The majority of data science projects fail to reach production (something like 20 percent of data science projects make production). Data science adoption is crucial to the success of an organization’s investment in data science. Thus, this 80 percent failure rate often is the result, not of the data science capabilities, but of the lack of the business’s willingness to integrate the data science solutions. Lack of infrastructure can play a role, but this is something that easily is remedied. Here we will discuss 3 core components of data science adoption – education, partnership, and trust.
Convincing the business of the mutual benefits that a data science solution provides is imperative to data science adoption occurring. A lack of adoption means a lack of impact, and the unfortunate response often is categorized as a failure of the data science effort. A number of challenges come with building a strong data science team, and the adoption element of the data science process tends to get overlooked. This should not be a surprise since many projects across many disciplines tend to face struggles on their way to production.
Education is Foundational to Data Science Adoption
The first component of data science adoption is education. There is not a universal understanding of what data science is. Many organizations lack a definition of data science in their business. Because of this there are often misaligned expectations of data science capabilities. Educating the business, at all levels, offers a strong foundation to generate proper understanding and enthusiasm for the impacts data science will deliver. Investing time engaging in recurring roadshows will offer a platform for proper data science questions to be discovered, scoped, and acted upon.
An area of focus in educating the organization needs to be the detailing of common steps in the data science project. Data science does not follow the expected project path that other disciplines, such as software development and engineering, may follow. The “science” in data science means discovery and experimentation. It is a bit of an adventure, and adventure, by nature, does not always come with a clear path to the goal.
Data science projects commonly are embarked upon without clarity regarding what the end result will look like. A staple of data science projects is the iteration needed to work through the problem. The data science process is summarized into the business question, the data science question, the data, the model, the evaluation, the adoption, and the monitoring. At any step, a discovery may be made that causes iteration on a prior step. Organizations can feel uneasy with the ambiguous nature of data science projects. Providing education on the process can set expectations and give leverage to the data science team to maintain their focus.
Strong Partnerships and Data Science Adoption
Partnership is the second component in generating data science adoption. of data science outputs. Partnership is the formation of relationships with business stakeholders that deliver growth to all parties involved. The partnership results in an expansion of the pie, not an expansion of an individual slice. Partnership is achieved through persuading the business regarding the benefits that data science brings to the business. It is formed when there is a shared set of goals and can be bolstered when there is clear executive buy-in to the effort.
This can be challenging to develop since strong partnerships tend to develop when there is a history of interaction between parties. Fostering an environment that is rich for the formulation of a strong partnership requires that the data science team is diligent in their interactions with those they are seeking partnership with. Diligence here means offering to take some of the load, where possible, from the others. This action displays the commitment the data science team has to succeeding and the passion being brought to the project.
Another means to a strong partnership is through the development of small prototypes intended to clearly reflect the potential benefit to be gained and entice greater input from the partners. Offering them a sample of the data science capabilities influences their appetite for the possible solutions that may come their way should the data science team be successful.
Trust from the Business Means Accelerated Data Science Adoption
The last core component of data science adoption is trust. Trust is essential in any relationship and it is no different here. The data science team is building relationships with the business that will allow for all groups to succeed. To build these relationships there must be confidence that each participant can rely on the others to keep their interests in mind. A new group entering the fold, in this case data science, means this group must establish their capabilities and the positive impacts they can have on the others.
Trust is best developed through gradual and incremental engagements. Taking on smaller projects that fit into current processes is useful for getting one’s foot in the door to greater trust. Organizations tend to want data science to step into the business and return immediate transformational outcomes. Such outcomes are rare without trust.
Transformational outcomes often require more than a simple algorithm. They require changes to processes and changes to people’s behavior. They are disruptive to the status quo and people tend to get uncomfortable when it comes to changing the status quo. Incremental projects that enhance current processes are less disruptive and easier for people to grasp since they fit into what is comfortable. Impacting the more incremental engagements will form the building blocks to the truly transformational outcomes.
The business will begin feeling comfortable working with data science through these smaller incremental engagements. Any errors that may occur will be more tolerable since the impacts are less disruptive to the larger systems. This can be a sort of playground to experiment with the relationships needed for transformational change. The challenge that occurs is how the leadership responds to the time needed to develop this trust. It is the data science team’s responsibility to inform the leadership of the need for environments that foster trust with the business. It is trust, layered with education and partnership, that allows adoption of data science outcomes to occur. Without adoption, there is no impact from data science.