Content thumbnail How to get the most from big data

How to get the most from big data

Article by Matt Ariker, Peter Breuer, and Tim McGuire | McKinsey

Published on

Big Data and advanced analytics are critical topics for executives today. But many still aren't sure how to turn that promise into value. This presentation provides an overview of 16 examples and use ...

Big Data and advanced analytics are critical topics for executives today. But many still aren't sure how to turn that promise into value. This presentation provides an overview of 16 examples and use cases that lay out the different ways companies have approached the issue and found value: everything from pricing flexibility to customer preference management to credit risk analysis to fraud protection and discount targeting. For the latest on Big Data & Advanced Analytics:

  • 1. Big Data and Advanced Analytics 16 Use Cases From a Practitioner's Perspective June 27th, 2013 Workshop at Nasscom Conference - by Invitation Any use of this material without specific permission of McKinsey & Company is strictly prohibited
  • 2. McKinsey & Company | 1 Big Data and Advanced Analytics Pyramid Make your own data, for the problem at hand Unfamiliar, unstructured data. Acted upon directly, often at scale Unfamiliar, unstructured data. Converted into structured data. Acted upon at scale Familiar, structured data. Acted upon at scale D C B A
  • 3. McKinsey & Company | 2 A. Familiar structured data, acted upon at scale Selected examples Campaign lead generation - finding the leads that are most likely to result in incremental telecoms sales 2 Pricing - offering competitive prices only to the most sensitive retail deposit customers, while maximizing value 4 Pricing - create transparency into B2B chemicals prices, to enable more targeted price setting 1 Customer experience - knowing my hospitality customer's individual preferences, wherever the customer is travelling 3
  • 4. McKinsey & Company | 3 Moving from across the board pricing to differentiated targets, just using historical prices 0 100 200 300 400 500 600 700 800 900 1,000 Price per unit Account sales 100,00010,0001,000100 0 100 200 300 400 500 600 700 800 900 1,000 Account sales 100,00010,0001,000100 Price per unit Differentiated price targets One-size fits all price target DISGUISED EXAMPLE From across the board pricing increase ... ... to differentiated target-setting at customer- product reflecting customer's willingness to pay SOURCE: McKinsey Value Advisor team 1
  • 5. McKinsey & Company | 4 Telecoms companies are investing in big data infrastructure, bringing together data from diverse sources New services Government, urbanization, and social good Operations Marketing and sales Big Data available to Telcos Socio and economic analysis Health care and disease prevention ILLUSTRATIVE 2 Source: McKinsey Telecoms Practice, Integrated Incumbent Example
  • 6. McKinsey & Company | 5 Create an integrated picture of the household, and its product/ brand holdings Full household product holdingFamily of 4 Rikke Hansen Husband Kasper Hansen and their two kids Storkevænget 8 2840 Holte Household ID: 3512697 Customer ID: 3525300699 1 X Voice 1 X BB & Wifi 2 X Mobile 1 X Kids mobile 2 X Tablets X3 HH ARPU 1 X TV + HH services & solutions First pilots with +20-50% take-rates Source: McKinsey Telecoms Practice, Integrated Incumbent Example 2
  • 7. McKinsey & Company | 6 This picture allows the telco operator to target offers tailored to the product holdings at each household 15.0 HHs in the country 5.0 Non- customers 10.0 Currently customer in Group 2.0 Fully covered HHs1 2.5 Own mobile and fixed 4.0 Own fixed only 1.5 Own mobile only + Competitor product holdings Further sales potential DISGUISED EXAMPLE Most of this data was in the phone book 20 years ago, but was not actionable First pilots with +20-50% take-rates Source: McKinsey Telecoms Practice, Integrated Incumbent Example Brand-product holdings of all households Households, millions 2
  • 8. McKinsey & Company | 7 Hospitality: know your customer... ...everywhere in the world ▪ Commercial details - Employer relationship - Travel partnerships - Payment/ credit card ▪ Room preferences - No smoking - Pool view - Ground floor ▪ Personal preferences - Welcome drink - Entertainment ▪ Usage history - Internet usage - Fitness usage - Restaurant meals The hospitality industry captures and acts on customer preferences on a multi-national basis 3 ▪ Provide the same of personalized service - Cloud based architecture - Traditional architecture ▪ The data is not different from what was in a paper based system SOURCE: McKinsey Marketing practice
  • 9. McKinsey & Company | 8 Deposit pricing based on statistical estimates of sensitivity allows smart pricing at scale 4 Business as Usual ▪ Prices are set regionally ▪ Promotional pricing offered on new term deposits ▪ When promotional pricing lapses - Some customers leave - Other roll-over their deposits ▪ Both promotional prices and go-to prices have varied significantly - Across regions - Over time - Relative to competition With 1-2-1 Pricing approach ▪ Statistically predict customer's sensitivity to the price reduction ▪ Target the right price for each customer LowHigh Price sensitivity Fund for retention offers, if needed Interestrates 1-2-1 pricing Traditional price DISGUISED EXAMPLE SOURCE: McKinsey Banking CVM Service Line
  • 10. McKinsey & Company | 9 Applying individual level price elasticities can yield significant impact -200 +800 +1.000 +200 Evolution of price list interest rate and cost of fund- ing vs. competi- tors' price list Index Yearly TD volume growth (Million €) After 3 months: 1st step of differentiation with 2-5 ratesBefore 1-2-1 100 100100 101 99 96 93 9292 Competitors' top interest rate Bank's top interest rate Bank's average booked interest rate1 Growth vs. market After 6 months: 2nd step of differentiation with 10 rates -15% - +15% Δ=40bps Δ=90bps 1 Average of contracts opened or renewed in the period SOURCE: McKinsey Banking CVM Service Line 4 DISGUISED EXAMPLE
  • 11. McKinsey & Company | 10 Big Data and Advanced Analytics Pyramid Make your own data, for the problem at hand Unfamiliar, unstructured data. Acted upon directly. Sometimes at scale Unfamiliar, unstructured data. Converted into structured data. Acted upon at scale Familiar, structured data. Acted upon at scale D C B A
  • 12. McKinsey & Company | 11 -25 -20 -15 -10 -5 0 5 10 Print + Online Net of price (change) impact Competi- tors Online Competitors TV Price Announcements Price Gap Retention Losses Advanced Marketing Mix Modeling identifies the impact of marketing actions on sales/ churn Churn (retention) model Thousands of customers per month TV 4 5 3 2 6 7 1 DISGUISED EXAMPLE 5 SOURCE: McKinsey Marketing Practice
  • 13. McKinsey & Company | 12 This approach captures social media "buzz", such as comments on facebook and twitter, as marketing inputs Breakdown of drivers of customer acquisitions by marketing activity Percent Print special Print general Base, incl. price Negative Social Media TVSearchDisplayAffiliate 7.3 8.3 3.6 -7.88.5 9.0 8.4 EUR -36 million profit loss, can be fixed with EUR 0.8 million investment 5 DISGUISED EXAMPLE SOURCE: McKinsey Marketing Practice
  • 14. McKinsey & Company | 13 Supermarket purchase data (captured through loyalty programs) Mobile phone usage data (pre-paid or post-paid) SME supplier data (e.g., brewery supply to stores and bars) SME customer data (e.g., eBay) Utility data (e.g., electricity consumption and payment) Case example: A supermarket JV in Central America ▪ 3 models built using only supermarket transactions and age (as loyalty program captures date of birth): - Risk - Income - Need-based segmentation ▪ Risk model is used for pre-screening and selective pre-approval (GINI: 37) ▪ Income model is used to assign lines (45% correlation with payroll income) ▪ Segmentation is used to target customers for specific campaigns (e.g., credit card vs. personal loan for specific appliances on sale) Shopping basket data provided a Latin American bank with rich insights into credit risk in the unbanked segment 6 SOURCE: McKinsey Risk Practice
  • 15. McKinsey & Company | 14 Advanced Next Product To Buy (NPTB) algorithms integrate long term behavior with the most recent data to make smarter offers 7 Market Basket Analysis Bayesian Rules Engine Classical market basket analysis is well known from leading "bricks and mortar" retailers "Market Basket Analysis" links conditions with product uptake, e.g., IF affluent AND increased monthly salary AND Family with <__> THEN x% probability of hiring mortgage Basket = single trip, i.e. customers buy A and B together in one trip Basket = all purchases of one customer within last year(s), i.e. customers who read A also read B "Market basket" includes years of transaction history, as well as the most recent web-browsing behavior Next recommendationPurchase history Next recommendationPurchase history Next recommendation Product portfolio Transactional behavior Contact history and other Socio-demogr aphics Basket = collection of customer specific data including ... iPhone SOURCE: McKinsey Marketing Practice
  • 16. McKinsey & Company | 15 Next-Product-to-Buy probabilities guide in branch/ store sales efforts, promotions and product recommendations SOURCE: McKinsey Marketing Practice Customer Likelihood of buying in the next month % by product Long term loans 32% Owns 15% Savings account Owns 89% Owns Pensions 39% 22% 15% Short term loans Owns 10% 21% Cards Owns 12% 40% Current account 87% 64% 60% Invest- ment funds Owns 97% Owns Product Probability 87% Investment fund 97% Current account 95% Recommendation - Customer View Current account Customer Probability 40% 32% 97% Recommendation - CLV view Recommendation engines (next product/service/ application) to buy can deliver 3-5% revenue uplift Long term loans Investment fund Credit cards 7 DISGUISED EXAMPLE
  • 17. McKinsey & Company | 16 Cross channel data integration tools like Click Fox* now allow firms to see customers' experiences across channels Business as Usual Multiple customer touch points, each with its own infrastructure and data Click Fox Integration* Brings the diverse data sources together Organizes into meaningful customer journeys Customer Journeys Product feature search Specific non-standard business process Service or dispute resolution Intuitive use case Manage customer experience and satisfaction Emerging use cases Estimate credit risk Estimate churn likelihood Target preferred channels 8 SOURCE: McKinsey Marketing Practice; * McKinsey has invested in Click Fox
  • 18. McKinsey & Company | 17 Big Data and Advanced Analytics Pyramid Make your own data, for the problem at hand Unfamiliar, unstructured data. Acted upon directly. Sometimes at scale Unfamiliar, unstructured data. Converted into structured data. Acted upon at scale Familiar, structured data. Acted upon at scale D C B A
  • 19. McKinsey & Company | 18 C. Unfamiliar or unstructured data, acted upon at scale, directly Selected examples Discount targeting - using location data to offer discount coupons redeemable to the nearest store 10 Discount targeting - using transactional spending data from banks or networks 12 Fraud prevention - by matching the location of mobile phone with a credit or debit card transaction 9 Discount targeting - using speech analytics to identify customers who are most likely to attrite 11 13 Advertising targeting - using browsing history to target web site visitors with the most relevant adverts
  • 20. McKinsey & Company | 19 Joint venture between a telecom operator and credit card issuer uses location information to reduce fraud How banks use telco data to fight fraud... ▪ A large EU bank is leveraging data from a Telco company to identify fraud by crossing Card transactions and mobile data The Bank sees a transaction in Spain from your Card The Telco operator knows you are in Norway Fraud? 9 ...but are still trying to forge a workable governance approach Does the telco have the right to sell a customer's location to a bank? Is the bank obliged to take the data from the telco? Preventing that fraud, help protect other banks and customers? Should the customer opt-in and instruct the bank to get this data from the telco? What does it cost? Who should pay? SOURCE: McKinsey Banking Practice, Interviews with industry experts
  • 21. McKinsey & Company | 20 A leading payment network's joint venture with a retailer is an early example of using real-time, location data to target customer offers SOURCE: Mobile Marketing Watch A payment network and a retailer use real-time transaction data to make time and location sensitive offers to customers... ... made possible by information management capabilities ability to process and analyze transactions in real-time...goes well beyond processing purchases to delivering critical information that benefits consumers, merchants, a nd financial institutions. ▪ Other payment providers and start-ups looking to provide similar offerings ▪ Role banks will play in this trend is not yet clear SMS coupon for closest retail outlet is pushed immediately to customer Customer makes purchase at juice bar while shopping using the network's card Customer can use coupon in the nearest retail location The network matches customer, purchase location, and qualifying offer in real-time 1 2 4 3 10
  • 22. McKinsey & Company | 21 Speech analytics enables Telco operators to analyze phone calls to help provide more tailored offers to customers 11 The privacy and legal questions are typically harder than the analytics Some companies are testing technology to analyze telephone calls on their network (electronically)... ... and trigger responses or actions based on key words ...and gets a long-duration retention offer? ...and gets a promotion on a Caribbean cruise? Customer mentions the name of a competitor... Customer mentions a forth- coming holiday... SOURCE: McKinsey interviews with industry experts
  • 23. McKinsey & Company | 22 Banks and card companies are looking to release the value in their transaction data assets Merchant-funded reward programs Banks are monetizing behavioural data to deliver highly-targeted offers to customers Insight development and services Card networks provide analytics services for merchants, that range from pre-packaged reports to customized consultations Customers opt in Offer Bank matches relevant offers & customers Offer Bank matches relevant offers & customers Reinforcement Response tracked & shared with merchants and customers Reinforcement Response tracked & shared with merchants and customers Redemption Offer redeemed by customers (e.g., mobile download, real- time at POS) Merchants set customer aspiration Presentment Offer presented to customers Presentment Offer presented to customers Full or pilot programs at multiple banks in North America Example services: ▪ Benchmark analytics: analyzes merchant performance against the industry category, or specific business competitors ▪ Portfolio Analytics: provides users access to own transactional data through an extensive set of reports in multiple categories ▪ Macro-economic spend indicators: reports consumer spending in multiple vertical industries, based on aggregate activity in the payments network, coupled with estimates for cash and checks 12 SOURCE: McKinsey Banking Practice, Interviews with industry experts
  • 24. McKinsey & Company | 23 A leading US based bank uses web browsing data to serve targeted pages to prospects or visitors Risk models Segment specific websitesInputs - information used How they enter the site Internet specific data Surfing history ▪ Use internet data and score customers before the website is loaded ▪ Need to score customers on models in <0.5s Low risk saver Higher risk borrower ▪ Location ▪ Aggregated social media data ▪ Cookie information on past sites visited ▪ Some sites associated with low risk ▪ Other sites with higher risk e.g. social media ▪ Natural search ▪ Sponsored search - Brand names, credit terms ▪ Banner adverts ▪ Aggregators 13 SOURCE: McKinsey Banking Practice, Interviews with industry experts
  • 25. McKinsey & Company | 24 Big Data and Advanced Analytics Pyramid Make your own data, for the problem at hand Unfamiliar, unstructured data. Acted upon directly. Sometimes at scale Unfamiliar, unstructured data. Converted into structured data. Acted upon at scale Familiar, structured data. Acted upon at scale D C B A
  • 26. McKinsey & Company | 25 D. Unfamiliar or unstructured data, acted upon at scale, directly Selected examples Pricing and Advertising targeting - learning the right price (i.e. odds) and the right landing page to show each visitor to a gaming website, using on-going experimentation 15 Advertising targeting - learning the right landing page to show each visitor, using on-going experimentation 14 Credit line management - learning the right credit line to both profitably and responsibly offer each account, with on-going experimentation 16
  • 27. McKinsey & Company | 26 A leading European Bank runs experiments on the bank's web site to find out which visitor should be served which page 14 Customer logs on to bank website Bank knows product holding, segment, his torical behavior Plus recent product enquiry/ browsing behaviour "Champion" offer suite ▪ Shown ~90% of the time ▪ Reflects, recent behavioural history, long term product holdings and customer segmentation ▪ Maximizes value, based on current knowledge "Challenger" screens #1-5 ▪ Each shown ~2% of the time ▪ Learn if customer preferences or market conditions have changed SOURCE: Bank investor day presentation + Behavioural targeting results in a 27% lift in banner click through, and 12% increase in sales
  • 28. McKinsey & Company | 27 Similarly, a gaming web site uses browsing history AND experimentation to learn about the right offers 15 ILLUSTRATIVE Customer logs on to gaming website Website knows his historical behavior "Champion" screen ▪ Shown 95% of the time ▪ Maximizes value, based on current knowledge "Challenger" screens #1-5 ▪ Shown 1% of the time ▪ Learns of customer's preferences or market conditions have changed SOURCE: McKinsey Marketing practice, Industry interviews
  • 29. McKinsey & Company | 28 Industry leaders invest making-your-own-data, even in sensitive areas like credit line management SOURCE: McKinsey Risk Practice, Industry interviews 1 2 3 4 5 6 7 8 9 10 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 Low line test: 2.5% of applicants Optimal credit line: 95% of credit applicants High line test: 2.5% of applicants ILLUSTRATIVE Credit line allocation by risk band Currency units ($, £ etc.) Low risk applicantsHigh risk applicants Industry leaders invest millions of $s in champion-challenger experiments, that mature over 3 or more years, to learn how their strategies can be improved 16
  • 30. McKinsey & Company | 29 Education  BA from St. Stephen's College, Delhi  MBA from University of Chicago, Graduate School of Business Work experience McKinsey experience includes:  Consumer banking  B2B marketing Functional focus Advanced Analytics, Customer Lifecycle Management, Credit Risk Sectors Financial Services, Consumer Products Prithvi Chandrasekhar Senior Expert, Marketing, London Office @McK_CMSOForum Stay Connected:

How to get the most from big data - Page 7
How to get the most from big data - Page 10

Over the past 30 years, most companies have added new C-level roles in response to changing business environments. The chief financial officer (CFO) role, which didn't exist at a majority of companies in the mid-1980s, rose to prominence as pressures for value management and more transparent investor relations gained traction. Adding a chief marketing officer (CMO) became crucial as new channels and media raised the complexity of brand building and customer engagement. Chief strategy officers (CSOs) joined top teams to help companies address increasingly complex and fast-changing global markets.

Today, the power of data and analytics is profoundly altering the business landscape, and once again companies may need more top-management muscle. Capturing data-related opportunities to improve revenues, boost productivity, and, sometimes, create entirely new businesses puts new demands on companies-requiring not only new talent and investments in information infrastructure but also significant changes in mind-sets and frontline training. It's becoming apparent that without extra executive horsepower, stoking the momentum of data analytics will be difficult for many organizations.

Because the new horizons available to companies typically span a wide range of functions, including marketing, risk, and operations, the C-suite can evolve in a variety of ways. In some cases, the solution will be to enhance the mandate of the chief information, marketing, strategy, or risk officer. Other companies may need new roles, such as a chief data officer, chief technical officer, or chief analytics officer, to head up centers of analytics excellence. This article seeks to clarify the most important tasks for executives playing those roles and then sets out some critical questions whose answers will inform any reconfiguration of the C-suite. Daunting as it may seem to rethink top-management roles and responsibilities, failing to do so, given the cross-cutting nature of many data-related opportunities, could well mean jeopardizing top- or bottom-line growth and opening the door to new competitors.

Six top-team tasks behind data analytics

Crafting and implementing a big-data and advanced-analytics strategy demands much more than serving up data to an external provider to mine for hidden trends. Rather, it's about effecting widespread change in the way a company does its day-to-day business. The often-transformative nature of that change places serious demands on the top team. There's no substitute for experienced hands who can apply institutional knowledge, navigate organizational hazards, make tough trade-offs, provide authority when decision rights conflict, and signal that the leadership is committed to a new analytics culture. In our experience, the concerted action that's required falls into six categories. Leaders should take full measure of them before assigning responsibilities or creating roles.

Establishing new mind-sets

Senior teams embarking on this journey need both to acquire a knowledge of data analytics so they can understand what's rapidly becoming feasible and to embrace the idea that data should be core to their business. Only when that top-level perspective is in place can durable behavioral changes radiate through the organization. An important question to ask at the outset is "Where could data analytics deliver quantum leaps in performance?" This exercise should take place within each significant business unit and functional organization and be led by a senior executive with the influence and authority to inspire action.

Leaders at one large transportation company asked its chief strategy officer to take charge of data analytics. To stretch the thinking and boost the knowledge of top managers, the CSO arranged visits to big data-savvy companies. Then he asked each business unit to build data-analytics priorities into its strategic plan for the coming year. That process created a high-profile milestone related to setting real business goals and captured the attention of the business units' executives. Before long, they were openly sharing and exploring ideas and probing for new analytics opportunities-all of which helped energize their organizations.

Defining a data-analytics strategy

Like any new business opportunity, data analytics will underdeliver on its potential without a clear strategy and well-articulated initiatives and benchmarks for success. Many companies falter in this area, either because no one on the top team is explicitly charged with drafting a plan or because there isn't enough discussion or time devoted to getting alignment on priorities. At one telecommunications company, the CEO was keen to move ahead with data analytics, particularly to improve insights into customer retention and pricing. Although the company moved with alacrity to hire a senior analytics leader, the effort stalled just as quickly. To be sure, the analytics team did its part, diving into modeling and analysis. However, business-unit colleagues were slow to train their midlevel managers in how to use the new models: they didn't see the potential, which, frankly, wasn't part of "their" strategic priorities.

As we have argued previously, capturing the potential of data analytics requires a clear plan that establishes priorities and well-defined pathways to business results, much as the familiar strategic-planning process does. Developing that plan requires leadership. At a North American consumer company, the CEO asked the head of online and digital operations, an executive with deep data knowledge, to create the company's plan. The CEO further insisted that it be created in partnership with a business-unit leader who was not familiar with big data. This partnership-combining a data and analytics expert and an experienced frontline change operator- ensured that the analytics goals outlined in the plan were focused on actual, high-impact business decisions. Moreover, after these executives shared their progress with top-team counterparts, their collaborative model became a blueprint for the planning efforts of other business units.

Determining what to build, purchase, borrow, or rent

Another cluster of decisions that call for the authority and experience of a senior leader involves the assembly of data and the construction of advanced-analytics models and tools designed to improve performance. The resource demands often are considerable. With multitudes of external vendors now able to provide core data, models, and tools, top-management experience is needed to work through "build versus buy" trade-offs. Do strategic imperatives and expected performance improvements justify the in-house development and ownership of fully customized intellectual property in analytics? Or is reaching scale quickly so important that the experience and talent of vendors should be brought to bear? The creation of powerful data assets also can require the participation of senior leadership. Locking in access to valuable external data, for instance, may depend on forging high-level partnerships with customers, suppliers, or other players along the value chain.

The radically diverging paths different retailers have chosen underscore the range of options leaders must weigh. Several retailers and analytics firms have established long-term contracts covering a broad sweep of analytics needs. Other large players, both brick-and-mortar and online, have invested in deep internal data and analytics expertise. Each of these choices reflects a dynamic set of strategic, financial, and organizational requirements that shouldn't be left to middle management.

Securing analytics expertise

Under almost any strategic scenario, organizations will need more analytics experts who can thrive amid rapid change. The data-analytics game today is played on an open and (frequently) cloud-based infrastructure that makes it possible to combine new external and internal data readily and in user-friendly fashion. The new environment also requires management skills to engage growing numbers of deep statistical experts who create the predictive or optimization models that will underwrite growth.

The hunt for such talent is taking place in what has become the world's hottest market for advanced skills. Retaining these valued employees and then getting them to connect with business leaders to make a real difference is a true top-management task-one that often demands creative solutions. The leader of a big-data campaign at a major consumer company, for instance, decided to invest in an analytics unit distant from company headquarters. This other locale had abundant talent and a cultural environment preferred by data scientists and engineers. The leader then closed the loop, ensuring that each unit of the analytics team had a direct connection to a business-unit team at the company.

Mobilizing resources

Companies often are surprised by the arduous management effort involved in mobilizing human and capital resources across many functions and businesses to create new decision-support tools and help frontline managers exploit advanced analytics models. An empowered senior player is vital to breaking down the institutional barriers that frequently hamper efforts to supercharge decisions through data analytics. Success requires getting a diverse group of managers to coalesce around change-encouraging alignment across a wide phalanx of IT, business-lines, analytics, and training experts. The possibility of failure is high when companies don't commit leadership.

Take the example of a second transportation company, where middle managers across product areas were tasked with identifying data-analytics opportunities and then pushing them forward. The analytics managers were routinely frustrated when data teams failed to deliver data on schedule or in usable formats. When it came time to embed the resulting analytics into customized tools, managers faced additional frustrations as urgent requests worked their way through routine budgeting and planning processes. The company gave the task of stepping up the pace of its analytics agenda to a top marketing and sales executive, who assembled cross-functional teams including database managers, analysts, and software programmers. The teams rotated across analytics opportunities, steering them from launch to implementation in six- to eight-week bursts. Through this rapid mobilization, the company checked off several analytics priorities only months after the marketing leader took charge.

Building frontline capabilities

The sophisticated analytics solutions that statisticians and scientists devise must be embedded in frontline tools so simple and engaging that managers and frontline employees will be eager to use them daily. The scale and scope of this adoption effort-which must also involve formal training, on-the-job coaching, and metrics that clearly define progress-shouldn't be downplayed. In our experience, many companies spend 90 percent of their investment on building models and only 10 percent on frontline usage, when, in fact, closer to half of the analytics investment should go to the front lines.

Here, again, we have seen plenty of cases where no one on the top team assumed responsibility for sustained ground-level change. Lacking senior accountability and engagement, one financial-services company weathered several waves of analytics investment and interest only to have efforts fizzle when training and adoption fell short. Dismayed, business-unit leaders then took charge, investing in ongoing training sessions for managers and end users, pushing for the constant refinement of analytics tools, and tracking tool usage with new metrics. Over time, thanks to the consistent application of analytics, the transformation effort gained the hoped-for momentum.

Putting leadership capacity where it's needed

As companies size up these challenges, most will concede that they need to add executive capacity. But that leaves unanswered important decisions about where, exactly, new roles will be located and how new lines of authority will be drawn. As we'll outline below, our experience shows that companies can make a strong case for leading their data-analytics strategies and talent centrally or even for establishing a formal data-analytics center of excellence. However, frontline activities (mobilizing resources, building capabilities) will need to take place at the business-unit or functional level, for two reasons. First, the priorities for using data analytics to increase revenues and productivity will differ by business. Second, and just as important, companies best catalyze frontline change when they connect it with core operations and management priorities and reinforce it with clear metrics and targets.

Beyond this bias for pushing frontline mobilization responsibility to business units, there is no single prescription for where and how a company should add leadership capacity. Given the relative immaturity of data-analytics applications, that shouldn't be surprising. Yet as leaders review their options, they needn't fly blind. Pushing for answers to three key questions, in our experience, brings strategic clarity to the needed organizational changes:

  1. Will a central customer or operational database be used across business units?
  2. Is there a compelling need to build substantial analytics resources internally to retain talent and build proprietary assets and advantages?
  3. Within each business unit, can the current functional executives handle the change-management challenge or should the company dedicate new executive capacity specifically for the data-analytics change effort?

We'll illustrate the importance of these issues through examples of companies that have addressed them in different ways.

When central data assets are key

At many consumer-services businesses, exploiting analytics involves combining transaction data across a number of businesses or channels. That approach allows these companies to shape insights such as how consumers engage with Web sites or decide between shopping online or in stores. These companies often have (or are building) new central data warehouses or data environments, as well as related data-management capabilities. In addition, they often are working through new rules of the road on issues such as how they can access data while protecting consumer privacy or ensure that key customers aren't hassled by unnecessary contacts.

In such cases, an enhanced role for the CIO-spearheading the development of the data-analytics strategy and talent building-is a popular path. Operationally, the CIO takes charge of efforts to develop the data and analytics infrastructure while letting the business units mobilize change aimed at exploiting it.

At one multibusiness consumer-services company, for instance, the board and senior-leadership team recognized that a significant step-up in performance could be achieved if it fully exploited analytics opportunities across business lines by harnessing its multichannel databases. Recognizing the overarching role that the central databases play in the company's agenda, the leadership designated the chief information officer to direct the effort and to define the data and analytics strategy.

The leaders realized that each business unit, by necessity, would have its own targeted analytic priorities, such as strengthening promotional offers or optimizing inventory levels. Moreover, a different group of managers would be applying the insights across business units. The leadership concluded that under these circumstances, managing analysis and frontline training from the center would be a mistake and decided instead that the CIO should partner with business-unit leaders, sharing with them a tiered set of responsibilities.

At present, the CIO is immersed in two key projects. The first is creating a new infrastructure that unites the company's multichannel transaction data with external social-media and competitive information and delivers the result to business units through an intuitive interface. The second involves building up analytics expertise that can be assigned to different business units but managed centrally, at least for the next couple of years as the effort gains critical mass. The analytics team is led by a deeply experienced executive who reports to the CIO and provides a crucial injection of top-management capacity. In parallel, business-unit leaders are hammering out analytics priorities and building the skills of frontline managers who will use new models to, for example, redirect spending across media channels.

When substantial internal analytics expertise is core to performance

We are also seeing a second approach, which shares some of the centralized aspects we touched on above but specifically involves companies that decide to build rather than outsource a critical body of advanced analytics expertise. That decision often leads organizations to locate the expertise centrally, where it serves as a common platform for creating value across business units.

At one consumer-facing company, analytics expertise and leadership were concentrated in the finance and risk-management team, which historically had accounted for significant data-related value creation. When the company began pursuing a more aggressive analytics strategy, the CFO took responsibility for several tasks, including defining the basic strategy, overseeing make-versus-buy decisions for the core risk-management analytics tools, mobilizing resources within the function's analytics team, and building expertise.

However, having made these primary decisions about analytics, the CEO and CFO soon realized that significant complementary efforts were needed to secure better data for the analytics team and to reinforce change efforts and revamp several processes across the business units. To lead these initiatives, they established a new position-chief data officer-within the CFO's organization. This CDO proactively manages information, working with business managers to identify both internal and external data they may not even realize exists. Delivered ready for analysis, the data can be applied rapidly to needed tasks by modeling experts and, just as important, continually refreshed for new experiments and broader application. Many companies may find they need this type of leadership to support business leaders as they identify sources of data-driven advantages, work through analytics priorities, and try to accelerate frontline adoption.

When managing scale and complexity within business units is paramount

Whether elements of the effort are managed centrally or not, much of the data-analytics heavy lifting will fall on business or functional leaders within individual business units. A core question at the business-unit level is whether to add a new role or ask a key functional leader (such as the CMO or the head of operations) to add new responsibilities to what in all likelihood is already a pretty full plate.

When the senior leaders of a large financial-services company took a wide-ranging look at its strategy, they decided that one business unit could gain a significant competitive edge if it doubled down on data analytics. To push the strategy ahead decisively, the company recruited a chief analytics officer, who reports to the business-line president and oversees a new center of excellence drawing on internal consultants, analytics modelers, and software programmers.

This approach, which represents a significant organizational change, is accelerating the business unit's data-transformation effort. As a top-team member, the CAO can drive a broad range of decisions, from setting analytics strategy to defining the responsibilities of frontline managers. Since the center of excellence spans multiple disciplines, the CAO can mobilize analytics and software-programming resources swiftly, which has sped up the creation of frontline tools. Meantime, operating from within the business unit has given him a deeper understanding of what makes it tick-its priorities, patterns of working, and ongoing challenges. This has paid off in sharper decisions about which tools to develop and a keener sense of the skills that training programs need to foster. The fact that the business unit's leaders are engaged with the CAO on a day-to-day basis helps keep them focused on their analytics and adoption agendas.

Building on this success, the company has recently taken the further step of adding another new role, a chief data officer, who reports to the CIO but works daily with the chief analytics officer to help knit together data and new analytics tools and to speed frontline change.

For companies pursuing the potential of data analytics, a decision about leadership capacity looms-regardless of where in the end they decide to place it. For some, such as the consumer-facing companies described earlier, current top-team members will be asked to step up and assume broader leadership responsibilities, often with additional support from new, senior lieutenants. For others, such as the financial-services company we explored, establishing one or more new senior posts to drive the analytics agenda will be the best solution.

At all companies, top teams, and probably board members as well, need a better understanding of the scale of what's needed to ensure data-analytics success. Then they must notch these responsibilities against their existing management capacity in a way that's sensitive to the organization's core sources of value and that meshes with existing structures. None of this is easy, but it's the only serious way to pursue data analytics as a new frontier for growth.

About the authors

Brad Brown is a director in McKinsey's New York office, David Court is a director in the Dallas office, and Paul Willmott is a director in the London office.

The authors would like to acknowledge the contributions of Matthew Ariker, Amit Garg, Joshua Goff, Lori Sherer, and Isaac Townsend to the development of this article.

How to get the most from big data - Page 14

After the 2013 holiday season, retailers were notably concerned about lower shopping volumes. There was a silver lining, though. As one digital marketing agency headlined a November 19 blog post, "The Holiday Season's Greatest Gift Isn't Big Cash. It's Big Data."

Simply collecting Big Data does not unpack its potential value. People need to do that, and those people are hard to come by.

Wal-Mart already captures from customer transactions more than 2.5 petabytes (2.5 quadrillion bytes-a 16-digit figure) every hour. That is, every 60 minutes, Wal-Mart stuffs the equivalent of 20 million four-drawer filing cabinets with data. Fifteen out of 17 US sectors now have more data stored per company than the Library of Congress. A McKinsey Global Institute study finds that this single data category has "the potential to provide more than $800 billion in economic value to individual consumers over the next decade." Some gift! Provided companies can successfully unwrap and actually make use of it.

Simply collecting Big Data does not unpack its potential value. People need to do that, and those people are hard to come by. Just 3.4 % of CMOs surveyed by McKinsey in 2013 believe they currently have the right talent to fully leverage marketing analytics, and 98.8% describe finding that talent as "challenging." In consequence, while few dispute the potential value of Big Data, CMOs surveyed say they use marketing analytics just 29% of the time to make decisions, and a paltry 3% say that analytics contributes "very highly" to their company's performance.

Finding Talent: Look beyond the usual suspects

The University of California, Berkeley, recently announced an online Master of Data Science degree, and, in August, IBM unveiled a Big Data educational partnership with more than a thousand colleges and universities. Such academic forays into data science are noteworthy because they are exceptional. Over the long term, we're hopeful that these sorts of initiatives will prove to be the first of many and that they will at least begin to meet the demand for data science talent.

The most effective Big Data specialists are "translators" capable of bridging different business functions and effectively communicating between them.

For now, however, many companies are trying to fill their ranks with candidates holding degrees in computer science, industrial engineering, and statistics. But there are just not enough of them, so employers need to improvise. Major companies are starting to turn to disciplines as diverse as physics, philosophy, psychology, economics, and even biostatistics to find Big Data talent. They seek people with analytical minds and deep curiosity - critical talents for working with data in business. We know of companies, for example, that have successfully transformed computational fluid dynamics engineers and West Point-trained ordnance engineers into data scientists.

Find translators and bridge builders

Cracking the Big Data enigma requires more than number crunchers. To turn the fruits of Big Data insight into value in the marketplace calls for an extensive mix of expertise. Companies need to fill multiple roles with specific skillsets. Specialist competence is essential but not sufficient. Look for specialists who are also "translators," capable of bridging different business functions, comfortably and effectively communicating between them. "Campaign experts," for example, focus on turning data models into specific marketing campaigns.

You need people with multiple skillsets, but, realistically, it is very rare to find someone with all the skills you need. More feasible is finding people with at least two essential skills. Think of them as "two-sport data athletes," and look for such combinations as computer programming and finance, statistics and marketing, psychology and economics-just to name a few.

A good data scientist is a uniquely valuable professional whom other companies will inevitably covet.

The next aspect to consider is recruiting for the entire Big Data "value chain." This will ensure that you have the people you need to get the insights all the way to the front lines. In addition to the core data scientists, whose prime expertise is creating predictive models from data, the team should include "data hygienists," who give preliminary form to unstructured data by ensuring that it is clean and accurate. They hand off to "data explorers" who filter data to identify information actually capable of predictive analysis. "Business solution architects" further structure this filtered data, prioritizing material most likely to yield actionable business insights in the hands of the data scientists.

Does every team require separate people for each role? No. But each role represents an essential process, and so every role must be cast and played.

Address your retention issue

Having recruited a data scientist, a "translator," or a two-sport athlete, recognize that you have a uniquely valuable professional, who other companies will inevitably covet. Research competitive compensation levels. You will want to make a credible financial offer-including an upward career path. This latter part can be a problem, since most data analysts do not aspire to become managers, the traditional business career objective. Better get creative, then, and offer an alternative. One financial services company, for example, created roles designated "subject matter experts," which command the pay and prestige of a senior manager.

Recruiting for the entire Big Data "value chain" will ensure that you have the people you need to get the insights all the way to the front lines.

Often as important as money is providing an environment of intellectual challenge, collegiality, and extensive connection with those outside of the analytical team. Like many of us, analytical folk want to feel they're having a positive impact on the business rather than being held in a back room to collect and analyze data. Data teams need to be integrated into the company with a full appreciation of the value they create for the entire business. Seasoned analysts accept the reality that, most of the time, they are expected to work backwards from a marketing and sales decision to provide insight on its impact (or lack thereof). But if 80% of the analytical task is "decision-backward," it pays to engage your insight team by allocating at least 20% of the work to innovation, challenging and engaging analysts to create "data-forward" insights that lead decisions strategies rather than merely course-correct existing strategies.

As data increasingly drives businesses, getting-and keeping-your Big Data people behind the steering wheel will require innovative ways to recruit, retain, and inspire them.

It's hard to overstate the importance of getting pricing right. On average, a 1 percent price increase translates into an 8.7 percent increase in operating profits (assuming no loss of volume, of course). Yet we estimate that up to 30 percent of the thousands of pricing decisions companies make every year fail to deliver the best price. That's a lot of lost revenue. And it's particularly troubling considering that the flood of data now available provides companies with an opportunity to make significantly better pricing decisions. For those able to bring order to Big Data's complexity, the value is substantial.

The secret to increasing profit margins is to harness Big Data to find the best price at the product-not category-level.

We're not suggesting it's easy: the number of customer touchpoints keeps exploding as digitization fuels growing multichannel complexity. Yet price points need to keep pace. Without uncovering and acting on the opportunities Big Data presents, many companies are leaving millions of dollars of profit on the table. The secret to increasing profit margins is to harness Big Data to find the best price at the product-not category-level, rather than drown in the numbers flood .

Too big to succeed

For every product, companies should be able to find the ideal optimal price that a customer is willing to pay. Ideally, they'd factor in highly specific insights that would influence the price-the cost of the next- best competitive product versus the value of the product to the customer, for example-and then arrive at the best price. Indeed, for a company with a handful of products, this kind of pricing approach is straightforward.

It's more problematic when product numbers balloon. About 75 percent of a typical company's revenue comes from its standard products, which often number in the thousands. Time-consuming, manual practices for setting prices make it virtually impossible to see the pricing patterns that can unlock value. It's simply too overwhelming for large companies to get granular and manage the complexity of these pricing variables, which change constantly, for thousands of products. At its core, this is a Big Data issue (exhibit).

Many marketers end up simply burying their heads in the sand. They develop prices based on simplistic factors such as the cost to produce the product, standard margins, prices for similar products, volume discounts, etc. They fall back on old practices to manage the products as they always have, or cite "market prices" as an excuse for not attacking the issues. Perhaps worst of all, they rely on "tried and tested" historical methods, such as a universal 10 percent price hike on everything.

"What happened in practice then was that every year we had price increases based on scale and volume, but not based on science," says Roger Britschgi, head of sales services at Linde Gases . "Our people just didn't think it was possible to do it any other way. And, quite frankly, our people were not well prepared to convince our customers of the need to increase prices."

Read the full article on McKinsey Insights & Publications

How to get the most from big data - Page 17
How to get the most from big data - Page 18
How to get the most from big data - Page 19
How to get the most from big data - Page 20
How to get the most from big data - Page 21