» » next post - David Meier: Good units produced (total parts – scrap) / (Available work hours – wait Kanban) = GPPH (good parts per hour)
« « previous post - Art Smalley: Productivity and Improvement
Michael Balle

Michael Ballé: Who needs to use the metric and to what purpose?

By Michael Balle, co-author of The Gold Mine and The Lean Manager - Last updated: Wednesday, January 30, 2013 - Save & Share - Leave a comment

There are two ways to read metrics:  one, to drive behavior, the other to better understand a problem – or both. Taylorist thinking is deeply ingrained in all our mindsets, and the usual fallback for any desired outcome is to slap an indicator-and-incentive on it. This usually works, but at the price of unexpected side-effects, which can often negate the very impact one sought. Metric improvement behavior is well studied, and if the reward is relevant enough, we now know humans will 1) do whatever they can to get the prize, 2) at the expense of all other variables, and more likely than not, the very thing we were seeking in the first place. Try to reward your kids with pocket money every time they bring home a good mark at math tests, and they will, but at the expense of any interest in mathematics.

Productivity measures are very prone to this sort of fallout. Incentivize your managers for parts produced by labor hour spent, and they’ll churn out products whether you need them or not, and at dubious quality. Conversely, reward front-line managers for producing exactly at takt time, and they’ll add extra labor and feather the nest to be sure to deal with any variation in the work. Incentivize gains in real hours against standard hours, and then people stat mucking about “real” hours (do we count material handlers, team leaders, in or out?) as well as “standard” hours  (whose standard? Before or after improvement? etc.)

On the other hand, not measuring such things is probably worse, as no one will have a clue on what productivity problems we actually face. The other way of looking at metrics is as a tool to better understand a problem – as a gap between a desired situation and the current one – and thus scope countermeasures in order to solve the problem. In this perspective, the question is not about what “lean” metrics to use, but what problem needs solving at which level.

Recently, I’ve had an on-going argument with the CEO of an engineering company about measuring sales/person. He kept arguing – correctly – that there is no known smart way to measure engineers’ productivity (how do you measure one person’s ability to find the solution that benefits the greater number of people quickly)? In the end, started tracking sales/person, and was very pleased to see the number steadily improving. He was very surprised when I asked him whether increasing sales/person  beyond a certain level was such a good idea: weren’t we overburdening engineers? We then looked at response-to-customer lead-time that was also creeping up. He suddenly saw that he was probably selling faster than his engineers could reasonably cope with, and would have to hire. In that sense, the productivity question is: what is the productivity standard we’re seeking?

Any metric should never be looked at on its own. Inventory turns or days should be looked at in the light of on-time delivery. Parts per hour per person, in the light of defective parts per million, and so on. Metrics as incentive make sense when a group of people is incentivized on a set of targets – which minimizes game playing within the team and being too focused on one or the other metric.

At management level, I tend to work with business metrics such as sales/person or value added per person when there is a very large purchasing part, and so on. The metric should be crafted according to the problem we’re trying to solve, not in absolutes.

At the cell level, I tend to work with parts per hour per person, because this is something supervisors usually find easy to calculate and quite intuitive to understand, although it is very sensitive to mix in work content of the products. At cell level, it usually works ok since products being built on the same cell tend to have roughly the same order of work content, but this is not always true. When there is a large variation of work content in the same cell we tend to revert to real hours versus standard hours – which is unsatisfactory, but best we can do. Whatever the calculation, the key thing is that the supervisor understands the ins and outs of how to compute it, understands the underlying outcome we’re trying to create, and takes responsibility for his or her own metric – identifying kaizen opportunities to improve it.

Contrarily to the other key metrics such as lost time accidents, accident frequency rates, customer complaints, defective parts per million, days of inventory and so on, productivity metrics are particularly troublesome because the almost never work both at cell level and at management level. For instance, parts per hour per person makes sense at the cell (once you’ve established the calculation rules such as counting the team leader but not the train operator), but it’s much harder to interpret at plant level due to mix issues. On the other hand, sales per person or sales per square meter make good sense at plant level and can be used to compare either the same activity or different activities, but doesn’t make much sense at the cell level.

Reporting systems were imagined to feed shop-floor data to senior managers in order to improve their decision-making. In many cases, this works well, such as with quality indicators. In other cases, it works poorly such as productivity and throughput measures. I haven’t yet come across with a satisfactory answer to that question other than:

  1. Be clear on who needs to use the metric
  2. Never use a metric on its own
  3. Progressively build a standard calculation clear to the users
  4. Never incentivize specific metrics without expecting major side-effects.

Post to Twitter

Share this post...Tweet about this on TwitterShare on Google+Share on LinkedInBuffer this pageShare on FacebookEmail this to someonePin on PinterestShare on Tumblr
Posted in Uncategorized • Tags: , , , Top Of Page

Write a comment

*