How to Create a Successful Data Strategy

With rapid advances in AI and data science, data has become an essential asset to every enterprise. Setting up a data strategy, therefore, has become every enterprise’s mission, particularly in the C Suite and at Executive levels. What is a data strategy and how do we create the right data strategy? I would like to dedicate this article to answer these 2 questions.

.

Before discussing data strategy, we need to understand what a strategy is. Using a simplified definition, a strategy is a thoughtful plan focused on changing the current state in order to reach a vision for the future. In other words, the right strategy needs to start with a vision, and the strategy is a way of making a series of changes, usually requiring innovation and out-of-the-box thinking, to achieve the vision. Every enterprise must have a business vision and a business strategy in place before having a data strategy. A data strategy should go in hand in hand with a business strategy and serve to realize the business vision. On the other hand, data lives with technology, while providing value to businesses and customers. Data strategy, therefore, is also the strategy of both data and technology.

.

Given this definition of what a data strategy is, we should also be able to say what a data strategy is not. For example, going to Cloud, adopting AI, and hiring data scientists are not data strategies; they are the means and tools I see that are often confused with vision or strategy. This confusion could either lead to costly efforts of “boiling the ocean” or doomed failures because of the lack of clear directions and goals in the first place. Technologies are advancing at an unprecedented speed, particularly in data and AI areas. Successfully adopting a new technology is important and meaningful only when it can help realize a business vision that the preexisting one cannot. Therefore, the ultimate business value a new technology can bring, relative to that of the existing system, is what justifies exploring the new technology and making it part of the strategy. On the other hand, a new technology could propel new ideas for new products or new revenue-generating opportunities. When this is the case, the new product or revenue opportunities need to be part of the business vision and business strategy, while the technology becomes an investment in the data strategy. In other words, the exploration and adoption of new technologies should be aligned with business objectives and business vision. 

.

There are 6 areas that we should focus on to create a successful data strategy: Alignment, Architecture, Process, Organization, People, and Long-term planning.

.

.

This is probably the most important aspect of data strategy. While the importance of data strategy is widely accepted, its implementation is still a big question for many organizations. This is mainly because often, business people are not confident in the technology and data scientists or technologists are  too focused on technology without contextualizing technological tools in the face of business challenges. Alignment can happen, therefore, only when both sides focus on business results:

.

Business objectives and strategies should focus around the results that need to be achieved for the customers and clients, and not restrict potential technical solutions. For example, when a business strategy says “deliver API to the company’s clients,” it immediately narrows the scope for the technologists to build API solutions for the end clients. While it is likely one of the possible solutions, it does not truly reflect what clients are looking for. It is more appropriate to state the problem differently as “clients want a convenient way of retrieving data from anywhere they want with minimal human effort.” With this objective, the solution would be very different and in fact, should lead to a completely different strategy which will emphasize more on user experience instead of API only. The lesson we have learned is that business people should state business problems by focusing on the desired business results, and avoid defining which technologies to use from the get-go.

.

Technologists always look for new tools, new libraries, and even new programming languages. From the moment a new technology is implemented, it can gradually start to lose its glory and attract criticism due to its limitations compared to even newer technologies. So a serious pitfall I have seen over many years is that implementations or updates are done only for the sake of technology advances, without solving business problems. Sometimes, this is necessary and unavoidable. On the other hand, we should try to avoid following the hype in the technology world, and instead focus technology advances on business use cases and make wise decisions based on business needs.

On a higher level, business needs for data strategy come from 3 areas: Products, Operational/Process Improvements, and Compliance. The data strategy for each of these areas can be very different. Products-related strategy requires capabilities of fast discovery and integration with new data sources and delivery of new user experiences. Operational/Process improvements are applied most internally and focus on efficiency, cost reduction, and quality improvement. Compliance focuses on data security. Effective data strategy, therefore, should have all these areas covered.

.

In the areas of AI and analytics where there are many opportunities, data strategy should align with business strategy in order to set priorities and determine which areas to start in and what use cases to plan for next. Clearly defining the priority of AI and machine learning (ML) is not an easy task, which requires a deep understanding of the methodologies, opportunities, and bold vision in the first place. In this case, business and product people should work closely with technology and data science leaders to create visions and strategies together.

.

.

When working on data architecture, understanding the existing and future data assets is key. Important questions need to be raised, such as, what type of data is needed to solve business problems, what are the typical use cases, and what’s the volume growth in 3 to 5 years. As I stated in the other article (Principles for Big Data Performance), working with small data volume is more straightforward, which warrants faster cycles and requires less investment. When working with large data volumes, however, the architecture will become much more complex, hence, the cost will be much higher and implementation will take a longer time with more risks. If data sizes are indeed different for different sources and applications, there should be different strategies for addressing each case, with different ways of thinking and planning from the beginning. Given the complexity of big data architecture propelled by the continuous evolution of technologies and increase in data volume, the strategy on big data architecture should be also far-reaching, with the vision likely not only limited to 3-5 years but also on the look-out for an even longer term.

.

A typical architecture for big data consists of 3 areas: 1) Data Sources and their staging area; 2) Data Lake and Data Processing Platform; 3) Delivery solutions. A successful data strategy for big data should plan differently for these 3 areas, and align each with business objectives accordingly. In particular, rebuilding the Data Lake and Data Processing platform for Big Data is very costly and can take a long time and many resources. Adding a data source or developing a new delivery solution, on the other hand, can be done relatively quickly and easily to increase revenue. Building a scalable, integrable, and sustainable data processing pipeline and the platform is a core strategy to enable the long-term growth of an enterprise.

.

.

It happens often that new ideas and innovation start from business departments because they are closer to customers and clients, while IT or centralized data technology groups have the expertise in change management, scalable production systems, process automation, and enterprise-level solutions. Having “decentralized” IT in the business departments is also called “Shadow IT”. The reason for its existence is clearly evident: IT has centralized resources that often cannot meet the business requirements of being fast and nimble to satisfy customers’ needs. On the other hand, the business department neither has the capability for a large production environment nor the budget for enterprise-level software solutions. This limitation could eventually lead to “growth pains” and wasted resources when the solution needs to be re-engineered and migration needs to happen.

.

This conflict becomes even more obvious in Data Science and AI fields. While small data science teams in business departments can quickly experiment with AI and machine learning algorithms, they should also plan out how to scale and productionalize their models at the same time. The whole life cycle, starting from experimentation of new algorithms through integration and productionalization,  need to be explicitly discussed and enforced in the data strategy, with the organization structure and partnership between business and technology departments incorporated. This exercise could lead to enterprise-level efficiency and faster innovations at the same time. For example, data scientists can focus on new ideas, but do not need to worry about supporting production, while IT/centralized Technology group focus on building and supporting the scalable integrated production platform with optimal performance. In other words, a good strategy allows us to look at the data life cycle as a whole, from the initial innovation through the eventual productionalization and business growth. It also prompts technology and business departments to work closely with the same objectives to achieve win-win outcomes.

.

The subject of any data strategy is the data assets for a company. A successful data strategy should have a clear plan on how to unleash the value of the data assets to serve business purposes. Data governance is the process of managing these datasets in terms of ownership, integrity, compliance, quality, content, and relationship with other datasets. Without established governance, an enterprise would lack clarity and insight into its datasets, which could result in inconsistent or overlapping data strategies. As stated in a previous article of mime (Master Data Management: an Essential Part of Data Strategy), to ensure Master Data Management as part of the data strategy, a data governance process is essential for the success of the strategy. Data governance is a process, the effectiveness of which is measured by the results it delivers. Its purpose is to provide transparency and efficiency, but not confusion and bureaucracy. Given its important function in managing the data assets, data governance should be part of data strategy, with a more concrete plan on how to establish, enforce, and continuously improve the data governance process for an enterprise.

.

.

People should be also part of any strategy for an enterprise, and this is no exception for data strategy. The difference for data strategy is that technical talent is hard to find and data also requires institutional knowledge to be productive. The assessment of the current talents and plan ahead of training or building the pipeline of new talents are extremely critical for a successful data strategy. In this space, talent retaining and finding new talents are equally important. Building teams with diverse and complementary skill sets and multiple levels of experience should be an important aspect of a successful data strategy. 

.

Another aspect is building the data competency and data culture in the organization. This is not to say that anyone accessing the data should be able to do SQL, Python programming, or run machine learning algorithms. One example is that most people are capable of working with data in an Excel spreadsheet. The same level of reach and competency should be established for people to access the data easily using a standardized tool within the organization, such that people can pull, analyze, and use the data to improve their decision making and work efficiency without the help of technical people. This certainly depends on several capabilities within the company: easy data access, clear data cataloging, availability of self-service tools with the plug-ins of accessing and analyzing the data, and a lot of training. Establishing data competency and strong data culture requires long-term efforts, but it is an important part of the data strategy for an organization to fully discover and realize the value of its datasets for business growth.

.

.

Data are the assets of a company, hence, working with data is a long-term effort, and data strategy requires long-term vision. In fact, a successful data strategy should have an over-arching plan that could surpass business strategy to enable a long-term foundation for future business strategies and growth. This is because establishing data platform, infrastructure, talent, and culture requires time, and the results become the company’s assets along with the data. Working with data assets is very different from building programs or interfaces. It requires careful management, long-term investment, and persistent effort. The stability with a solid foundation and capability is essential for a company to grow efficiently and quickly with minimal interruption and setbacks. While we do not know exactly what the technology landscape will turn out to be, a successful data strategy will help the enterprise prepare and embrace it instead of reacting to it with stumbles. In this aspect, the strength of an effective data strategy emphasizes the value provided by the data assets and the people, but not new technologies, which are simply the tools and approaches to extract and realize the value of the data.

.

.

While data strategy has the same components as business strategy, it is still relatively new in the strategy world. Setting up a right data strategy requires first a business vision and the alignment with a business strategy. Data strategy should also be considered as an enabler of the long-term vision, setting the cornerstone that future business strategies can rely on. Lastly, preventing a data strategy from being purely technological or IT-focused is the key to its success. In other words, an organization should take a holistic approach to adopt a long-term data strategy, with optimized technological investments, people, and processes, to enable continuous business growth. After all, as Jim Rohn puts it: “Success is 20% skills and 80% strategy. You might know how to succeed, but more importantly, what’s your plan to succeed?”.