There appears to be a recurring pattern: large investments in agency-wide data platforms often struggle to secure user buy-in. A decade ago, the push to avoid "Big-Bang" deployment failures led to the rise of the Digital Services movement in the Federal Government. Today, as organizations shift toward data-centric decision-making and adopt new platform strategies, similar challenges are resurfacing. Big-Bang data platform deployments risk failing in adoption, mirroring the shortcomings of past large-scale systems. What lessons can we learn, and why are these Big-Bang approaches causing adoption problems? Below are are top 6 problems and recommendations to avoid them.
(1) Large Upfront Price Creates Unrealistic Instant Value Pressure
MOST IMPORTANT: Adoption is easier when you prove value and scale - not start with high costs and then try to prove value. Big upfront costs create a pressure for Big-Bang release in value - which is counter productive to adoption. The platform isn't proven - until it is proven to each individual. The pressure to shove all the things in the new expensive platform, amplifies early big mistakes increasing the likely hood of user rejection.
Example: An organization launches a new data platform promising brilliance. So much money and time goes in to configuration, building and setup - without consistent demonstrations of increasing value (Insights, Productivity Gains, Outcomes). Pressure builds and the large Big-Bang Data Platform becomes a four letter word in the Agency.
Solution: Scale the price and complexity of your platform along with the proven value. Deliver iteratively using Agile/Lean methodologies.
(2) Lack of Integration with Existing Tools and Workflows
Data professionals often rely on familiar tools such as Python, R, SQL, Jupyter Notebooks, and visualization software like Tableau or Power BI. If a new data platform doesn't seamlessly integrate with these tools, it can disrupt established workflows and decrease productivity.
Example: A data scientist accustomed to using scikit-learn in Python may avoid a platform that doesn't support Python integration, forcing them to learn a new programming language or framework.
Solution: A loosely coupled ecosystem platform that allows many tools and processes is better designed for tomorrow and will get more adoption today.
(3) Steep Learning Curve and Usability Challenges
Complex interfaces, proprietary technologies, or insufficient documentation can make new platforms difficult to learn. When the time investment to become proficient is high, data analysts and scientists may prefer to stick with familiar tools to meet project deadlines.
Example: A platform requiring mastery of a new query language without offering comprehensive tutorials may deter users who are under pressure to deliver results quickly.
Solution: Allow for adoption of new processes and skills needed for tool in to expectations of delivery. Buy or build platforms that are intuitive and don't require users to become platform experts - you will get faster adoption.
(4) Performance and Scalability Issues
Data platforms must efficiently handle large volumes of data and complex computations. If the platform is slow, unreliable, or cannot scale to meet growing data demands, it hampers the ability of data professionals to perform analyses effectively.
Example: An analyst experiences significant delays when running queries on big datasets, causing setbacks in providing timely insights to stakeholders.
Solution: Ultimately, users will adopt platforms that work well. However, the platform needs to make their work life easier not harder. Make sure the users define "working" and not the builders.
(5) Data Quality and Trust Concerns
Trust in data integrity is crucial for accurate analysis. Platforms lacking robust data governance, version control, or data lineage tracking can lead to skepticism about the accuracy and reliability of the data.
Example: Without clear data provenance, a data scientist cannot verify the source and transformations of the data, making it risky to use for building predictive models that inform critical business decisions.
Solution: Build transparency for the users into the platform.
(6) Insufficient Support, Training, and Organizational Buy-In
Adoption is hindered when there's a lack of training resources, user support, or encouragement from leadership. If the organization doesn't actively promote the platform or allocate time and resources for its integration, employees may not see the value in transitioning.
Example: An organization rolls out a new data platform but doesn't provide workshops or allocate time for teams to learn it, leading to frustration and low adoption rates.
Solution: Use Design practices like Service Design, Product Design, and Human Center Design to build with early adopters and have them encourage others. Make sure platform value is proven at each step via real outcomes helping the users.
By Greg Godbout from Flamelit
Comments