Solutions

One way organizations can change this narrative around their analytics skillset is to create a Citizen Data Science Program (CDSP). This program fits in the middle, between the business analysts and the data scientists, and follows the concept of a ‘guild,’ which is a virtual community of the data teams’ capabilities that stretches across multiple business units. (Kalmane, 2022). This guild provides the tools, technology, and cultural adoption for upskilling existing employees to produce comparable results as a regular data scientist.

Implementing a CDSP involves change across the organization, with ramifications for:

Talent acquisition.
Training.
Promotion and career advancement.
Knowledge management.
Budgeting.

Fundamentally, a citizen data science program changes the way of working as well as the roles and responsibilities of members, both for the LOB analytics functions and the central data team. Each organization must define what the CDSP will look like to suit their needs, but here are examples of the supporting changes and capabilities they commonly use to enable and support a CDSP:

Get the Right People: To source external candidates, you will need to update your HR criteria for hiring proficiency in business and data skills. Hiring managers should be trained to identify business analysts that show potential to be high-performing citizen data scientists (CDS). There might be many skilled information analysts in your own organization who can take on additional tasks and be part of the CDSP. By getting the right people, you will have analysts with the domain knowledge to provide input on the complete deliverable along the data pipeline.

Organizational Training:

Stand up a lightweight upskilling program that LOB analysts can take advantage of.
Build out professional development plans at scale for each LOB analyst across the enterprise and provide visibility into their development over time across the organization (involves collaborating with HR and with various LOBs).
Extend data literacy education to the peers of CDS in their respective LOBs, so they have a better understanding of the results and insights generated.

Team Collaboration:

Enable cross-collaboration between LOB citizen data scientists and the central data team.
Provide a way for LOB analysts’ insights to make their way back into the central data team.
Share best practices among LOB citizen data scientists across LOBs and generate a feedback system for improvements.

Data Processing: By implementing such a CDSP, your LOB analysts will be capable of analyzing their existing data sources and skilled in handling and manipulating slightly more ambiguous types of data like unstructured, streaming, and graph data. You gain the opportunity to add new data sources and remove old ones to the data modeling pipeline that might work better for your needs. Breaking down your data silos will benefit this effort as it will allow for better data integration. The LOB analysts will also be able to curate more meaningful datasets within these sources, since they could exploit their business acumen to select the most appropriate data relevant to the business domain they are already experts in.

Technology/Tools: You can infuse these capabilities into your regular business analysts by using tools that involve the full span of the data science workflow, from data sourcing, preparation, analysis, and modeling to model training, validation, and even deployment—fully automated. With tools and platforms readily available today, such as Knime, Dataiku, DataRobot and Alteryx (Scalefresh, 2020) these citizen data scientists, many with no or minimal coding experience, can now use drag-and-drop software canvases to create and derive insights from data models. Using such tools can make it much easier for your LOB analysts to manipulate data compared to using an Excel spreadsheet. (Boulton, 2018)

Roles and Responsibilities: Clearly articulate the level of the roles and responsibilities for all involved (technical, management, executive). The overall support for the analytical ecosystem will probably become more complex when adding a CDSP. But you can reduce the confusion and foster collaboration by having everyone understand both the exact boundaries of their tasks and whom they’ll be working with.

Data Governance: Review your data governance policies and make any changes around items such as the scope of understanding, which illustrates what model the data teams are building, what the model should do, and for whom. This will also outline what data you should be using and how to use it.

Trust and Transparency: Creating standards for transparency about how data and models are handled is critical to building trust in the results. These standards are the processes for protecting the organization’s users and its associated datasets. They include the policies for security of the model and algorithm storage as well as those who can access the data for testing and training. You should also have policies that address the user’s privacy and the company’s ethics standards, which should include definitions for strong explainability of your analytics and how you conduct bias detection.