Long example but hope it helps (I have a TON of these examples of the power of observation by the way so this isn’t some isolated once in a decade occurence)
Here’s an example of the power of observation: at a moderate volume, very high mix electronic board assembly line a new board was experiencing a high level of ‘cracked’ through hole connectors, discovered after final test. The failure rate was concerning as it involved a new connector version that was going into high volume use on many of the boards we were producing. (This was a prototype line from final Customer testing and manufacturing capability testing prior to very high volume manufacture over seas.). The Problem was assigned to an engineer in my group. After a few weeks he had made no progress. I sat with him in his cubicle and discussed the Problem.
He had talked to the supplier’s product engineer who gave him several potential causes. 1) aggressive insertion and extraction of cables form the connector during final test, 2) aggressive handling fo boards as they are transported form operation to operation, 3) aggressive handling at soldering operation, 4) excessive heat at soldering, 5) excessive thermal cycling at soldering 6) aggressive handling as inventory is distributed to soldering…basically blame the Customer.
The engineer had decided to run a full factorial to determine the ‘cause’. (They had just taken a DoE class and gotten some statistical software they were eager to use) This particular experiment had several factors at the soldering station and if nothing was discovered then the engineer would look at the aggressive handling ideas. I informed the engineer that the experiment would take awhile to run because production was behind int heir build plans and interfering with production would be difficult. I then asked if he had observed the process of cable insertion & extraction and looked at the failed components and non-failed components to see if there was any pattern. The average failure rate was about 25% so it wouldn’t take long to observe the failures. The engineer said they hadn’t done that yet as they were busy with the Supplier’s engineer and the design of the DoE. About this time several other engineers had poked their heads over their cubicle walls as they knew what was coming.
I metaphorically grabbed the engineer by the ear and dragged him down to the floor. I first took him over to the pile of cracked connectors and told him to LK at the connectors and see if he could see anything non-random. After a couple of minutes he asked what the ‘B’ was on the connector. I told him that it was a mold mark identifying which cavity the part came from…He then said that all of the parts in the pile had a ‘B’ on them. We couldn’t see the mold mark on the boards as it was on the bottom of the connector so I took him to the inventory location: there were 4 cavities A, B, C, and D.
Now what are the odds that the temperature variations and aggressive behavior would only effect the B connector? Well, zero of course. A second look at the cracked connectors showed that the crack always occurred in the same corner. Subsequent measurements of the thickness of each corner on all cavity parts showed that the cracked corner on B was very thin compared to the other corners; the parts actually straddled the minimum spec for the corners.
Staring at computer screens is not the way we solve problems…
Full factorial DoEs are great. But where do you get the factors? Where do you get the levels? Observation is the best ‘clue generator’. And there are a lot more study designs that are great at narrowing down causal factors. Data is great, BUT we need to look at tit real time, over a long enough period (the brief look you posted isn’t helpful) and the data collection must be Appropriate…you have yet to tell us anything about that.
