Like many buzzword topics across industries and disciplines, data science faces its similar challenges. It is not unusual for an organization’s leadership to blindly jump into the data science space based on the notion that they “have data”. Data by itself does not mean that they are ready for it to deliver the transformative impacts usually sought from expensive groups. Prior to jumping into the data science space an organization needs to assess the following – what are the short/mid/long term goals of it, what sorts of questions do we intend to apply it to and is the business ready to adopt/be challenged by the output of data science? The outcome of this will allow you to begin properly thinking about the readiness of your organization for data science.
Defining Goals for it
It is likely there are broad expectations for it in the organization. Often, these include notions of being the catalyst for data-driven decision making, uncovering some multiple of revenue gain or automating all of the decision making processes using artificial intelligence. These are grand, macro goals that serve a necessity but tend to lead to leadership down a misguided path.
Effective data science needs to be paired with a robust data strategy and data management framework. This fosters the creation of a data science strategic journey in which more concrete goals for it may be defined.
The core goals for it begin with establishing a data strategy and putting in place the resources to create robust data Management practices. data strategy will define how the organization intends to develop and transform their data into a resource that is more than a by-product of normal business operations. This strategy should include a focus on the generation, collection and storage of data for data science needs. It requires data in a format that is different than the formats used for storage and traditional reporting.
Data management builds on data strategy and implements the initiatives laid out in the data strategy. Here, data management will identify the needed infrastructure and systems that must be put in place to allow for data science to flourish. data management is the driver of the transformation of data from some by-product into a true resource for the organization.
Along with establishing a strong data strategy and foundation for data management, the goals of it now have a platform to stand on as the journey to organizational impact begins. In general, each data science goal should impact some element of increased revenue, cost reduction and customer satisfaction. Goals that can be linked to these three elements will have a direct ROI that can be tied to them. And presenting data science success in terms of ROI is a powerful tool to drive continued support from organization leadership. Ultimately, any goal that is set for it, needs to be linked to some business impact. And doing so in a language that is familiar to business leaders tends to make the bridge easier to cross.
Finding the Right Data Science Questions
Identifying the relevant data science question is crucial to ensuring that the deliverables from it have the intended impact and are positioned with the greatest likelihood of adoption. The defining of data science questions often begins with meeting with business stakeholders to define business objectives. From these objectives we can apply a sort of first principles approach to find what the true questions are that must be asked.
Setting a proper business objective is vital to the success of the it’s project. This can be a struggle due to the lack of clarity from the business regarding what is wanted of the project. A useful practice to derive a strong business objective is to ask the questions that seem the silliest.
This could be questions ranging from simply asking for a definition of some acronym to asking why a particular person has to review a certain document. These silly questions often will cut to the heart of problems and have answers that may have been known, but never were brought to light. The business is forced to think about the simplest steps. And in doing so, the objective of the question being asked will begin to appear. That question may transform into something different.
This leads to a key step in the data science process, which is the translation of the business objective into a specific data science question. One business objective can be translated into many questions of it. Take the business objective of “increasing close ratios.” The focus of this question is to understand what drives closing ratios for the sales team in making a sale. When translated to a data science question, we end up with many possibilities for us to pursue. Is the target an understanding of the probability of a close occurring? Do we want to know how much sales will change with this knowledge? Or maybe the goal is to improve lead scoring for the sales team to enhance the selection of customers they pursue?
Each of these questions are valid and each question likely would lead to an enhancement of knowledge. However, each question must follow its own path through its process. There may be transferable learnings gained at certain points in the path, but the goal of each question is distinctly different.
If the question is left too vague, then you risk not solving a problem that is useful for the business to understand, or you may not capture the necessary nuances of the business that may be required to have an impactful outcome. The knowledge gained through asking those silly questions during the defining of the business objectives becomes invaluable when it comes to translating the business objective to a question. The silly questions often result in a revelation of esoteric knowledge that is not widely known but can shape the questions data science is seeking to solve.
Adoption is Key to Success in Data Science
Adoption of outputs is likely the most important element in establishing impactful data science, advanced analytics and data management practices in the organization. All the foundational components may be in place, but without a willingness from the business to take up the outputs of advanced analytics, any hoped-for impacts are missed.
The notion of adoption gets lost in the excitement around artificial intelligence, machine learning, advanced analytics and the other buzzwords of the data space. A primary challenge in driving to successful adoption is working to solve the “soft” issues that exist in the business. In other words, these are obstacles that go beyond delivering an algorithm to production. There may be legacy policy in place that lead to the business users not willing to challenge their current practices.
Adoption requires trust, education and partnerships. A level of trust must be established and nurtured between the data science group and the organization. Trust is built over time. This is something that can be difficult to convince the leadership group of, specifically if they are confined to the notion that data science should be easy.
Education requires an interest from all parties involved. The data science team is not the only one doing the educating. A reciprocal environment is needed, where the data science team is educating the business regarding the discipline of data science, while the business is educating the data science team. Both groups gain a level of trust that is obtained through honest interaction.
Partnerships inherently require trust and education as well as a desire to see the other party involved in the partnership succeed. The data science team must be able to convince the business that growing a partnership with the data science group will lead to greater success for their part of the business.