5 Things Stakeholders Must Do Before Embarking on a Data Science Project

Businesses often jump into a data science project with poor understanding and uninformed expectations about what they will get out of the project, creating friction for the delivery team, frustration for the stakeholder, and waste of precious resources. Here are five must-do’s to avoid these issues.

In our previous blog posts, we looked at what data science is and how it benefits data projects. We also discussed how to build a team to drive business goals with data, including data engineers and data analysts. In this blog post, we want to look at five key things stakeholders should know when considering a data science project.

Y. Karen Zheng, Associate Professor, Operations Management at MIT Sloan states, “One of the biggest mistakes companies make about analytics is the disconnect between the technology and real business decisions.” Indeed, businesses often jump into a data science project with poor understanding and uninformed expectations about what they will get out of the project. Collecting data for the sake of having data won’t drive value or increase ROI for your business. Stakeholders must determine what data projects they intend to undertake based on how they intend to use the data and analytics to inform business decisions.

Therefore, it is accurate to state the most successful analytics organizations are always decision-driven. Zheng states, “They start by asking what business decisions they need data and analytics for, then investing resources to collect the right data and build the right analytics.” To achieve the best results, stakeholders can keep five simple items in mind to support their data project to ensure they are deriving the most helpful information and not wasting precious resources.

1: Define project responsibilities

When undergoing a data science project, stakeholders and data experts need to be clear on responsibilities - the business (stakeholders) define project objectives and business goals while data experts work with the data. Take this a step further and clearly outline the responsibilities of each stakeholder and data expert from data collection and preparation to business analysis, algorithm selection, model building, and evaluation. 

In addition, agree on expected deliverables before beginning work. Jordan Levine, Lecturer, Operations Research and Statistics at MIT Sloan teaches his simple strategy to analytics, “First, identify three sources of use cases and start to build them.” These use cases can be determined based on the deliverables stakeholders would like to pursue and can drive value for the business.

Levine’s three use cases include:

• Use cases that support C-level metrics (think revenue, cost, and risk).
• Business processes that can be supported by self-serve analytics and dashboards.
• Compliance must-do activities. 

Utilizing data visualization tools at key junctures to clearly illustrate accomplishments and areas where improvement may be needed during different stages of the project lifecycle. As each of the initial data goals are achieved, more can be undertaken with greater support from the business as the results drive improvements in sales, ROI, and actionable insights.

2: Understand your existing data

Stakeholders who need data science services often do not know where and how data is organized within their enterprise. Project scoping and planning activities should uncover what kind of data already exists, where it is stored, how to access it, and who is responsible for each part of the data. Ideally, your data team and stakeholders should be prepared to answer some, if not all, of these questions before engaging in the data science journey. This fundamental understanding will help the data team advise you on whether and how business goals can be met. And it will drastically reduce the initial amount of work the data team has to do to set out on the journey of answering the business question.

3: Ensure data is accessible

Oftentimes, the data science team appears to be deadlocked from accessing all required sources of data due to them being locked by third-party software providers, part-time system administrators, or complex security measures set up by corporations. Make sure that your data and information technology teams are talking to each other and have a shared understanding of the access required to support the data stream.

4: Set realistic delivery deadlines

Collecting and organizing data usually takes much longer than customers anticipated due to the complexity of the existing data and new features that are needed to complete complex analytics. As a result, you need to set realistic goals and timelines that are agreed upon by all stakeholders keeping in mind that data can often take longer than stakeholders may think based on how organized, clean, and error free the data is.

5: Remember - data doesn't lie

Data analysis reflects the actual state of affairs, which may not align with what stakeholders want to see. For example, if a business wants to increase sales, data analysis could reveal that its current marketing efforts are ineffective or that its pricing is too high for its target market. While this may be something that stakeholders do not want to hear, this is precisely why data is so powerful - to show us how things are, not how we wish them to be. With this information in hand, the business can make more informed decisions about increasing sales, such as adjusting its marketing strategy or lowering prices. 

In closing

Stakeholders who follow these guidelines feel confident that derived insights were valuable while their resources were used effectively and efficiently. If you’re embarking on a data project for the first time, ChaiOne’s data experts can serve as solid, reliable guides.