Bill Sequeira, PhD
Data Strategy and Transformation Leader and CEO, Alida, Inc.
A former CTO and COO, Bill Sequeira is an experienced digital and tech executive who helps companies adopt advanced data science methods and rapidly transform their data into a strategically valuable asset.
Companies across industries are realizing the potential value of their data, but unlocking that value is no simple task. In our latest Expert Q&A, Bill talked to us about how business leaders can craft an effective data strategy, tackle the complexities of a large data transformation, foster a culture that values data, and, ultimately, turn data into a core element of the business and a valuable asset.
When companies get it right, what kind of value can they unlock from their data? Can you share an example you have seen of an organization that figured out how to extract extraordinary value from their own data?
There is much value to unlock. Data can be used to create a 360º customer view, allowing companies to:
- Extract more revenue, provide better service, improve loyalty, and identify critical patterns in their market.
- Understand relationships between key players and entities to anticipate patterns of use and identify where costs can be better managed or where latent but unrealized revenue can be extracted.
- Provide B2B or B2B2C data services with useful and timely data that businesses or consumers use for improved decisions.
- Secure the knowledge of key customers and enable new services that lead to an increase in revenue.
- Increase enterprise value by harnessing strategic market sector customer data.
That’s just the start, and often several value dimensions can be realized at once.
To illustrate, a healthcare company was facing challenges with population health management in its core business. With some help, the company was able to mine the healthcare data it had to create add-on solutions focused on newly targeted clinical subpopulations. By repurposing its own data and staff, the company was able to deploy new services to existing customers in as little as six months. Over time, the company developed a reputation as the most innovative company in its market, client retention rates went up, and new products were introduced.
Turning data from an afterthought into a core element of the business can be an overwhelming task. What are the keys to overcoming the paralysis that sometimes happens when contemplating a data/digital transformation?
Paralysis often comes from an overwhelming sense of not understanding the breadth, depth, and cost of change required to achieve the desired value, and from the multitude of interrelationships necessary to make it work.
The first pitfall is not knowing what you don’t know. Sometimes you have the intuition that using data to generate value is possible, but how to get there is unclear. The second is fragmentation. Multiple organizations and teams. Multiple technologies. Perhaps multiple acquisitions and legacy systems.
To illustrate, imagine five business areas. Now introduce core data as a sixth business area. It will impact the prior five. Core data will change the relationships between the prior five areas (5×4), plus create five new relationships, for a total of 25 relationships and processes that will change.
And because data transformation is multidisciplinary, a wide range of functions such as business development, marketing, venture investment, IT, data science, technology, infrastructure, cloud, AI, mathematics, analytics, architecture, systems analysis, customer engagement, et al. will also converge.
It’s hard for companies to tackle such complexity without help. So where to start?
Excellent planning is key. An external expert can help you avoid paralysis by creating transformation maps that outline your current state vs. future state, internal business relationships, change areas, bottlenecks to overcome, and the investment to be done. They can also help define the process required to maintain a high level of service while the core of the organization is being transformed, as well as the gradual steps to be taken forward to manage disruption.
What are key elements of a data strategy? What kinds of things should business leaders keep in mind when crafting one?
Not a simple question to answer as organizations vary in size and strategic objectives. But I would say these are some useful tenets when crafting a sound data strategy:
- Understand where the market is heading to ensure data assets will support the organization’s market position.
- Identify the strategic objects where data is key to achieve success.
- Create data-specific objectives for your organization and pinpoint where true data value resides.
- Measure data ROI — measure data value, data-related objectives, and success.
- Identify a before and after of the data strategy to visualize end goals and their contribution to your success.
- Identify reasonable steps to go between the before and after.
- Identify the data producers and data consumers in your organization, internal and external.
- Assess the impact of change your data strategy will have in your staff, and the need for new skills, knowledge and experience.
At a technical level, ensure that a sound data architecture is established and a good set of metrics is created. Also verify that data technologies, tools, and infrastructure are within reach and budget, and identify any external partners that are needed.
Lastly, a data transformation is a very complex process. Make sure that you have experts at hand that can support you and your organization throughout the transformation.
What are the biggest obstacles to the successful implementation of a data initiative? How can business leaders anticipate and overcome them?
Some common data transformation obstacles are:
- Failure to secure transformational skills, knowledge, and experience to help craft, plan, and execute on a data transformation. The proverbial not knowing what you don’t know. Most organizations do not possess the skills required to implement a transformation on their own or to operate post-transformation.
- An organization that is reluctant to change its culture, decision-making style, power base, accountability, project-management style — all can derail much needed transformations. We’ve seen executives hampered by a company culture that does not want to move forward.
- Insufficient diligence to identify, understand, and set objectives, establish realistic timelines, and measure progress and success due to a lack of knowledge.
- Overemphasizing the technical aspects of a transformation but underestimating the difficulties associated with process (and consequent culture) transformation when transitioning into new ways of doing business.
- Underestimating the cost (time and money) of transformation, and in particular new staff costs (new hires with new skills and expertise), loss of staff, delays introduced when deploying new processes, the staff’s learning curve related to new processes, etc.
In general, the key to a successful transformation resides on identifying expert knowledge to provide guidance throughout the transformation process.
A data transformation typically affects many — if not all — employees in some respect. How can leaders convince people to change old habits and foster a culture that manages data in a cohesive way?
There are two approaches that can help the process — one technical and the other administrative.
The fastest way to affect technical change is to introduce new useful tools with related processes that staff can start operating right away. If accompanied by clear purpose, direction, and the outlining of results, the transition is easier. Habit and culture change are the hardest to accomplish, and it can’t happen in a vacuum.
The administrative side can help too, by including explicit transformation goals for staff associated with performance reviews, tied to performance targets, salaries, and bonuses.
An unfortunate but real consequence of a transformation is that some staff may no longer fit within the new paradigm. Therefore, be ready to rethink staffing as the transition evolves.
What role can AI play in ensuring the quality of data?
A big challenge for AI is good data. The data adage “garbage in, garbage out” plays a role in AI applications.
So what can AI do to increase the quality of data? Good data requires good interpretation, and the challenge is having the right context for a correct interpretation.
A past client was building a massive data pipeline to serve millions of customers and was concerned that 5–10% of the data presented to them had the potential for being incorrect due to ambiguous interpretation and selection. We built a deep metadata framework — ~240 dimensions — to drive intelligent data-driven processes that first identified critical data prone to ambiguity and then constructed useful “views” (context) to interpret and disambiguate. In such situations AI plays an active role — detecting data that is prone to ambiguity, detecting abnormal cases, isolating conditions and understanding patterns of use. This example goes beyond light-weight ML applications. AI can have a profound advantage if implemented correctly.
What can independent experts offer to this process?
Experts bring valuable knowledge, skills, and experience that can help throughout the process of transformation and achieve strategic objectives desired:
- Time to Solution — Experts know what to do and where to do it, and therefore can save much time in identifying the right steps to be followed.
- Cost — Experts are knowledgeable of the process and its pitfalls and experienced in mapping and driving transformation processes that will ultimately bring the overall cost down by preventing unnecessary mistakes, conflicts, and project failure.
- Team Transformation — Experts familiar with the organizational impact of data transformation can provide support and prepare the staff for change.
- Solution Quality — Multidisciplinary experts are effective at identifying relevant long-term issues and cross-boundary decisions. Organizations in the process of transforming seldom have the skills, knowledge, or expertise to do so. The result is a better integrated data transformation process.
We’re seeing more and more enterprises creating “data lakes” as real-time repositories of structured and unstructured data from all of their available sources. What are the most important things business leaders can do to ensure they will be able to leverage these data lakes for actionable insights?
Data lakes are useful in the right situations. But from a strategic viewpoint they solve a data processing problem but not the underlying business, strategic data problem.
IT organizations have struggled to centralize, standardize, and normalize data, with varying degrees of success. The most common concerns we hear from our clients are the very long time required to provide new solutions, the difficulty to integrate fragmented data (often the result of M&A activity), and the struggle when new insights are direly needed within a window of time.
But the very nature of data is to be distributed across organizations, to replicate, and grow. So how to reconcile data processing with distributed data, data integration, and data agility?
Seek data lakes that can provide a distributed, virtual view of data within your organization and can rapidly integrate new repositories into their standardized flow. Having the ability to accommodate disparate data repositories and integrating the data rapidly, on demand, is a must. M&A does not stop, neither does the market.
Also realize that while data lakes aggregate data, the power for strategically valuable insights comes from algorithms, not just the data. Many lose track of this important component. So when selecting a data lake, carefully review how the tool incorporates algorithms, how algorithms and algorithmic tools are managed and integrated, and how easily can algorithms be changed to adapt to new requirements.
How should companies approach data monetization opportunities? What are the risks to keep in mind?
It is important to first determine if you are monetizing a data-driven service, a product enabled by data, or data as an asset.
When building a new data asset, it is wise to follow a path typical of new business venture development — construct a business case, measure product acceptance, understand the product or service packaging, and identify viable adoption models, product positioning, etc. Do these right and you have a better chance of building a valuable asset and significantly de-risking the opportunity. And to maximize the opportunity of building data products or services, find effective ways to make data come alive. Algorithmic power brings out the value locked inside data, its meaning, value, and use.
Monetizing data also has an important legal dimension — ownership, use, liability, transactions, duplication, distribution, licenses. Underlying is the need for well-defined legal relationships between providers and consumers and the trust between them.
Another aspect of risk comes from your regulatory environment and your market, operational, and legal frameworks — compliance, privacy, governance, security, data breaches, and data leaks. As data use becomes widespread, the risks of cyber injury (the consequences of using online data against someone), cyber liability (exposure to liability through the use of data), and the use of “fake” data in data-driven systems will increase.
Careful planning and design goes a long way to manage risk during implementation and operation. The use of Blockchain can also alleviate some of the data-related risks. And the use of “smart” processes to actively monitor data transformation networks can help de-risk processes.
As an example, a company was looking to increase its enterprise value by repurposing its existing healthcare data to build an information services business to serve non-healthcare companies with marketing insights. The company incubated the effort with the help of outside expertise, reporting directly to its board. The two key de-risking steps taken were 1) fielding a marketing study as a proof of concept that insights generated were additive to existing marketing data and data models to lower product risk, and 2) by aggressively looking for customers and Letters of Intent (LOIs) to measure acceptance and lower market risk.