Unintended consequences and perverse incentives are as old as social organization. [See Wikipedia's unintended consequences page for an interesting and informative list.] Many of them could be avoided by asking a few simple questions ahead of time.

1. Can this metric be manipulated without improving the underlying business?

If so, employees are inventivized to take the short cut.

Example: PM schedule compliance

As commonly measured, PM schedule compliance allows dates to be moved or schedules to be adjusted or old work orders to be completed out in order to bring the numbers up.

2. Can this metric be improved through undesirable or unintended behaviors?

If so, then the benefits of improving the metric will be offset in ways probably not measured.

Example: MTTR

MTTR encourages maintenance employees to store and hide extra supplies or take short cuts in workmanship to get jobs done faster.

3. Can this metric be improved by sacrificing long-term productive capability?

If so, then short-term results will turn into long-term decline.

Example: earnings per share

If earnings per share is growing at five times the rate of revenue or more, watch out!

Are these metrics, therefore, invalid?

Not necessarily. But consider selectively sharing the information or de-emphasizing them in benchmarking efforts in order to mitigate the risks, MTTR in particular is especially useful for planning purposes, but not for incentive schemes.

From TPM for Supervisors by Kunio Shirose:

OEE Improvement Process

1. Measure the extent of each of the six big losses
2. Determine how much each loss affects OEE
3. Find out what problems stand in the way of improving availability, performance rate, and quality rate
4. Determine targets and orientations needed to solve problems discovered in step 3
5. Find out how higher equipment effectiveness will affect cost-cutting and profit-boosting

The Six Big Losses

1. Breakdown losses
2. Setup and adjustment losses
3. Idling and minor stoppage losses
4. Speed losses
5. Quality defects and rework
6. Startup/yield losses

From TPM for Supervisors by Kunio Shirose:

1. Improvement activities designed to increase equipment effectiveness. This is accomplished mainly by eliminating the six big equipment losses.
2. An autonomous maintenance program to be performed by equipment operators.This is established as operators are trained to know their equipment.
3. A planned maintenance system. This increases the efficiency of traditional preventive maintenance activities.
4. Training to improve operation and maintenance skills.This raises the skill levels of equipment operators and maintenance workers.
5. A system for PM design and early requipment management. PM design helps create equipment that requires less maintenance. Early equipment management gets new equipment operating normally in less time.

Similar to the generic job description for a maintenance engineer, here is a template for a reliability engineer (or manager) paraphrased from Making Common Sense Common Practice: Models for Manufacturing Excellence. In MBM terms, these would be the responsibilities and expectations for the role of reliability engineer.

• Loss Accounting [Use existing databases to perform Pareto analysis and follow up to improve results. Since everything can't be done at once, this item is the foundation for prioritizing all others.]
• Root Cause Failure Analysis (RCFA) [Perform and/or facilitate RCFA and prospective solutions based on RCFA.]
• Managing the Results of Condition Monitoring Functions [Ensure quality data is being collected and that condition-monitoring technology is applied and used in an integrated way.]
• Overhaul/Shutdown Support [Review equipment condition to ensure the right work, and only the right work, is performed. Commission equipment during startup using applicable technologies.]
• Proactive Support [Working with all other departments to ensure good practices in design, purchasing, maintenance, stores, and operation.]
• Facilitator/Communicator [Deal with gray areas where responsibilities are shared by two or more groups. Find and implement solutions while avoiding finger-pointing.]

If a manufacturing facility has immature processes and systems, one tends to see a lot of emphasis on total production or equipment uptime. More mature facilities will have a broader perspective including safety, reliability, quality, etc. along with production.

One best-practice metric for manufacturing lines is Overall Equipment Effectiveness (OEE), which is the product of

• availability: uptime within the line’s control,
• quality: percent good product, and
• production rate as a percentage of ideal (or best proven) rate.

One advantage (depending on your perspective) of OEE is that problems are tough to hide. You can’t slow the machine down to improve uptime and expect to get away with it. You can’t slack on quality in order to improve rates. Any production problem will ultimately show up in OEE numbers.

OEE and Other Operating Metrics

There are many variations on operations metrics, but if you have the capability to measure OEE, you can learn from looking at other factors as well.

OEE can be seen in the context of many operational metrics in the diagram below. Definitions follow.

• Asset Utilization: % of ideal operating rate achieved. Would typically be used by business management, not operators, due to effect of market demand, which is mostly outside of their control.
• OEE: asset utilization plus market losses. By adding market losses back into utilization, certain forces beyond the control of operators and maintenance employees are backed out. Therefore, OEE will be meaningful to a wide variety of operations, maintenance, and planning employees.
• Quality Utilization: OEE plus changeover losses. Backing out changeover losses yields a metrics that considers downtime, production rate losses, and quality losses. Since retooling and changeovers are not considered, quality utilization is a measure of steady-state operations efficiency. If there are frequent changeovers, this will not impact quality utilization.
• Potential Rate Utilization: Quality utilization plus quality losses. Backing out quality losses measures only downtime and production rate losses. This metric has limited application, but might be used if systems to measure quality defects are not in place.
• Asset Availability: Potential rate utilization plus production rate losses. By subtracting only downtime (both scheduled and unscheduled), asset availability shows how much the equipment was available for production. This metric is commonly used as a measure of reliability.

Steps to Measure OEE

1. Select system boundaries. Managers need to clearly what lies within the responsibility of their areas.
2. Define the output. For the sake of calculating OEE, inputs do not need to be known. Therefore, wasted energy, materials, and labor do not factor into OEE. If there are multiple outputs that can not be consolidated, then there will be more than one OEE number.
3. Decide on loss subcategories. There are always three categories: downtime, rate reductions, and quality defects. In the example below, downtime is split into changeover time, scheduled downtime, and unscheduled downtime. Process rate and quality issues are not subdivided.  Depending upon the situation, management might want to split the “due to”s differently. Losses of any kind might be allocated to internal or external causes. Scheduled operator breaks might get its own category. It’s up to the management team to decide what is most useful and informative.
4. Record losses. Record and track the data.
5. Calculate metrics. As the data begins coming in, compile operational metrics on a daily, weekly, monthly, or other time basis.
6. Benchmark. A world-class facility is supposed to be around 95% OEE. Where do your assets stack up?
7. Improve. This is the most important step. Use the data to create a Pareto chart of operational losses and fix the problems.

Example Calculation

Widget, Inc.’s B line produced 1274 widgets in one day. Due to a lack of demand B line was only scheduled for one 12-hour shift. A lot of 75 widgets was found to be defective. Ideally, 200 widgets per hour are produced.

The shift log shows that 2.55 hours were down for scheduled breaks and a planned repair. 45 minutes down was caused by an unexpected actuator jam. 1.12 hours were used to change the size of the widgets being produced.

In the following table standard Excel cell formats are used: for input cells and for output cells. Calculated values are explained below.

Initial Calculations

•  Actual Production Rate = (Total Units Produced) / (Running Hours)
• Process Rate Loss = (Ideal – Actual Production Rates) / (Ideal Production Rate)
• Quality Losses = (Defective Units Produced) / (Total Units Produced)
• Nonrunning Hours = Sum of Downtime Losses
• Demand Hours = 24 – (No Demand Downtime)
• Running Hours = 24 – (Nonrunning Hours)

Calculated Losses

• Downtime Losses are calculated as hours down divided by the relevant time period, which is either 24 hours or Demand Hours.
• Rate and Quality Losses are equal to the given percentage times running hours divided by the relevant time period.

Calculated Operations Metrics

Operations metrics are calculated as described above using the given numbers.

Plotting Losses

The losses in the example can also be plotted to show the relative impact of each loss type. A waterfall chart based on both 24 hours and demand hours would look like this:

Challenges in Measuring OEE

1. Recordkeeping. OEE requires that all losses be recorded and correctly allocated. Depending on the complexity of the process, automated systems can help, but employees may need some convincing that correctly allocated losses will not simply become a stick to beat them with. The emphasis has to be on improving, not on placing blame. Once a system is in place, operators must be accountable for complete and accurate records.
2. Subdividing Categories. Too many loss subcategories creates confusion. Too few subcategories makes data interpretation difficult. Dividing the loss types into categories should be done based on need.
3. Analysis Not Performed. Data collection is a means, not an end. If the data is not used to set priorities and drive decisions, then it is wasted effort.