The Any Analysis Framework
When you do any analytics, dashboarding, financial modeling, predictive modeling, and other data work, it can get out of hand quickly.
To keep everything in order as you work through a business problem, I have leveraged a tweaked version of the CRISP-DM method containing the following phases Business Understanding, Data Understanding, Data Preparation, Analysis, Validation, Communication & Deployment.
Each phase requires specific tasks and activities to be completed before proceeding to the next phase. This approach helps users stay focused on the task while providing guidance and structure throughout the project.
Additionally, it allows for proper documentation of all decisions made throughout each step along the way so that future team members can understand what has been done previously. Ultimately, this leads to better outcomes and more successful strategies that can be later operationalized in the business.
Business Understanding
Take time to understand the business objectives and ensure that your analytics project is aligned with these goals during this step.
Understand the business context and what the customer or stakeholder wants to achieve.
Identify needed resources, such as data sources, staff, budget, and time.
Become aware of potential risks, how to manage them, and any associated gaps.
Work with stakeholders to understand how the analysis results will drive impact through more revenue growth, better operational efficiencies, and/or to assist with other business decisions.
Data Understanding
Understanding data is exploring its structure, relationships, and underlying patterns. It is an important step in the analysis process because it helps us identify potential problems and opportunities for improvement. Data Understanding involves visualizing data, summarizing key characteristics of a dataset, evaluating its quality, detecting outliers and anomalies, and transforming raw data into a more meaningful form.
In this step, I like to dig into the data line by line and understand how that data relates back to actual business activities. I sometimes even go to data engineers to understand how this data came to be and how it was ETL/ELT-ed. I also find it helpful to work with business owners to get their perspectives on the data.
The key here is to be skeptical of the data and ask a lot of questions about it!
Data Preparation
Data preparation is the process of transforming raw data into a format that is suitable for further analysis. It involves cleaning and organizing data and applying various techniques to make it more useful and easier to work with. Data preparation generally includes activities such as data integration (combining multiple datasets), feature engineering (creating new features from existing ones), data transformation (scaling or encoding values), and the imputation of missing values. Preparing your data properly ensures accurate insights are extracted in the analytics phase.
For this step, I usually like to understand by first trying to understand what I need the data to look like to do the analysis and then start the data preparation. Pretty much working backward, so I know what the data prep should look like.
Analysis
Data analysis examines and models data to discover useful information, informing conclusions, and support decision-making. It involves reviewing raw data to identify patterns or relationships between variables to gain insights about a particular phenomenon. Data analysis usually involves applying various techniques and algorithms to help you gain insights about relationships between variables and understand how changes in one variable may affect the outcomes of other variables or processes.
By understanding these relationships between different pieces of data, organizations can make more informed decisions and optimize their operations.
Validation
Analysis validation is an important part of the analytics process. It is necessary to verify that the analysis is as expected and does not produce inaccurate results due to errors in data or assumptions made during the analysis.
Validation with stakeholders is an important part of the analytics process. To ensure stakeholders understand and trust the analysis, it is important to explain the data clearly, provide visuals or graphs to illustrate trends, and present a narrative of how the data were collected and analyzed. Additionally, regular conversations with key stakeholders can help identify potential issues early on and ensure everyone has a consistent view of the results.
Finally, when presenting your findings, highlighting implications for decision-making and areas for future exploration. This will help stakeholders stay engaged and make informed decisions based on the data.
Unfortunately, if your sniff test with stakeholders does not pass in this step, you might have to go back to the data understanding step to understand why the analysis does not look valid.
Communication & Deployment
Communicating the results to stakeholders is one of the most important parts of any data analysis. Communication is key in making sure everyone involved understands and can take action on the data insights. To ensure effective communication, there are a few tips to keep in mind when sharing your data analysis with stakeholders:
Provide Context - Before presenting your results, provide context for what the analysis was looking at and why it’s important. The “WHY” is the most important here because It does not matter how good the solution is if it does not solve the problem. You should clearly understand why you are solving this particular problem and what outcome you expect from your analysis.
Focus on Clarity - Make sure all the presentation's numbers, graphics, visualizations, and terminology are clear to everyone involved. Avoid jargon or technical language that may confuse people who don't have extensive experience with data analytics.
Include Actionable Next Steps - Make sure also to include actionable next steps in the presentation. Explain what decisions can be made based on the data insights and how stakeholders can move forward with these decisions. Even talk about how you would operationalize your findings in the business, how long it could take, and how the ROI would be calculated and defined.
Incorporate Visuals - Use visuals such as graphics or infographics to help explain the data analysis and make it easier for people to understand quickly.
Address Potential Questions - Consider any potential questions stakeholders might have about the analysis and address them proactively during the presentation. This will help ensure everyone has a good understanding of the results and their implications.
The Analysis Framework
Here is a visual map of how the analysis framework works to bring everything together in this post.
