By Pranay Agarwal and Betsy Romeri.
“They just don’t understand me. I provide them with all the necessary details, all the data and analysis possible, but they don’t seem to get it.” These were the comments of a data scientist at one of the largest CPG companies in Brazil as he vented to me over coffee and described how frustrating his last year had been. He definitely was feeling the heat from the business stakeholders. None of the major analytics initiatives that his team had worked on had really taken off, and a lot of resources and time had been spent.
We often hear about companies investing millions of dollars in hiring the best data- science team and still not receiving the expected ROI. Why is that? The complexity is often due to lack of effective communication between the data-scientist team and the business stakeholders.
The communication problem starts in the inception stage. Proper expectations are not set. Business stakeholders often communicate very vaguely and fail to provide clear and concise goals for their projects. They are savvy business executives, but often underestimate the complexities that reside in their data.
On the other hand, data scientists are typically trained to solve challenging business problems using data, but may not be skilled communicators; thus, they often are unable to present their process or findings in a way that makes the connections with specific business metrics and goals.
Based on my experience interacting with both data scientists and business executives, I’ve compiled a short list of guidelines for both sides to follow to improve communications and, importantly, the success of their joint initiatives.
For the Business Stakeholder
Qualify the Problem: Business stakeholders must be very clear about the specific use-case they are trying to address. They need to involve the relevant stakeholders (from C-level to end users of the insights and prescribed actions) who understand the problem in depth and engage in a very collaborative manner with the data-science team to distill the problem down to specific questions to be answered. For example, what data are available to be used and what are the targeted performance parameters for the use-case? Significant and relevant questions (see sample questions below) should be discussed to achieve full alignment on a use-case.
- What is the single objective question the analytical model needs to answer?
- How does answering the question impact the business? What processes and areas will be impacted?
- How has the business been addressing this question up until now? Why is the current method(s) not sufficient?
- What level of “improvement” is required and what is the required ROI?
- What data are available to answer the question?
- Which business decisions are directly impacted by the calculated answer?
- How often are the targeted business decisions made?
- In which format(s) must the insights and/or prescribed actions be provided to the decision-makers?
Answering the questions above will help to define and qualify the use-case(s) in a very concise manner, allowing for (1) a better understanding of the resulting insights, predictions, or recommended actions and (2) a more direct connection of the results to the business decisions and operations involved. In addition, the work of the data- science team will be accepted and adopted more readily when they can confidently present their work as precise and comprehensive regarding the given question.
For Data Scientists
Listen, Ask Questions, Frame the Project Scope, Define Measurables:Business executives often throw many use-cases against the wall and wait to see what sticks. They don’t have the data-science background to know which use-cases will benefit the most from the application of data analytics, nor do they understand which cases will be “quick wins” versus long-term projects. This is where the data scientist can help with good communication skills.
Analyst teams should spend the time to understand each use-case suggested by the business executives. Learn the nuances of each problem they are experiencing, including any assumptions they are making regarding why they have the problem, how they have tried to solve it in the past, and what data they have been using. This will help (1) frame the use-case to be addressed with data and analytics and (2) define the measurable parameters that can be followed to determine accuracy, success, and—importantly—when the projects is DONE.
A common problem I have observed in my experience working with both business executives and data-science teams is that they are not always in agreement when a given project is done. Defining “done” at the beginning of the project is extremely important.
I suggest developing a detailed roadmap with a timeline of expected, achievable milestones and their benefits, starting with potential “quick wins” that can be achieved early on to improve the chances of “buy-in” by stakeholders. Quick wins also build the credibility for your team that might help when the longer and more complex projects take longer than expected.
Finally, a Scope of Work (SOW) should be written to document and discuss variables such as location and integration of multiple data sources, integration of solutions with legacy systems, security and governance requirements, as well as how the results can be accessed and utilized. The SOW is a great opportunity to communicate every aspect of your proposed work on a given use-case. A well-written SOW can provide the roadmap as well as align all stakeholders on roles, responsibilities, and expected outcomes.
Provide Transparency: Once the use-case is defined and all stakeholders are aligned, there is another level of communication required of the data-science team. Business executives expect to see tangible results and see reports on the progress of any initiative. They prefer to be fully informed and understand the milestones of any project and how the results apply to the business.
With a transparent approach, companies are able to build a culture where the data- science team and business executives collaborate. Methodologies such as scrum with frequent releases of progress reports, including setbacks, along the way, allow for a high degree of collaboration and obtaining buy-in from executives at key points.
With this level of communication, executives feel they are actively contributing to the overall success of the project and are less resistant to changes, compared to being involved only in the boardroom, looking at summaries of the results in the form of graphs and charts after months of silence and no idea how they were arrived at.
My best advice on transparency for the analytics lead is to consider providing a very detailed project plan with key milestones and deliverables and frequent checkpoints with the project committee—before the project begins.
Conclusion
Data scientists are in great demand and at the forefront of the next generation of business intelligence using machine learning and AI. Aspiring data scientists should consider honing skills in business and developing skills in communicating what a successful analytics project looks like and how it can be achieved. On their side, business stakeholders should strive to get acquainted with the language and capabilities of data science to hone their ability to identify a viable and valuable (and potentially quick win) use-case for their business. Taking the time to understand and communicate saves time, money, and frustration in the very complicated and fast-moving world of advanced analytics and AI. Great communication between business executives and data-science teams also benefits the sales team’s success in getting projects sold!
