Recently, a well-established, rapidly expanding beer company invested heavily in a modern, state-of–the-art brewing facility. The new facility dramatically improved quality and productivity, and also reduced costs through the application of new technology. As a next step, the beer company began exploring methods of achieving a further quantum jump in performance. Recognizing that technology and added investment might offer diminishing returns, they decided to explore total quality management (TQM) principles as a means of achieving this ambition in the manufacturing arena.

The resulting project, which involved several Lean Six Sigma analysis tools, was created to demonstrate what sort of benefits TQM and Lean Six Sigma could generate to help the company decide whether to expand such continuous improvement methods across the organization.

Beer Brewing Primer

Before going further into the TQM process, here is a quick description of the basic process used to manufacture beer. The key steps in beer brewing are as follows:

  1. Soak malted barley and other ingredients in hot water to make wort, a liquid extract that contains the sugars that will be fermented into alcohol.
  2. Boil the wort, add hops as a bittering agent and filter the resulting mixture.
  3. Once the wort is cooled, add yeast to ferment, mature and lager the beer in fermentation tanks.
  4. Bottle the beer.

Selecting the Theme

To begin, the senior manufacturing management attended a two-day quality mindset program to get an introduction to TQM, understand why and how it works, and, most importantly, to open their minds about exploring change. The group was then asked to brainstorm their priorities and select a theme for the project, which, if successful, could help demonstrate the potential benefits of TQM.

The team selected “Quality Improvement” (QI) as the theme and chose one of their large, modern locations (referred to here as “Factory A”) as the venue for the project. A cross-functional team from Factory A’s management was selected to take part in the two-day quality mindset program.

The project was structured around TQM’s seven-step problem solving method, defined as:

  1. Define the problem
  2. Conduct root cause analysis
  3. Generate countermeasure Ideas
  4. Test the ideas and implement in production
  5. Check the result
  6. Standardize procedures
  7. Prepare a QI story

In this case, however, many of the above steps were merged or conducted in a non-linear fashion at various points in the process.

Step 1: Define the Problem

The Factory A team brainstormed and prioritized the key areas of quality that needed dramatic improvement. Eventually, “improving consistency of taste “emerged as the target area. After using a ranking exercise to review the various possible attributes of taste, the team chose “bitterness of beer” as the parameter that would be tackled in this project.

In the units of measurement for bitterness prescribed, the standard was 10*+/-2 units (Note: for confidentiality reasons, 10 is not the actual value but just an arbitrarily assumed base). On examining past quality control (QC) measurements, 75 measurements over one month revealed the following:

Average bitterness (B) = 10.2
Sigma (s) = .029

After seeing these measurements, it was puzzling why the team perceived a need for improvement in consistency of bitterness. The team decided to reevaluate the bitterness measurement system and the sampling scheme.

Measurement: The measurement of bitterness involves four stages: degassing (removal of all carbon dioxide gas from the sample), shaking, centrifuging and measuring. Observation of the measurement process yielded obvious inconsistencies:

  1. The degassing was done manually by tipping the beer from one tumbler to another.
  2. It was not clear when the beer was degassed enough.
  3. The centrifuging was being varied from operator to operator.
  4. The specified times in the standard operating procedure (SOP) were not being followed.

Countermeasures: 1) A magnetic stirrer was introduced and the time of degassing was standardized, and 2) the necessity of obeying the SOP timings was emphasized.

The repeatability and reproducibility of the system was checked using gauge R&R analysis and the variation of .07 was deemed satisfactory.

Sampling: Past performance was based upon a sample size of one bottle per shift for a production of 25,000 bottles per hour. The team suggested that workers should check the bitterness much more intensively to confirm performance. An hourly sampling was done for two days. The results are shown in Table 1.

Table 1: Sampling Results

  Number of samples Average Bitterness Avg.
+ 3s
Avg.
-3s
QC sampling 75 10.2 11.1 9.33
Hourly sampling 43 11.2 12.3 8.15
Control limits     12 8

 

 

 

 

The hourly data showed that the process did not deliver even 3-sigma quality (i.e. 99.7 percent of products within the control limits) over 40 hours. Therefore, it was very unlikely that the process could deliver 6-sigma quality over one month.

An X-bar control chart was developed, as shown in Figure 1.

Hourly Bitterness – Mild Beer Bottle
Hourly Bitterness – Mild Beer Bottle

From these measurements, two important mindset changes were achieved:

  1. The process was not perfect, as previously thought; there was room for improvement.
  2. The measurements could now be trusted to mirror reality.

Regular plotting of the chart was commenced, and 100 hours was selected to represent the population variation. The average and variation of the first 100 hours of readings measured the current state and helped define the problem, using the TQM formula: problem = desire – current state.

Step 2: Finding Root Causes

The average measurement (10.8) of this batch of beer ranked at the top of the bitterness scale for “mild beer.” Because the high average fell within the overlapping range for the “strong beer” category, the team saw that the beer was going out of range.

Generate and test countermeasure ideas: Reducing the bitterness average required a reduction in hops added by 100 grams per batch. Once this was implemented, the average came down progressively to 10.3 in about 150 hours (see Table 2).

Table 2: Progressive Reduction in Bitterness Over Time

Hours 50-150 75-175 100-200 200-300
Average bitterness 10.8 10.6 10.4 10.3

 

 

 

As the project progressed gradually, the hops dose was fine-tuned; reducing the hops addition by another 100 grams achieved an average of 10. The project’s first objective had been largely achieved, along with minor cost savings. The stage was now set for the much more difficult task of reducing the variation by 50 percent.

Brainstorming generated a list of possible causes of variation, which were ordered using an Ishikawa, or fishbone, diagram. Most key variables and recipe components were controlled automatically and were remaining within tight limits. Only two processes were manual, and were varying:

  1. Preparation of a hops solution and the time of its addition to the wort.
  2. Weighing of hops – a balance that needs calibration, cleaning and careful usage.

SOPs were developed for the above two factors and implemented.

The bitterness of the beer develops through the process in three key stages:

  1. Wort making
  2. Fermentation, maturation and lagering – two three-stage processes carried out continuously in the fermenters.
  3. Between fermentation and bottling, there was a change in bitterness.

Regular measurement of bitterness of each wort batch was introduced and the team developed a control chart to review the 3-sigma limits after every 50 batches. The experiential standard for the wort’s average bitterness was 20 (Note: Again, this level is different from the actual value due to data confidentiality reasons).

For the first 50 wort batches, the average bitterness was measured at 21, with a sigma of 1.31. The average was being adjusted in line with the bitterness of beer.

With the basic process stabilizing, control chart plotting was transferred from quality assurance to the shift brewer. A small line team started to meet and question every out of the ordinary peak or trough daily and killed the sources of variation. Gradually, the sigma reduced further and over two months (200 hours) by another 50 percent – from 0.68 to between 0.35 and 0.38.

Step 3: Check the Results

When recording started for this project, the state of wort bitterness was as follows:

Average: 10.7
Sigma: 0.11
3-sigma limits were 10.37 to 11.03

The improvement in wort bitterness occurred in three stages:

Phase 1
The average wort bitterness improved dramatically, from 10.7 to 10.8. When team members queried why, two causes emerged:

  • One of the ingredients of the recipe was a thick liquid received in drums. The team realized that about 5 kilograms were remaining in each drum and could be removed by washing with hot water. This process was implemented to enhance yield.
  • Batches of 10.6 (the lower end of the bitterness scale) ceased to appear and were replaced by some 10.8s, hitherto not present as the process gradually standardized for bitterness variation control.

Phase 2
The TQM team then wondered why if one batch could be 10.8, why should the next one be 10.7? Relentless search and elimination of minor causes of variability gradually led to an increasing number of 10.9 readings. By end of Phase 2, the average wort bitterness measurement had moved to 10.85.

Phase 3
Thereafter, regular reviews and gradually increased tightening of process parameters raised the wort bitterness average further to 10.9. A few batches touched 11, but most were at 10.9 and a few were 10.8. The improvement achieved is summarized in Table 3.

Table 3: Final Project Results

  Initial state After change
Average wort bitterness 10.7 10.9
Sigma 0.11 0.04

 

 
These results translated into a major cost savings of $150,000 annually.

In the future, the team may consider measuring the variation ratio of pre- and post-fermentation bitterness to try and make the fermentation process even more consistent.

About the Author