As I wrote a few weeks ago, quality control analytics can save serious costs. I listed 5 examples of such cost savings in that article. Now, the question is, what can manufacturing and purchasing managers do to utilize data for quality management? How to start?
Should you look at the data they have already collected, and look for hidden insights? In that case, they have a ‘mystery’ on hand.
Or should you gather more data, and the right conclusions will become obvious once you have the missing data points? That’s a ‘puzzle’.
In this post, I’ll walk you through both scenarios and the approach to dealing with the data in their cases.
The difference between puzzles and mysteries
Bob Bruner explained it clearly:
The distinction between puzzles and mysteries was drawn by Gregory Treverton, a national security expert. A puzzle has a “simple, factual answer” and is solved by getting more information. The location of Osama bin Laden is a puzzle. Finding the answer requires increasing the collection of intelligence. A mystery is different—it may have many contingencies and is solved by more analysis rather than more fact-gathering. What will happen to Iraq in 2007 is a mystery. Solving puzzles depends on what we are told; solving mysteries depends on what we hear, on listening well.
Mysteries: looking for insights in data already collected
Some companies are already drowning in data. Or they can easily request a lot of data already collected from their service providers (e.g. inspection companies). For example, you probably already have this information:
- Smart categorization of products (by material, by complexity, by category…)
- Order basic data (customer, supplier name, quantity, date…)
- Factory audit scores
- Results of product inspections and laboratory testing, with the details of the reasons for failures
- Major complaints from customers, product returns, etc.
And, guess what? You can already pull many insights from these data.
Step 1 is gathering them in 1 database (for example Power BI)
That’s often where people stop because it poses an enormous challenge: many documents may be hard copies, may be in emails, etc., and that work may necessitate hundreds of man-days of work. Not to mention the commitment to keep doing it, moving forward.
One way to go around that challenge is to implement a new tool that will support the work of the employees and gather all those data. That’s one of the main goals of a quality inspection app.
Step 2 is running analyses
You probably have a lot of theories in your mind, and you would like to see if the data support them, right?
For example, “this factory is better than that one when making this type of product”, “seasonality does play a critical role”, or “when we do an inspection during production and then just before shipment on the same batch, we often have big discrepancies in the findings”.
You may have to make use of a “data science” consultant if your managers are spread too thin to work on this analysis. But your company can probably make better decisions this way. And you might find some surprising correlations… or lack of thereof.
Puzzles: gathering more data, to see the whole picture
In some cases, you gather some data but you can’t really make an informed decision, yet.
For example, your quality standard is not very well developed, so your QC inspections may be ‘passed’ or ‘failed’ but you are not sure it means the batches are good or bad.
One piece of data that is often missing is an approximation of the total cost of poor quality.
(When a batch is of unacceptable quality, it tends to have a negative financial impact due to chargebacks, expedited productions & shipments, loss of business, and so on. Few companies track all those costs.)
Without the cost of poor quality, many questions may remain unanswered.
Another example is the nature of the returns from the users/consumers. Many companies are too busy responding to complaints and do not spend the time to analyze the failures and to categorize & save them. That lack of data prevents comparison with the issues found in pre-production reviews, in production inspections, etc.
Data for quality management is seldom one or the other
Data can be messy:
- Having a lot of data does not mean you have all the data you need. Spending time gathering some other information may be justified.
- The data you have may not be accurate.
- Having a lot of data can be a problem in itself. Different ways of looking at them might lead to different conclusions. It can be quite unclear.
Are you already using data for quality management? Have you faced mysteries or puzzles? What experiences do you have to share? Let us know by leaving a comment, please.
This free eBook shows importers who are new to outsourcing production to China or Vietnam the five key foundations of a proven Quality Assurance strategy, and also shows you some common traps that importers fall into and how to avoid or overcome them in order to get the best possible production results.
Ready to get your copy? Hit the button below: