How to build a data-driven company - MIT Sloan News

komoros.blogspot.com

While transformations can be difficult, the value of embracing data is clear — which is why Cindi Howson, the chief data strategy officer at analytics platform provider ThoughtSpot, urges more companies to think about what’s stopping them from becoming data-driven.

Speaking in August at the MIT Chief Data Officer and Information Quality Symposium, Howson said data-driven companies enjoy increased revenue, improved customer service, best-in-class operating efficiencies, and improved profitability. “It sounds like what we all want and why we collect data at all,” she said. But according to a study from the Harvard Business Review, only 20% of companies are actually empowering frontline workers with data, Howson said. “That is an unacceptable situation for the state of the industry after 25 years. So, what does it take?”

There is no single path to becoming a data-driven company — some firms might focus on building the right data team, while others invest in the right technology or build analytics into their digital transformation strategy. Startups and newer companies have an advantage as they’re able to build data into the framework of their organization. Incumbents face challenges like entrenched business processes or reluctant leadership.

Creating a data culture is one of the keys to becoming a data-driven firm, according to Howson and other experts who spoke at this year’s symposium. In a survey over the last year of analytics leaders from around the world, Howson found that 61% said the culture — not technology or people — is the biggest barrier to becoming data-driven. During the pandemic, as companies all try to unlock the value of data faster, leaders continued to identify culture as the main block, she said.

Culture is usually set by company leadership, Howson said; grassroots culture changes bubbling up from lower-level employees are possible, but more difficult.

Here are six ways leaders can strengthen their company’s data culture, according to the experts.

1. Know what it means to be data-driven

Data is also no longer just a byproduct of transactional systems, said Andrea Gibbons, the deputy chief data officer at the Export-Import Bank of the United States. Applications and technology are now designed around what data is needed for a business to make decisions.

Gibbons said companies should organize their systems and applications to give the right people access to data so they can make decisions quickly, instead of having data “dumped somewhere and then massaged and arranged for better analytics to make a decision a week or a month or a year later.”

2. Embrace new technology

While the right technology is not an easy shortcut to a data culture, technology and culture are two sides of the same coin, Howson said, as both depend on being open to new ways of doing things.

Companies with outdated technology — reports and dashboards, on-premises enterprise data warehouses, siloed data marts, and email as the primary source of communication — often have a culture where data is hoarded, Howson said. And often there is a lack of leadership: resistance to change, complacency, an unwillingness to use technology created elsewhere (also known as “not-invented-here syndrome”), and fear of failure — all impediments to adopting a data culture.

Instead, companies should invest in a modern technology portfolio with AI-driven insights, data lakes, collaboration tools like Slack, cloud-based, logical data warehouses, and augmented analytics, Howson said. This often correlates to an innovative culture where data is democratized, there is transparency and trust, expertise and insights are shared, and people feel empowered and energized to embrace the new.

3. Disrupt your culture

Building a culture is tough in part because it requires soft skills, Howson said. She outlined five concrete ways leaders can disrupt their workplace:

Bring in a change agent. This can mean hiring someone like a chief digital officer, who are often “serial innovators,” Howson said, coming in to a company and shaking things up. This is a role with a high churn rate, she warned, as change agents tend to achieve success and then move on to the next challenge.
Incentivize innovation. Companies should ask themselves how they approach innovation, making sure employees are not punished for trying new things. Successful firms embrace hackathons or invite teams to tackle hard problems and even celebrate failures.
Confront brutal facts. How does your company culture respond when the numbers paint a bad picture? Successful companies avoid “vanity metrics” and hiding data that reflects poorly on their performance, Howson said, and instead address issues directly.
Identify “what’s in it for me.” This means flipping the equation from a focus on the data to a focus on what the company needs — from “what data do I need?” to “what problems can I solve using data?”
Organize for collaboration. Achieving results requires a collaborative approach, Howson said, which means that some companies will need to reorganize around data and analytics. This might mean combining business people with technical employees and coders. While there isn’t a single best organizational model, the most successful firms have data and analytics embedded inside every business unit with some degree of centralization, Howson said.

4. Make your organization’s data FAIR: findable, accessible, interoperable, reusable

“Can people even find the data?” This fundamental question often addresses many underlying problems, according to Milind Kamkolkar, chief data officer at therapeutics firm Cellarity. This is especially a problem at organizations with business silos.

One way to look at the relationship between employees and data is FAIR, guiding principles for data created by a team of researchers in Europe. In applying those principles to the workplace, Kamkolkar said, firms should make sure data is:

Findable: Findable data is often a main business problem and focus for CDOs, Kamkolkar said, especially as silos lead to data repetition and lack of coordination. Security and data privacy are also dependent upon findable data.
Accessible: This relates to security roles, responsibilities, and ownership, and ensuring that data professionals are tasked with focusing on accessibility. In highly regulated businesses, it is important to ensure access is happening at the right discretionary level based on rules and governing principles.
Interoperable: Interoperability can be difficult, as it requires integration, but it presents an opportunity for firms to create relevance across data sources, Kamkolkar said. If done right, it allows you to develop your data initiatives under a semantic model that will create an interoperable language with your business as well.
Reusable: Once a company has generated great data, it is important to ensure it can be used to its full potential value, Kamkolkar said.

5. Build data literacy

To build a data culture, everyone on the team needs to be speaking the same language. This can be difficult in a field where language is often imprecise — the terms machine learning and artificial intelligence are often used interchangeably, the experts noted. But data literacy allows everyone in the company to speak in specific terms, instead of generics, about how to use data.

It can be a hard sell, and some are offended by the term data literacy, Gibbons said — data acumen is another way to phrase it — but she believes it’s part of the change that needs to happen for an organization to truly be data-driven.

6. Don’t look at data as a separate part of the business

In order to build trust across different business units and adopt a data culture, companies should reframe how employees view data. “We’re basically integrating our data organization within the business function,” said Asif Syed, vice president of data strategy at The Hartford Steam Boiler Inspection and Insurance Company. “Our data office is part of the business organization and any business problem we try to solve starts with a line of questioning focused on what kind of analysis or data is necessary to solve this business problem.”

Kamkolkar said he takes a similar approach. “I would argue being a data-driven business is simply [good] business; the information assets we talk about are critical,” he said.