Part 1 - Data Analysis for Craft Breweries
Every brewery has unused production data whether it's on paper, google sheets, or in a software platform. When analyzed, production data can become an integral part of a QC program through statistical process control (SPC). Larger breweries have been doing this for decades with significant financial and quality returns. As one of the cheapest forms of QC, there is no reason small breweries can’t do it too. In part 1 we’ll review a few basic principles of statistical process control (SPC) before giving you examples to work through in part 2.
Breweries Applying Process Control
Before getting into the nuts and bolts of process control, here are a few breweries using these tools successfully to make better beer with less money.
The Concept
Let’s start with an example before getting into the complexity of making good beer.
These example bolts need to be manufactured at a certain width.
Every manufacturing process has inherent variability.
The manufacturer needs to reduce the variability enough so that each bolt is an acceptable width.
If the variation in bolt width is normal, it will look like the bell curve below. 99.7% of measurements will fall within 3 standard deviations of the average, which we’ll call the control limits. That range is what we are reliably capable of producing, so it’s called the process capability .
Let’s flip this bell curve on its side and plot the points on a graph. Now we can see as we’re plotting measurements if they fall within normal variation.
If a width falls outside of control limits, we’ll know that it is abnormal variation. Once identified, we can identify the cause and get back on track.
If we apply this same concept to control points along the way, we can identify out of control parameters before they become an issue in the final product. View the charts below. If Machine Feedback 3 were in control, our final product may not have been out of width specification.
The upper and lower control limits are calculated values based on what the manufacturer is producing. Sometimes these limits do not match up with the actual goal. In that case we would label the process incapable of meeting demand.
In this example, only a portion of the bolts that we are producing work. We’ll have to throw away all of the products on the right side of the red line.
To address this, we have to reduce the variability in our intermediate products until the final product is less variable. Then we can adjust one element of the process at a time, to shift the mean.
Applying SPC in Brewing
The ultimate goal is to make good beer consistently. To do that we need to control the intermediate products. We’ll walk through how to define what those are, and monitor them.
Defining Intermediate Products
Beer is complicated, with hundreds of variables effecting the final product. If we view it in pieces, the task becomes less daunting.
The ASBC has defined the intermediate products for us. Under the sampling plan you’ll see the brewing process split into stages. If we control what comes out of each of these stages, we control the final beer.
2. Deciding what data to collect
Every piece of data collected in the brewery should be for the purpose of controlling intermediate products. If we are collecting and analyzing the correct data, we should be able to trouble shoot unanticipated results with little downtime.
At Grist we recommend that breweries with limited resources start with values that give the most powerful information for the least time and money. We have helped define what those variables are - feel free to reach out for a prioritized brew log.
3. Organizing the data
Unless data can be easily manipulated and compiled, it’s next to useless (read; paper logs or brews on separate spreadsheet tabs). In Grist, you can store information however you’d like and it will be organized and analyzed for you. If spreadsheets are being used, batch information should be stored in one row. This allows us to see all values for the same variable in one column, and easily analyze that data.
4. Analyzing Data
Now that we are collecting and recording data, it is easy to format each value into a control chart. This will allow you to trouble shoot process issues by identifying parameters with excessive variation. This is visualized below in Grist, but can be done in Excel to a certain extent if data is stored properly. Use the form below to receive an Excel control chart template.
5. Making Data Driven Decisions
The benefit of having data organized, analyzed, and visualized is that everyone in the brewery can be involved in making data driven decisions. Experience and intuition can help fill in the gaps, but data should not be ignored. Creating a culture where data is approached without expectation or ego will be important. We will touch on this more in Part 2 of Data Analysis for Craft Breweries.
