# How metrics like OEE often dictate behavior

By Ron Pereira Published on May 11th, 2008

A reader of the blog asked me the following question via email.  By the way, if you ever have a question please feel free to drop me a line.

I am a production engineer for 3 packaging lines and our overall indicator is overall equipment effectiveness (OEE). Our OEE is calculated using the maximum speed that the line will run for the particular product.

A welcome side effect of some recent machine upgrades is some of our products can now run faster, but due to planning rates (and people factors) this is not always required.

Would you advocate (or is it widespread) increasing line speed but not changing the specified speed to give an overall OEE boost?

Overview of Problem

I am assuming the line speed is being used in the “performance” piece of the OEE calculation. For example, at line speed “fast” we should be able to produce 100 widgets per day. Then, at the end of the day, we count up the widgets and see how many we have. If we have only produced 75 widgets our performance is 75%.

The problem with this, as the questioner alludes to, is that the demand for this particular product may not be 100 per day. It may, for example, only be 60 per day in which case they need stop the line after completing the 60th widget, lest they overproduce (the mother of all wastes).

Stopping at 60 is the right move but the team must now take a hit on their OEE metric since a “fast” line speed means they should be able to produce 100, not 60, in one day.

Recommendations

My suggestion to this individual, and anyone else battling this type of issue, is to remember that metrics like OEE (one of my favorites by the way) should be used to support the business – not drive the business.

Here are my specific recommendations.

1. Calculate takt time. Hypothetically speaking, if there are 8 working hours (28,800 seconds) in the day and we are asked to produce 60 widgets, our takt time is 480 seconds per piece (28,800/60).
2. Calculate the optimal crew size for the assembly line which is simply the sum of manual cycle time for one product divided by takt time. If, for example, the total manual cycle time for one widget is 16 minutes (960 seconds) our crew size is 2 people (960/480).
3. Balance the work so that each operator has no more than 480 seconds (takt time) of work to do at a particular process. This is where line speed enters the picture. The team will need to understand how fast the assembly line should run in order to accommodate this particular pace. A fixed position stop system, similar to the way Toyota operates, may be worth investigating.
4. Set up a hourly production board to track “plan to actual” allowing the team to determine if they are on track or need assistance as the day progresses (before it’s too late).

With all this said, the performance piece of OEE should be “judged” against what the team was supposed to produce on that particular day.

In other words, we set line speed to what we need it to be for that particular day. Some days we may run it faster and some days we may run it slower. In the end, it’s all about what our customer asks of us on that particular day.

Of course, this production team could also decide to study their demand and begin to smooth production using the lean concept of heijunka. This would allow them to stabilize the process and work more consistently day to day.

Would you approach it the same way? Do you have any other advice for this individual?

If you enjoyed this article please consider subscribing to our full feed RSS. You can also subscribe by email and have new articles sent directly to your inbox.

1. #### Paul Cary

May 12, 2008 - 7:35 am

I also think that if demand is less than available capacity, you may have “scheduled downtime” deducted from the overall calculation. This method may work better with a piece of capitalized equipment rather than a production lime.

2. #### Ron Pereira

May 12, 2008 - 9:39 am

Thanks Paul. So you are saying this planned downtime is deducted off the top to ensure they are not penalyzed in their OEE measurement, right?

3. #### Mike Z

May 12, 2008 - 12:08 pm

Ron alluded to the point that the business model should not be driven off of OEE scores. It needs to be said that Lean and Six Sigma provide dozens of measurements that can be used as estimations versus a baseline or target rather than an evaluation criterion or score.

Your reader should be trending the OEE vs. “Required” OEE rather than desired OEE. If used for an evaluation of the line’s productivity, it needs to be measured on a daily basis for a discrete manufacturer. Some days only require 30% yield while others demand a 85% yield. This can support your mixed constraint scheduling process if you’re using one as well. We have instituted a similar scheme by determining staffing levels on a daily basis in assembly based on the “Required OEE” and distributing our flexible staff to areas with a higher “Required OEE” for that day.

4. #### Quinn

May 12, 2008 - 1:43 pm

Perhaps for a future video you could illustrate the principle of heijunka or level loading. Demonstrating how it could be applied to a varied product line of say envelopes stuffed with red, white or blue paper. (envelope size could also be a variable). Just a thought in response to your request for possible future video ideas.

5. #### Ron Pereira

May 12, 2008 - 1:51 pm

Oh… I like that Quinn! Let me see what me and my Pereira Lean Production Team can come up (aka my kids)!

6. #### Paul Cary

May 12, 2008 - 1:55 pm

I have used time card racks for work centers or capital equip, each slot representing a time increment i.e. 1 hr. place the order in the slot and if the estimated time is 3 hrs. leave 2 slots these blank plus the slot would equal 3 hrs. of capacity.

7. #### Lester Sutherland

May 12, 2008 - 2:31 pm

Ron,
I agree that OEE is not the metric to use for most situations. If the equipment was a bottleneck or constraint OEE would make sense. It sounds like the original question was should he increase the performance (speed) indicator to the new higher speed, or be able to run faster and get a “boost” in the measureable because his increased speed 110% would offset some other problem like availability or quality. The old game the numbers issue. I always answer that the best speed is whatever the best achievable speed is…. But don’t use the measureable where it does not make sense.

8. #### Mark Graban

May 13, 2008 - 5:20 am

If you just divide cycle time by takt time, you’re going to need 100% efficiency… not likely even in the best of processes. It’s better to use a factor (I’ve seen 80% or 90% used) to account for “normal” downtime (even in a “lean line”). You need to be able to run faster than takt to make up for downtime you do incur (or be prepared to run overtime to make it up).

9. #### Lester Sutherland

May 13, 2008 - 6:53 am