Data Analytics

How to Launch an Advanced Analytics Pilot

Here’s what to keep in mind as the Office of Personnel Management builds its enterprise analytics platform.

Peter Dunn is Federal CTO for CDW Government.

The Office of Personnel Management and other agencies planning advanced analytics pilots must partner with vendors to tailor solutions to their specific mission needs.

OPM revealed that it is building components of an enterprise data and analytics platform and piloting new data products in the data strategy it released in March, and most agencies lack the expertise to develop such a platform completely in-house.

Advanced analytics refers to the sophisticated techniques and tools used to interpret complex data sets. Traditionally, agencies starting pilots assemble tiger teams consisting of data scientists, domain experts and IT professionals to define required analytics, data engineering and government processes. That’s because only agencies fully understand their own priorities, and no one-size-fits-all commercial platform exists that can completely address their problems.

Click the banner below to learn how Backup as a Service boosts data protection.

The Best Analytics Pilots Have a Problem to Solve

The best advanced analytics pilots begin with a specific problem an agency needs solved, from logistics to operational movement issues. For instance, a Department of Defense pilot might focus on improving the movement of troops or equipment.

Conducting Big Data analytics with artificial intelligence will help the Army make better informed staffing decisions and implement a zero-trust security architecture, according to its digital transformation strategy.

The only way for the AI-enabled cybersecurity tools supporting the Army’s zero-trust framework to become efficient is for the machine learning programs behind them to make data available for training the models. For instance, a solution that monitors all logs and defines user behavior analytics to flag anomalous network activity requires months of work by data scientists to develop the patterns for it to learn effectively.

That work must be balanced against project budgets and deadlines, which will vary from pilot to pilot. The Army’s digital transformation efforts won’t align with those of other agencies because of differing mission priorities.

When all is said and done, the average pilot could take anywhere from 12 to 24 months to create.

DIVE DEEPER: Why data literacy is a core capability for federal agencies.

Challenges Agencies Face Launching Analytics Pilots

As OPM evolves its enterprise analytics platform to improve the experience of federal employees and customers, it will likely incorporate machine learning and AI, but it can’t simply buy ChatGPT off the shelf and put it to work. That’s due to concerns about the accuracy and availability of the data that generative AI runs on.

Generative AI wasn’t on most agencies’ radars a year ago, but users can find solutions online that design images, resumes, white papers and presentations. While the code behind those solutions might be right for an agency, each output comes with its own set of requirements to define.

Another challenge agencies face with AI is the fear surrounding its use, which takes educating leadership and users to overcome.

OPM planned a pair of data science hiring actions for 2023 to bolster its analytics workforce because it needs technical talent capable of setting pilot objectives and success metrics and maintaining operational security, despite needing private sector partners. Pilot priorities and expected results shouldn’t be skewed by a government contract.

The agency must further weigh whether it’s getting the right technology stacks, conduct a risk assessment, and apply data governance and compliance measures. External resources can make suggestions and assist with the roadmap and pilot, but they shouldn’t be in the driver’s seat steering toward a solution that won’t meet OPM’s requirements.

This article is part of FedTech’s CapITal blog series.