Data-Driven Transformation at Microsoft

Large complex companies must pursue two kinds of transformation—digital transformation and data-driven transformation—to succeed in the digital economy. In digital transformation, a company reorients its business model to deliver digitized solutions and outstanding customer experiences. In data-driven transformation, a company learns to run on analytics-powered business processes, fueled by employees who habitually use data to inform their work and by automation that draws upon data science techniques. Digital transformation establishes the company’s direction and goals; data-driven transformation guides execution. 

Historically, large complex organizations have cultivated employee communities with entrenched practices to help the employees attain shared goals in similar ways. Although communities interact, they are each distinguishable by shared interests, activities, and skillsets. Communities that are data-oriented, i.e., “data communities,” are composed of employees with rich data-related skills, such as data manipulation and computational analysis. These employees share a passion for solving problems using evidence-based approaches.[foot]For an example data skill portfolio, see Wixom et al., “The Current State of Business Intelligence in Academia: The Arrival of Big Data,” Communications of the Association for Information Systems, 34 (2014).[/foot] “Domain communities” are composed of employees who share business-oriented goals for a defined domain (e.g., North American sales, consumer marketing.)

To achieve data-driven transformation, a company must establish data communities[foot]MIT CISR identified five ways that companies establish data communities: 1) local data communities grow organically; 2) a strong local data community is elevated to the enterprise level; 3) an enterprise-level data community is assembled from local community best practices; 4) an enterprise-level data community is grown organically; and 5) an enterprise-level data community is established through acquisition.[/foot] and then link them with its domain communities to create new and/or shared evidence-based work practices. Current MIT CISR research[foot]MIT CISR researchers conducted twenty-six interviews with Microsoft executives in Q1–Q3 2016. The researchers used a grounded theory approach to analyzing the interview data, and they drew upon dozens of Microsoft-provided and publicly available articles, documents, and other artifacts to triangulate their analyses.[/foot] has identified five common connecting structures that companies use to erase barriers between data and domain communities (see figure 1): (1) services, (2) advisory, (3) embedded, (4) multidisciplinary, and (5) networked. Services and networked structures promote shared behaviors and focus on maximizing scale economies. Multidisciplinary and embedded structures unify disparate teams to enable innovative behaviors. Advisory structures influence both innovation and efficiencies. Each type of connection is formed using reporting structures, workflows, or social ties. 

As described in a recent MIT CISR case study,[foot]I.A. Someh and B.H. Wixom, “Microsoft Turns to Data to Drive Business Success,” MIT Sloan CISR Working Paper No. 419, July 2017, https://cisr.mit.edu/publication/MIT_CISRwp419_MicrosoftDataServices_SomehWixom.[/foot] Microsoft utilized all five connecting structures to catalyze data-driven transformation. 

Achieving Data-Driven Transformation at Microsoft

Headquartered in Redmond, Washington, in 2017 Microsoft Corporation was an $85 billion technology company that employed 114,000 people and operated globally. In 2014, Satya Nadella became chief executive officer of Microsoft and accelerated its transformation into a cloud services company. Historically, Microsoft contained silos of exceptionally well-run businesses (e.g., Xbox, Windows) that focused on optimizing product revenues. To facilitate the transformation to services, Microsoft leadership consolidated core business functions (e.g., Sales, Marketing) to encourage open rather than siloed product-oriented perspectives. Management adjusted employee incentives so that one of the three core pillars of assessing an employee’s performance was based on how well he or she collaborates across the organization. And leadership communicated the importance of data in achieving Microsoft’s digital transformation goals. 

Microsoft's IT organization, comprising core enterprise services and dedicated teams, was instrumental in the company’s data-driven transformation.

Creating Generalized Practices

Microsoft IT established four enterprise shared services groups—Data, Data Science, Change Management, and Business Intelligence (BI)—to build common data-related capabilities. The groups identified and then promoted the widespread use of data, reports, algorithms, and processes across the company. The Data group, for example, helped create a “single source of truth” for data shared across Microsoft, such as financial metrics provided to Wall Street and a common list of countries. This group worked with local data governance committees to define, clean, and democratize common data. To influence user adoption, service leaders used marketing techniques including market segmentation, IT partnered with the newly consolidated core business campaign planning, and user survey collection. They learned functions to develop streamlined, evidence-based ways of how to identify and create strategies to remove “usage blockers," or barriers to adoption. This structured approach helped service leaders prioritize their efforts, measure progress, and nurture company-wide use of data-related capabilities.

Connecting structure: Microsoft’s data-related enterprise shared services groups advocated for the enterprise perspective. They harmonized data and created tools and process capabilities that reflected best practices and had widespread relevance and application across Microsoft.

Creating Customized Practices and Sharing Generalized Ones

Chief Information Officer Jim DuBois[foot]Jim DuBois stepped down as Microsoft CIO in July 2017.[/foot]established a three-person team to develop IT-specific metrics and dashboards. The dashboards contained traditional metrics, plus new ones such as actual services usage (drawn from telemetry data extracted from the cloud platform) and user sentiment regarding services (derived from textual data from crowdsourcing and enterprise social networking). DuBois recognized that the team’s expertise could benefit his executive peers, so he offered them the team's services free of charge. Executive access to real-time data via easy-to-use dashboards improved responsiveness and efficiency.

Connecting structure: DuBois’s advisory dashboard team, which became quite popular, catalyzed widespread use of dashboards at the senior management level.

Co-creating Customized Practices

IT partnered with the newly consolidated core business functions to develop streamlined, evidence-based ways of achieving desired business process outcomes. For the sales unit, IT helped create a new sales platform for managers, executives, and support staff that would cull and consolidate sales data to produce 360-degree views of Microsoft’s relationships with corporate customers. Sales leaders appointed a team to design multiple user interfaces, each customized to a unique sales persona, and created predictive models that forecast key sales process outcomes such as a sale’s likelihood of closure. The new system saved ten to fifteen minutes per sales opportunity by eliminating the need for Microsoft salespeople to manually search for and prepare data. Over time, the salespeople learned how to forecast more accurately, which led to better sales pipeline data and improved pipeline management.[foot]Barbara H. Wixom and Jeanne W. Ross, “How to Monetize Your Data,” MIT Sloan Management Review, January 9, 2017, http://sloanreview.mit.edu/ article/how-to-monetize-your-data/.

Connecting structure: Embedded data groups identified data and analytics requirements from the groups’ proximity to business processes. Equipped with strong business domain understanding, these dedicated units created intuitive, function-specific tools that helped users more easily draw upon meaningful data when performing work tasks.

Assembling Customized Practices 

Microsoft’s enterprise shared services groups’ penetration into business units offered insight into company needs and resources, which group leaders drew upon to assemble fruitful project-based collaborations. For example, the Data Science group and Facilities organization collaborated to create “smart” building heating and cooling solutions to explore ways to optimize the company’s energy consumption. 

In a few select cases, collaborations produced data services that were eventually productized and sold. For example, the Data Science group created cybersecurity solutions with the Security and Risk team and the Product team after rounds of experimentation and validation. Their collective work produced repeatable, scalable solutions that were offered to customers by Microsoft Consulting Services and Field Sales people. 

Connecting structure: Opportunities for Microsoft’s enterprise shared services groups to collaborate on projects arose from the groups’ regular engagement with business units. The groups and business units assembled multidisciplinary teams to jointly solve problems, identifying complementary business processes and data sources that deepened the impact of data and analytics approaches.

Amassing Customized Practices 

Enterprise shared services leaders at Microsoft relied on scalable communication mechanisms such as online office hours, email newsletters, and Yammer to reach large audiences.⁷ Yammer, an enterprise social networking service, was a particularly effective way to connect communities of employees to share experiences and support each other’s progress with data-driven work practices. To build active, BI-specific Yammer communities, the BI group identified champions across Microsoft who advocated for data and acted as trusted local experts. The group held competitions to incentivize sharing and participation. The BI group also monitored Yammer conversations for user sentiment regarding enterprise shared services offerings. 

Connecting structure: BI set up networked communities using a social networking service to permit community interaction. These communities created transparency—and surfaced previously elusive data practices across the company. 

CONVERGING CREATES VALUE 

By 2017, positive impacts from Microsoft’s transformational efforts were clear: 61 percent of Microsoft’s workforce was using the company’s self-service data offerings monthly, and the company’s stock price had nearly doubled since 2014. 

Microsoft’s achievements are not unexpected. First, the company made business model changes consistent with its aspiration to transform into a services company. Second, by leveraging connecting structures, the company facilitated convergence between its data and domain communities. 

Data-driven transformation calls for individuals from data and domain communities to connect, transfer knowledge, and change work practices. Changing work practices requires incentives, such as a stick, a carrot, or social pressure—or all of these, when the desired change is extreme. Convergence creates financial value for the company, either from innovative work practices that produce new value or shared work practices that create scale economies (see figure 2). 

When incumbent domain communities absorb data communities’ mindset, behaviors, and tools, the result is a data-driven company that pervasively monetizes, or creates financial value, from data. We consider this kind of company a data-driven high performer. Desperate attempts by non-digital companies at achieving this status—hiring a Chief Data Officer, establishing an analytics center of excellence, or acquiring a scrappy analytics business that models desired behaviors—mostly fall short. These moves alone do not reshape habits quickly or pervasively enough to generate data-driven transformation; they must be combined with connecting structures and associated strategies that deliberately erase barriers between data and domain communities, along with incentive to change.