We’re hearing a lot about how Big Data is the solution to everything. It all depends on what you do with it. How are you integrating it into problem resolution in your business?
Problems are taxing, time consuming, and a combination of human performance and supply chain issues. As a business owner or manager, you know that they are omnipresent and inevitable.
While gut instinct works for simple problems, it’s notoriously unreliable for complex business problems. Witness the fall of many CEOs and companies that made decisions without tapping available data and analytics. A good problem solving process follows a structured approach to reach the best possible solution.
After years of Big Data management, we learned we can’t make problems go away, but we can teach a better process for tackling the issues. ProCogia uses an 8-step approach to better decision making.
We have seen better outcomes for our clients when they combine their data with structured problem solving using each step below.
1. Define the Business Problem
Make sure you know what you are trying to solve by spending the time to adequately describe the issues. Albert Einstein once said, “If I had an hour to solve a problem I’d spend 55 minutes thinking about the problem and 5 minutes thinking about solutions.” Understanding what are you trying to solve and what state you want to reach is the first and most important step in structured approach towards problem solving.
Are you losing customers? For SKU questions, do you know if it’s the record keeping system that doesn’t work or is it the demand forecasting that needs to be changed?
2. Free thinking and Hypothesis formulation
Don’t jump too fast to causes and solutions. Once you’ve defined the problem, bring together a team and “work out loud” on the issues you think led you here. Don’t censor anyone but if you feel it’s getting off track, make sure you record “off topic” things separately to address later.
Once you’ve listed the causes, develop a hypothesis to test proposed solutions. This is where the data gets used. In real life, what are all the factors are likely to affect customer turnover?
Are SKU shortages caused by surge in the demand, or is it a purchasing and timing issue?
3. Data selection and Data quality analysis
Identify relevant data to validate your hypothesis. Are you using the right information to test? Once you select the data make sure it was accurately gathered and consistent in terms of variables, collection sites, etc. Test and checks to filter for outliers or errors.
Is age a factor in customer turnover? If so, are we collecting age correctly for our customers? Is a particular category of SKUs always short? Are they being recorded properly in the system?
4. Clean and transform data_
Data validation is a critical step to ensure data accuracy. Are the correct quality controls in place? As you test, make sure that you correct misclassified and add missing data. You may need to collect and transform data to a different format than what originally existed to be able to follow the next step.
For customers where we are missing age data, is it possible that it was gathered and not entered? In that case, it can be added. If it is missing, how do we ensure good testing? What work-around do we use: average age, or looking at similar customer profiles to calculate the missing value?
Review the SKU data to make sure that it’s all been correctly entered and classified before running tests.
5. Data modeling and analysis
Just as good problem definition is essential, picking the right data model is equally important. Different model alternatives should be explored and ideally the one that gives the most accurate results should be deployed. However, in real life, there is always a tradeoff between timeliness, budget and accuracy. Pick the best model you can given those real life constraints. Do we need to implement K- Means or hierarchical clustering to define our customer buckets for the turnover problem? For our SKU problem, do we start by improving the inventory policy or by implementing pooling and/or do both?
6. Findings and Insight generation
If you picked a good model, the findings and insights unlock hidden issues or causes that can be addressed. Was your hypothesis validated or did you find out something totally unexpected? Either way, you have results to apply towards a solution. For example you will find out if age relates to turnover or not. If it does, which particular age group? Which particular SKU or product categories need the most attention?
If your hypotheses was invalidated, you aren’t lost – it’s iterative. Go back to problem definition and re-define based on what you learned.
7. Evangelize the solution
Solid solution? The next step is to share the results with stakeholders or leaders and get buy-in. If there is dissension, make sure that there is agreement on the problem definition and steps to solution. Assuming that these stakeholders have been participating in the process, you may need to iterate or make small adjustments. However, given this structured problem solving process, there should be less dissension than with other methods. For example, if the results validate your hypothesis and there’s still dissension, is that bias against change or a valid concern?
8. Enable Resolution and Action
Once you implement a solution, continue to monitor and retest until the problem is history. Don’t ignore this crucial step. For example, are we looking at ways to address the particular age bracket that is likely to turnover? Are we looking to advise our procurement team to aggressively order the SKUs that are causing inventory shortages?
Conclusion
In addition to providing good business solutions, this approach helps organizations in team building, strategy development, and future problem resolution.
Overwhelmed by the thought of tackling this on your own? Give us a call at (425) 753-4770, email us at meharpratap.singh@procogia.com, or visit us online at ProCogia.com. We’ll help you use your Big Data with these steps to solve what’s keeping you up at night.