How do you use data to solve problems?

In my experience, it is very important to make sure that you understand the questions you have been asked to answer before you begin your analysis. The bigger the dataset you are working with, the more likely you will have a vast amount of insights to provide. It can be very tempting to find insights that you think are valuable. However, even if the insights you are providing are valuable, they are not the answers to the specific questions your stakeholders have. You also don't want to give a bad impression of being someone who doesn't prioritize listening to understand the objective before working on a project

The data must answer specific questions, otherwise it is superfluous to requirements. 1. What questions do we need the data to answer? 2. What data needs to be collected to answer these questions?

Problem well stated is half solved. Approach it from various levels: 1. How does solving this impact the overall system? What is my end goal here? 2. Is this the correct problem to solve? We tend to see the symptom as a problem. Analyze thoroughly. 3. What assumptions are we making here? Clearly define the knowns/unknowns in the problem.

In my experience, the lack of clear definitions for terminology had been a root problem for many organizations. This problem can present itself across multiple disciplines and departments within a single organization or cause confusion when trying to align new objectives with existing processes. Without the proper foundational framework and planning, all subsequent initiatives are subject to much higher risk of failure.

Data is an important source of information when solving problems. It can be big data but doesn't have to be so. Even gathering some clues the processes or parts show can be very important for clue generation. It's like bringing the parts of a puzzle. The puzzle can have a small one with a few parts or it can be a very big one with thousands of parts.

In my experience, prior to collecting data we create a data management plan that includes, at minimum, the following: metrics name, metric definition, source of data, exclusions, inclusions, sample size, numerator, and denominator. By creating a data management plan and training all data collectors on how the data should be collected as well as the numerator and denominator you can ensure the data is collected correctly each time and that everyone is on the same page.

Creating a frame or a matrix where all the definitions are well captured through organizing data in the right categories . Analyzing this frame for reasonability and accuracy. Relying final results using the matrix and communicating them to the right parties to complete the cycle.

In my experience, people jump too quickly to hypothesis. Even in this step you should continue collecting data and extract valuable information using data to converge down to a few, maximum 3-4 root causes. In most cases it' s possible to converge down to 2 probable causes. Until you reach this stage you should use the power of statistics which means looking at the data from different perspectives.

Visualizations, such as graphs and charts are always helpful. When creating graphs and charts be sure that they can stand alone by including a title, time period, source, x and y-axis titles.

Before this step it is important to explore and think about all possible root causes of the problem. Many people skip this step, yet proper root cause analysis can be the most insightful to considering how to properly solve the actual problem. Dig into the data, ask questions about how the data was collected and the circumstances it was collected under. Examine the real world scenario- is it properly represented in the data? Is your data good quality? Once you identify possible root causes of your perceived problem, make sure the problem has been accurately identified. Sometimes going one level deeper after doing root cause you may find the problem is actually something else than originally thought.

A great tool to leverage in this step is Plan-Do-Study-Act (PDSA). The team should gather enough information during the 'Do' and 'Study' step that once at the 'Act' step they can make a decision to adopt, abort, or abandon the brainstormed solution. If 'adopting' the solution, be sure that the solution idea was tested under as many conditions and circumstances as possible. For instance, a hospital may want to test a solution idea with as many nurses as possible, to include seasoned and new nurses. They would also test the solution on day shift and on night shift. Testing under as many conditions and circumstances as possible helps the team to ensure viability of the solution before implementing widely.

1. You can use A3 Report. This is a four-stage cycle: Plan-Do-Check-Act (PDCA cycle). 2. You can use the SMART (Specific, Measurable, Achievable, Relevant, and Time-bound) method. 3. You can use the DMAIC (Define, Measure, Analyze, Improve, and Control) method.

If the problem at hand is a complex one, the problem solving team should better try new and creative ways to understand and solve it. Usual tools like A3, 8D, Is/IS-Not are almost a century old, but they still work. Additional to them there are some open source programs which can be employed also after some training in them. For a complex problem where more than one root cause exist and interact with other ones, decision tree approach is a powerful method.