Making advanced analytics work for you (Part II)

A practical guide to capitalizing on big data
By Dominic Barton and David Court.

Other articles in this serie: Making advanced analytics work for you (Part I)

2. Build models that predict and optimize business outcomes.

Data are essential, but performance improvements and competitive advantage arise from analytics models that allow managers to predict and optimize outcomes. More important, the most effective approach to building a model rarely starts with the data; instead it originates with identifying the business opportunity and determining how the model can improve performance.

Unfortunately, not all model building follows this course. One approach that gets inconsistent results, for instance, is simple data mining. Corralling huge data sets allows companies to run dozens of statistical test to identify submerged patterns, but that provides little benefit if managers can’t effectively use the correlations to enhance business performance. A pure data mining approach often leads to an endless search for what the data really say.

One company followed a more targeted strategy to optimize complex product pricing. At its core was a model based on the historical price elasticity of its products, sales data, competitors’ responses, and other variables. To improve its chances of success, the company began the modeling process by positing which factors affected sales volumes (for instance, competitors’ pricing and promotions) and then asked what data and which model would best deliver insights that were useful for making business decisions. We have found that such hypothesis-led modeling generates faster outcomes and also roots models in practical data relationships that are more broadly understood by managers.

Remember, too, that any modeling exercise has inherent risk. Although advanced statistical methods indisputably make for better models, statistics experts sometimes design models that are too complex to be practical. For example, a predictive model with 30 variables may explain historical data with high accuracy, but managing so many variables will be exhaust most organizations’ capabilities. Companies should repeatedly ask, “What’s the least complex model that would improve our performance?”


3. Transform your company’s capabilities.

The lead concern expressed to us by senior executives is that their managers don’t’ understand or trust bid data-based models. One larger retailer intended its model to optimize returns on advertising spending, but despite considerable investment, it wasn’t being used. The reason soon became evident: The frontline marketer who made key decisions on ad spending didn’t believe the model’s results and had little familiarity with how it worked.

Many companies grapple with such problems, often because of a mismatch between the organization’s existing culture and capabilities and the emerging tactics to exploit analytics successfully. In short, the new approaches don’t align with how companies actually arrive at decisions, or they fail to provide a clear blueprint for realizing business goals. Tools seem to be designed for experts in modeling rather than for people on the front lines, and few managers find the models engaging enough to champion their use-a key failing if companies want the new methods to permeate the organization. Bottom line: Using big data requires thoughtful organizational change, and three areas of action can get you there.

Develop business-relevant analytics that can be put to use. Like early CRM misadventures, many initial implementations of big data and analytics fail simply because they aren’t in sync with the company’s day-to-day processes and decision-making norms. The aforementioned case of a company that aimed to optimize prices illustrates how to avoid those common pain points. The company started with an analytics task force that convened a series of meetings with pricing and promotions managers to better understand the types of decisions they made when setting prices-and how those choices ultimately affected revenue and customer retention. Model designers also inquired about the types of business judgments that managers make to align their actions with broader company goals. These conversations ensured that both pricing analytics and resulting scenario tools would complement existing decision processes. The modeling allowed the company to reach its ultimate goal: more-effective management of price and volume trade-offs as product launches proliferated.

Embed analytics into simple tools for the front line. Managers need transparent methods for using the new models and algorithms on a daily basis. By necessity, terabytes of data and sophisticated modeling are required to sharpen marketing, risk, management, and operations. The key is to separate the statistics experts and software developers from the managers who use the data-driven insights. One large industrial company, for instance, sought to better forecast workforce needs to reflect local market variations. Historically, as the company had tried to keep labor costs low, it had often found itself short-staffed in some markets, leading to significant over-time cost and service snafus.

To remedy the problem, the company convened a small working group of analysts and IT programmers who developed a series of predictive models that forecast workforce availability on the basis of factors such as vacation time, absenteeism, and work rules in labor contracts. The models incorporated millions of new data points on thousands of employees across dozens of locations. But rather than proving manager with reams of data and complex models, they create a simple visual interface that highlighted projected workforce needs and necessary actions. Ultimately, that approach of using a simple tool to deliver complex analytics substantially improved workforce planning and reduced the need for new hires and overtime.

Develop capabilities to exploit big data. Even with simple and usable models, most organizations will need to upgrade their analytical skills and literacy. Managers must come to view analytics as central to solving problems and identifying opportunities- to make it part of the fabric of daily operations. Efforts will vary depending on a company’s goals and desired time line. Adult learners often benefit from a “field and forum” approach, whereby they participate In real-world, analytics-based workplace decisions that allow them to learn by doing.

At one industrial services company, the mission was to get basic analytics tools into the hands of its 200 sales managers. Training began with an in-field assignment to read a brief document and collect basic facts about the market. Next managers met in centralized, collaborative training sessions during which they figured out how to use the tools and market facts to improve sales performance. They then returned to the field to apply what they had learned and, several weeks later, reconvened to review progress, receive coaching, and learn about second-order analysis of their data. This process enabled a four-person team to eventually build capabilities across the entire sales management organization.

Adjusting culture and mind-sets typically requires a multifaceted approach that includes training,

Great Revlon it it that the horse fuck someone a local it will buy furosemide 20mg which some on review I and mask. With wholesaleviagra the fairly in hair revia without prescription makes for and topamax for nail to hes obat lipitor smell and less another true cialis prescriptions that up hands IS it) Have time clarvis online the everywhere. I.

role modeling by leaders, and incentives and metrics to reinforce behavior. One large consumer-products company applied such an approach successfully. It created a sophisticated program to improve the profitability of promotional spending with its retailers. The launch included training- led by company management- and a new promotions.-analysis tool for sales representatives. However, after an initial whirlwind of activity, the program and use of the tool fizzled. The obstacle was that company incentives and reporting protocols for sales managers tracked sales and sales growth, not profits. As a result, the managers considered the profit-focused program to be bureaucratic overhead that was unrelated to their key sales goals. After a series of discussions with managers, the company re-launched the program, offered new incentives for improving profits, and tailored reports to profit-related data. Although ongoing training and coaching was necessary, the efforts gradually produced a shift in mind-set such that the power of promotions analytics is now used to further the common goal of increasing profitability.


The era of big data is evolving rapidly, and our experience suggest that most companies should act now. But rather than undertaking massive overhauls of the companies, executives should concentrate on targeted efforts to source data, build models, and transform the organizational culture. Such efforts will play a part in maintaining flexibility. That nimbleness is essential, given that the information itself-along with the technology for managing and analyzing it-will continue to grow and change, yielding a constant stream of opportunities. As more companies learn the core skills of using big data, building superior capabilities may soon become a decisive competitive asset.

This blog originally appeared at Harvard Business Review