How Do You Build A Data And Analytics Organization?

Business models are innovating towards an increased reliance on Data and Analytics for competitive advantage. The growth D&A is driving requires an organizational approach. Without it, managing the complex landscape of platforms, talent, and projects becomes a barrier to progress.

Vin Vashishta | Originally Published: May 23rd, 2021

Book time with me for Career Coaching or sign up for my Business Strategy for Data Scientists class.

The Data Science Team is scaling in many businesses. When I started building out teams six years ago, I was focused on the core capability which was sufficient for early Data Science business needs. Teams were three to five people.

About three years ago, that changed as the number of roles rapidly expanded and the reliance on Data Science for efficiency and revenue growth increased.

When I examined all the pieces being assembled company wide, I quickly realized it is more effective to look at this as a Data and Analytics (D&A) Organization rather than a Data Science Team that interfaces with supporting pieces scattered throughout the business.

Step one in building out an organization is taking inventory of what D&A talent, infrastructure, and products are already in place. That leads to consolidation of ongoing projects under a single umbrella. Most of the problem statements come from both efforts:

  • What projects should the business be taking on?
  • What talent does the business need to execute on those projects?
  • What infrastructure does the business need to execute and support those projects?
  • What is the ROI for existing, both internal efficiency and external customer, data products?

  • These problem statements define a DandA organization. How does the business go from problems to functional organization?

    The Data Science Product Roadmap

    In the first round of Data Science Team ramp up, step one was probably hiring a Data Scientist. The second round often starts with hiring a senior leader for the growing team. An organization is a strategic unit, and the third round of ramp up is defined by the business model and strategy.

    What projects should the business be taking on? For Data Science to succeed, model metrics need to be connected to business metrics. Data Science Product Roadmap planning starts with examining the KPIs the business uses to measure growth and progress towards strategy goals. Project one is building an evidentiary support for KPIs. Essentially, are these the best measures for progress?

    From there, products are split in three lanes.

    Internal strategy products model the business systems that drive KPIs. These are decision support products that give the business clarity into the best strategy for achieving business goals.

    Intelligent automation products are internal efficiency focused. They reduce operational costs and improve time to outcomes.

    External products are high revenue drivers where the functionality could not exist without the Data Science component.

    The Data Science Value Stream

    Technical value streams are essential to planning both talent and infrastructure. Jumping straight from Product Roadmap to scaling creates a static organization. The midpoint is mapping value generating activities.

    The business has clarity on what projects are highest value. Now the question is, what activities create value? With that framing, you can hear the link between strategy and execution; the connection between “this is what we should be doing” and “this is how we do it.”

    The Data Science Value Stream is an abstraction of project specific execution steps. Instead of defining how the business will create value for a specific project or set of projects, it defines how the D&A Organization creates value. The key is to make activities value centric rather than project or technical centric.

    With the Value Stream defined, the D&A Organization can build out its workflow. The Data Science Workflow is split into three components.

    The Data Workflow covers the phases and key activities that produce high quality datasets that are ready to be used for research and model development.

    The Research Workflow covers the phases and key activities that produce novel models and unique datasets (research artifacts). This is where a business converts high quality data into sustainable competitive advantages. Artifacts are used across multiple projects and they are the highest value work products.

    The Model Development Workflow converts research artifacts into productionized models. This covers activities that closely mirror traditional Software Development Lifecycle activities.

    Defining Talent and Infrastructure Needs

    The workflow provides a framework for evaluating existing capabilities and defining capabilities gaps. The missing pieces are grouped into two solution buckets.

    Infrastructure can manage some of the Data Science Workflow. The traditional buy vs build decision happens at this step. Much of the Data Workflow and Model Development workflow can be automated. Automation in this case provides the capability to scale project complexity without scaling team resources. For some projects, parts of the workflow would not be practical without key infrastructure and automation components.

    Talent is required to perform the key activities for each phase of the Data Science Workflow. Defining the workflow makes the staffing plan more certain. Many hiring efforts fail because the business is hiring a generalist who can cover many potential business needs. The successful hiring strategy is to build certainty around work activities and hire specialists who are capable of generating the specific value needed.

    The Organizational Built Out Plan

    The original inventory of talent and infrastructure now needs a plan to scale from where the business is to where it needs to be. The Product Roadmap provides a timeline. The build out plan is a phased execution schedule to deliver just in time capabilities to execute on projects.

    The most important step is creating a talent pipeline, similar to a supply chain. Most companies have access to junior and mid-level talent. However, advanced capabilities are out of their reach because they lack access to highly skilled knowledge workers.

    D&A talent is not difficult to source when the business understands its specific capability needs. Hiring must be based on work products and outcomes with a well understood infrastructure stack. Highly skilled knowledge workers are looking for roles where the projects and workflow are well defined. They gravitate towards master planned organizations with infrastructure to support advanced projects.

    When a hiring manager can have detailed conversations about the role today and how it will grow in the next two to three years, high end talent sees an employer of choice. With most businesses operating at a low level of D&A maturity, the businesses with clear strategy stand out.

    Even with all this planning, Data and Analytics Leadership roles are difficult to fill. The scarcity of talent can overwhelm the best of planning. Interim leadership is typically required to fill in the gap. This can come from internal promotions or external consultants.

    However, it is worth spending money on compensation to attract top leaders early in the planning phases. While most Data and Analytics leaders are qualified to take over an already built team, organizational leaders want a hand in building it from the earliest stages.

    Project Consolidation and Improvement

    The existing landscape of D&A projects, platforms, and products needs to be owned by the D&A Organization. This is the final stage of ramp up planning. The cost savings from this phase often pay for much of the ramp up.

    D&A talent is distributed through the business and they have been building solutions across the business. Many are functionally redundant. Levels of quality vary. Platforms overlap between organizations and niche business processes have led to elaborate customizations that are expansive to support.

    Much of what is required to align the goals of different organizations into a cohesive digital transformation strategy is accomplished by rolling up the D&A initiatives to be owned by a single D&A organization. Projects are owned from business case to deployment and support. Everything flows through the Data Science Value Stream and Workflow. Project timelines are brought in and the resulting products are delivered with consistency.

    The process is more easily managed after consolidation. Transformation is not stuck behind RPA that will need to be reengineered to Intelligent Automation. The D&A Organization can leapfrog the middle step accelerating timelines and providing higher returns.

     

    As the scope of D&A’s influence on the business model increases, it makes sense to move from team thinking to an organizational approach. That breaks down many of the silos preventing projects from moving from business need to functional solution.