The benefits of generative AI for data analysts are often emphasized in relation to hands-on, daily tasks such as debugging formulas or writing macros. Additionally, AI assists in conceptual tasks like choosing the most appropriate chart type based on the data and the intended audience.
While these everyday applications are indeed significant, the potential of generative AI extends much further. It plays a crucial role in aiding analysts with strategic tasks, helping them to effectively frame problems and develop suitable workflows for analytical challenges.
In this post, we’ll explore some renowned tools and techniques for framing data problems, and demonstrate how Microsoft 365 Copilot can be leveraged to think through and guide real-world projects. For each scenario, we will apply a well-known problem-framing framework and use Copilot to explore how to implement it for our specific needs.
I’ll provide a basic explanation and resource for each of these techniques, but if you’re not familiar with them it’s also not a bad idea to learn about them through generative AI itself! As you’re going to see, each of these frameworks involves some iteration and multiple steps. To make the most out of using this as a sounding board, you should be well-versed in the workflows and in crafting prompts. Otherwise, they won’t be very meaningful if you have absolutely no understanding of the problem, the framework, prompting techniques, and so forth.
Setting SMART goals
A SMART goal is a specific, measurable, achievable, relevant, and time-bound objective designed to provide clear direction and benchmarks for progress. For a data analyst, using SMART goals in problem framing ensures that the analysis is focused and aligned with the overall business objectives.
By defining clear parameters, the analyst can accurately measure progress and outcomes, ensuring that the goals are realistic and relevant to the organization’s needs. Time-bound deadlines add urgency and accountability, helping to prioritize tasks and maintain momentum. This structured approach facilitates more effective decision-making and helps in delivering actionable insights that drive business success. You can read more about SMART goals here.
- Problem statement: “I’m a data analyst at a company observing steady traffic into its ecommerce website, but not enough conversions into sales. My company wants to improve online sales numbers. What are some questions I should ask my business partners to convert this into a SMART goal?”
Cross-Industry Standard Process for Data Mining (CRISP-DM)
CRISP-DM is a widely used methodology in data mining that consists of six phases: business understanding, data understanding, data preparation, modeling, evaluation, and deployment. It provides a structured approach to planning and executing data mining projects. You can read more about CRISP-DM here.
Let’s take a look at walking through framing a data analytics problem with CRISP-DM and Microsoft 365 Copilot. Here are some prompts we could use. Note that we are going to break the problem down into multiple steps, one for each leg of CRISP-DM, with some stage and expectation setting at the beginning, explaining exactly who you are and what you are looking to achieve.
- Scenario and business understanding: “I’m a data analyst at a retail company. I’d like to optimize inventory levels to reduce carrying costs while ensuring product availability. My goal is to predict product demand more accurately.
To get started, can you help me understand the potential impacts on profit and customer satisfaction versus understocking in retail?”
- Data understanding: “What types of data should be collected to predict product demand effectively?”
- Data preparation: “Generate a checklist for cleaning sales data including handling missing values and outliers.”
- Modeling: “List and describe three predictive modeling techniques suitable for forecasting product demand.”
- Evaluation: “What metrics would be best to evaluate the accuracy of demand forecasting models?”
- Deployment: “What are the steps to deploy a demand forecasting model into a retail business environment?”
As you can see from the previous workflow, this is an iterative process. As you refine your approach, you may need to revisit or adjust previous prompts. For example, once you’ve chosen a specific tool for presenting and sharing the data, your prompts might become more specific to that tool. However, this is just a starting point. There will be trial and error and iteration, which is the nature of both CRISP-DM and generative AI.
Current State-Future State (CS-FS) Analysis
The CS-FS framework is useful for mapping out the current status of a business process or situation and envisioning a desired future state. It helps in identifying gaps and planning strategic actions to bridge these gaps. You can learn more about the CS-FS framework and specifically the CS-FS map here. Let’s see how we can get Copilot to help us walk through the basics of a CS-FS framework.
- Scenario and current state: “I’m a data analyst at a financial services firm. My department is experiencing slow report generation times, impacting decision-making processes.
What do you need to know about the current process and tools used for generating financial reports to help me build a current state vs future state framework?”
- Future state: “Imagine an optimized process for report generation at a medium-sized financial services firm for the types of reporting tasks I’ve mentioned to you so far. What would this look like?”
- Gap analysis: “Identify potential gaps between the current and future states of report generation that you have mentioned thus far.”
- Action plan “Suggest actionable steps to transition from this current to future state in report generation.”
Issue Trees
Issue trees are a logical framework that help break down complex problems into manageable, component parts through a hierarchical structure. They start with a broad, overarching question or issue at the top and branch out into detailed, contributing factors or sub-questions. This method helps identify the root causes of a problem and the relationships between different elements of the issue.
For data analysts, issue trees are invaluable in framing problems because they clarify the pathways for analysis and guide the systematic exploration of data. By using issue trees, analysts can ensure that they cover all aspects of a problem, prioritize their analyses based on the branches that could have the most significant impact, and maintain a clear focus on the ultimate questions they need to answer. You can read more about issue trees here.
Let’s take a look at using Microsoft 365 Copilot to help us get started and thinking about an issue tree.
- Scenario and breakdown: “I’m a data analyst at an e-commerce company noticing a significant drop in customer retention rates. Create an issue tree that breaks down the potential causes of declining customer retention rates in e-commerce.”
- Hypothesis generation: “Generate hypotheses for each branch of your issue tree on why customer retention might be declining.”
Root Cause Analysis
Root cause analysis (RCA) is an approach used to identify the underlying reasons for faults or problems. By focusing on the root causes rather than just the symptoms, it aims to resolve issues in a way that prevents their recurrence. Data analysts can use this approach when they encounter inconsistencies, errors in data sets, or unexpected results in their analysis. By applying RCA, analysts can understand not only what and how an issue occurred, but also why, which helps in devising effective solutions that ensure data integrity and reliability.
The “five whys” technique is a popular tool often used within RCA, particularly for its simplicity and effectiveness. It involves asking “why?” repeatedly (usually five times) to peel away the layers of symptoms which can lead to the core of an issue. You can learn more about RCA here.
Let’s look at using RCA with the assistance of Microsoft 365 Copilot:
- Scenario and problem identification: “I’m a data analyst at a manufacturing company facing frequent equipment failures that halt production lines. List common reasons for equipment failures in a manufacturing setting.”
- Root cause inquiry: “Provide some questions I can use to delve deeper into each potential cause of equipment failure to uncover the root cause.”
- Solution brainstorming: “Suggest practical solutions or improvements to address the root causes of equipment failures.”
Conclusion
As you’ve seen through various examples and methodologies, Microsoft 365 Copilot not only enhances the capabilities of data analysts but also transforms the way we approach and solve complex problems. By integrating generative AI into your problem-solving toolkit, you’re equipped to navigate through the layers of data more efficiently, ensuring that every strategic decision is backed by solid, data-driven insights. Do you have any questions about the frameworks discussed? Or are you curious about implementing Microsoft 365 Copilot in your own projects? Don’t hesitate to leave a comment or get in touch.
Leave a Reply