An enterprise guide to data and analytics

The potential consequences of failure can make organizations hesitant to embrace new and disruptive technologies. But this results in a slow and often overly cautious approach to adoption. This cautionary approach slows the benefits that can be derived from new technologies and approaches. Sandboxing is one way to test new ideas and prove out value without impacting other systems in operation.

It’s important to establish standards and procedures for data management, but also to identify key roles and responsibilities for data governance. This helps to ensure the quality and accuracy of data that is critical to the success of any data transformation project. A data governance framework should account for the entire lifecycle of data, from ingestion and transformation, to storage, retention and ultimately disposition.

Ensure protocols, procedures and solutions are in place to protect sensitive data and restrict access to authorized users—as well as comply with industry or government regulations to prevent hefty fines for non-compliance and mitigate risks of reputational damages. A data governance framework should ensure proper authentication, authorization, monitoring, and auditing of data.

There are a lot of data management tools and technologies on the market. When considering your options, cost will be a factor, but it shouldn’t be the main driver. Rather, equally important factors for a solution also includes whether it will be able to seamlessly integrate into your existing technology landscape, and the user experience a solution will be able to deliver to your users. The right fit for your organization helps accelerate adoption and helps you solve for the business problems you are building this solution around faster.

That being said, you’re not limited to just one platform. Historically organizations gravitated toward a single-platform approach. Recently, more organizations are taking a multi-solution approach and opting for two or three different platforms to meet the needs of stakeholders across the enterprise. When choosing multiple platforms, ensure you are accounting for whether these platforms can interact and create a seamless data experience.

To cope with large, growing, unwieldy datasets, many organizations are moving away from big and lengthy data transformation programs. Instead, start small, look for areas where you can quickly gain value, and solve for those first — ensuring that each iterative win builds toward your larger digital transformation goals. This sprint-based approach quickly delivers measurable results, keeps stakeholders engaged, and serve as cases for new and evolved data-informed and evidence-based ways of working across the organization.

Machine learning and artificial intelligence algorithms are continually ‘retrained’ on new data. Inherent bias can and often does exist in historical data. Instituting data governance that takes into account data integrity, “explainability”, fairness and resilience is important so that algorithms can learn in a way that doesn’t rebuild and propagate bias.

Don’t forget about the ‘last mile’ of analytics. Building a feedback loop process that feeds decisions made back into systems can help you understand what decisions were made as a result of insights and recommendations surfaced from data analytics. Closing the loop is imperative to the retraining and refinement of algorithms for accuracy and performance, or whether to retire an algorithm if business needs have shifted.

A data-driven culture values data literacy, analytical thinking and evidence-based decision making. Fostering a culture that balances innovation with risk management in tandem is important in driving a data-driven organization, which in turn is critical to the adoption of new and emerging technologies.