering insights, this method works consistently well — and the CSC-Microsoft IML solution allows it to be done quickly and on an enterprise scale.
So, what’s the sequel?
Using this combined CSC – Microsoft offering, enterprises can succeed where others have failed and derive real value from machine learning projects. But that’s not where this story ends; it’s just the next step in the journey to a cognitive future.
We know that technology is becoming better at “people tasks” than actual people. Cognitive computing has made it possible for machines to, in many case, outthink us, integrate broad information sets, find correlations and predict best possible outcomes. Cognitive computing is making its way into the enterprise, the boardroom, the physician’s office, the factory and beyond.
Successful cognitive initiatives start with a specific data story and grow out of successful pilot projects that prove the value of cognitive computing. With the approach I outlined here, CSC and Microsoft can help your business deliver immediate success and a vision for the future – and that’s a story still being written.
Jerry Overton is head of advanced analytics research in CSC’s ResearchNetwork and founder of CSC’s FutureTense competency, which includes the Predictive Modeling Research Group, Advanced Analytics Lab and Predictive Modeling School. Connect with him on Twitter.