Companies have a lot of data, but it isn’t always clear how to turn that data into added business value. In this webinar, Xomnia analytics translator Tim Reus explains how to identify a good data science use case. He also demonstrates how Xomnia applies a use case canvas to our data science projects.
Why use cases are important
A use case is a question that can be answered by using data. Tim provides three reasons why finding the right use case is essential to the success of your data science project. First, he says understanding your problem from an end-user perspective helps with finding the right data-driven solution. A good use case will also help you define how you measure the project’s success and ensure that your solution adds business value. Lastly, it will enable you to go beyond data science in a technical sense and view it from a business perspective.
Where to find a good use case
There are four common areas for businesses to discover good data science use cases. One is within corporate strategies. Examples of these use cases are recommender engines, data-driven marketing, and predictive maintenance. However, these strategies are often too high level for a viable use case.
Operational problems are the most common sources of data science use cases. Some possibilities are stock inventory predictions, picture recognition for data entry, decision support tooling, and pre-validations in processes.
Another way to uncover a potential use case is within data insights. Often a data analyst or data scientist will notice an anomaly or something that stands out within the data. This can be translated into a use case. At Linkedin, for example, a data analyst noticed that the data on users’ connections could be used for the platform to suggest relevant new connections
External factors can also facilitate a good use case. The COVID-19 pandemic has certainly been a change catalyst worldwide. We’ve seen data science implemented in South Korea’s drive-through virus testing sites where image recognition and heat sensors are used to determine if someone has a fever. Innovation in the market can also pressure a business or industry to change their strategy to avoid their losing market position. We saw this develop when Deliveroo disrupted the casual dining industry.
Measuring your use case
Tim recommends measuring the use case against both business and technical key performance indicators (KPI). This will ensure that your data science project adds business value. You should also assess the use case at the beginning, middle, and end of the project.
Business KPI examples:
- Average order size
- Customer satisfaction
- Time on page
Technical KPI examples:
- Mean Squared Error
- Response time
The use case canvas
When working to map out a good data science use case with Xomnia’s clients, Tim implements a use case canvas. He walks step-by-step through this process in his presentation. As you implement the use case canvas in your business, try to answer the following questions:
- What goal is supported with this use-case?
- What is the objective of the use case?
- How do we measure success?
- Who is the use-case owner?
- Type of models possibly used?
- Potential impediments? (e.g., legal, privacy, third parties)
- Who will use the product in the end?
- What data is required?
- Where would the product run? Which infrastructure or environment?
- Who are other important stakeholders to involve in the project?
- Any other things worth noting?
Tim emphasises that engagement is fundamental to every use case. So, make sure you put the people in the centre of the implementation. Data science is always a tech problem, but never just a tech problem.
Click here for a PDF version of Xomnia's use case canvas.
Would you like to watch the full presentation? Just enter your email address below and we’ll drop it in your inbox.