How to Approach Any Analytical Business Question
The following overview is written as a guide to those new to analytics or new to solving business problems.
1: Determine the List of Business Questions
Many of us had to say this for quite a few years before it became a mainstream message but documenting the list of business questions to be answered is very important. Without this documentation, it’s like running a project without a defined project scope. If you don’t document the specific questions to be answered, here’s what can happen.
People requesting the analysis may keep asking for more. That’s called scope creep and can impact your ability to deliver results on time. If this happens, you need your documentation to show that the additional work is “out of scope.” The response to such requests should be, “We can do that but since this adds to our project scope, we will need to add a week to the delivery date for the results. Are you okay with that or should we consider this a Phase 2 to talk about after we deliver the original scope by the agreed deadline?”
Another thing that can happen is that enthusiastic analytical professions professionals create scope creep. We love enthusiasm, but sometimes analytical experts get caught up with the “coolness” of the data and begin to analyze everything about everything. Time is limited, so keep your team members focused on the original business questions. There’s always time after meeting your project deadline to “go exploring.”
2: Collect and Check the Data
This step is about making a list of the data you need to be able to answer the business questions. Gather the data and, before you spend 20+ hours doing a detailed analysis, spend some time creating some high-level visualizations. This makes it easy to see if any of the data looks irregular. Investigate outliers or data that doesn’t make intuitive sense. For example, I often see employees you have aged 5-7 years old. These are errors in the data that should be corrected if there are many. If there are only a tiny number of these relative to the size of your data, then these errors are unlikely to affect the results of calculations involving this field. Data is rarely 100% perfect.
You may also find that you could not obtain all of the data you wished to have to answer the business questions. Can you proceed with answering the questions without that data, partially, or do you need to set up a new data collection activity?
3: Determine Which Analysis Methods Should be Used
Now that you have your data, or enough data to proceed, document the plan of which analysis methods will be used to answer the business questions. Sometimes there are multiple approaches to answering the same question. Will you create visualizations? Will you proceed with hypothesis testing?
Special note: In the image shown with this article, you will notice an arrow between steps 2 and 3. There are some cases where it is best to document the analysis methods you will use before you collect and check your data. One example of this is when you need a tool like a paired t-test during your project which requires some up-front data collection.
4: Conduct the Analysis and Interpret the Information
Proceed with performing the list of actions you need to take to answer the business question. This can be time-consuming but sometimes this is the easier part. The harder part is often the ability to interpret what you see in the analysis and apply it to obtaining an answer to the desired questions. How do those results apply to your business?
5: Tell the Story
The analytics world is divided on whether to use this phrase (I’m not a fan of it), but what this really means is to communicate your results to the proper audience and to tailor your communication to that audience. Presenting your results to a room full of engineers and mathematicians is not the same as presenting your results to a room full of executives. One group prefers a high level of detail to be convinced of the validity of your study; the other prefers that you “cut to the chase” and focus on the implications to the business.
While the devil is in the detail of each of these steps, I hope I have provided a useful guiding document to which people can refer for planning purposes.
Tracey Smith is a recognized analytics expert, speaker and author. She is the President of Numerical Insights LLC, a boutique analytics firm addressing the needs of businesses large and small. If you would like to learn more, please visit www.numericalinsights.com or contact Tracey Smith through LinkedIn.