Speak With Data

I worked with a group of ten women who worked on an assembly line and had been given the opportunity to meet weekly to discuss, with their supervisor, what could be done to improve their assembly line.

The product that they assembled was a model of a remotely controllable, outside rear view mirror.  The control was a mechanical device in which three wires moved the mirror when a knob was rotated. The ends of the wires had ferrules that were crimped into the knob assembly by a crimping machine.

There was a consensus about the biggest problem.  The crimping machine often broke down, bringing the line to a halt, sometimes several times a day.  When the line stopped there was a wait for maintenance to arrive and tinker with the machine.  The team took pride in reaching its production targets, and this machine interfered with that goal far too often.

The supervisor agreed that this was a problem.  He too was frustrated that the fix was never permanent. He had complained up the line with no results. 

We considered how we might get management’s attention. I proposed that the operators document their down time.  The team was enthusiastic.  A member volunteered to be the recorder.  Every time the line went down the date, time, duration of the delay and the reason the line had stopped would be recorded.  The record was kept for a month.

The data demonstrated what the team already knew. The crimping machine was the biggest source of downtime.  The data also showed the line had lost several person days of productive work in the recorded month.  I helped the line tabulate and summarize the data on a single page.

At lunch, I was sitting next to the Plant Manager when a volunteer from the line suddenly appeared behind him and dropped the report of their downtime over his shoulder.  Almost shouting she said, “That’s what we mean!”  The report very nearly landed in his soup.

The Plant Manager picked up the sheet of paper and examined it.  He thanked the team member and promised to take a careful look at it.  At his staff meeting the following morning he shared the report with everyone, and asked a manufacturing engineer to see what needed to be done to eliminate the problem.  A brand new crimping machine was installed on the line within a week.

There is no guarantee that a data collecting initiative such as this one will lead to similar results, but the probability of getting action will increase.  Curiously, months earlier, a manager had put a log card at the end of each line, in which he asked that all downtime be recorded.  The assembly workers complied for a time, and then gradually abandoned the effort.  There were no benefits to filling out the cards and no consequences for not doing so.

Speaking with data is not the same as collecting data.  Speaking with data is a way to get the process to tell us what we need to know.  In this example, the workers already knew where the major problem was, but management needed to know–to understand—the importance of the problem and what it was costing.  Simply complaining about the machine going down did not accomplish this.

Speaking with data makes the invisible visible.  This is not to say that we should not examine data critically.  Kaoru Ishikawa, a Japanese quality guru is quoted as saying, “When you see data, doubt it!” Sometimes data is collected to please someone (or to get them off our backs), and the data collected loses its meaning.  Statistical process control sheets, with data collected and recorded by operators sometimes contains what the operator thinks management, or the quality department, wants, rather than showing the real variation in the process.

Deming mentions an organization that had not had any injuries in an entire year, yet people on the shop floor had casts, bandages and other evidence of accidents.  It turned out that a bonus was paid for a year without accidents.  Reporting accidents would eliminate the bonus. There were accidents, but none had been reported.

I had the opportunity to be one of the hosts to a visit from Hino, a Japanese truck manufacturer, when I worked at Cummins Engine.  Over lunch the first day, the Cummins Plant Manager asked the visitors, through an interpreter, for their outstanding impressions after the morning tour of the engine plant.

The visitors huddled and reported that one of their impressions was that at Cummins we measured everything.  All sorts of data was collected.  The visitors were even struck by the fact that there were many barrels throughout the plant into which the data was thrown away. (The computer printouts were deposited in the barrels for recycling.)

We do not measure everything, the Hino visitors told us. We take measurements, when we have a problem to solve. When the problem is solved, we stop taking measurements.  We go and solve another problem.

Our visitors where using data to have the process speak to them.  While there are certain data that it makes sense to collect and examine on a regular basis, we often collect and record far more than we can use.  Fortunately, we no longer have to put all of it on paper to dispose of in large recycling barrels, as it can now be accessed on a terminal.

Data can help us communicate effectively with the people who can make things happen. Data can help us understand what is going on in our processes, to make the invisible visible. Data can be misleading.  When we see data, we should doubt them.

Notes:

The guote of Kaoru Ishikawa is in Maasaki Imai, Gemba Kaizen, p. 29.  I am almost certain that Deming’s story is in The New Economics, but I have not located it yet.  I will post an up-date when I find it.