What has been your experince with the quality of data you receive from people on the front lines of your organization?
Front lines being technicians or production, although it can apply to anyone.

I have had some very interesting experiences:

A) Data includes many nonsensical points
1) Way out of range (some not even possible)
2) Not the right type (attribute where you want variable, etc.)

B) I'll give you want I think you want
1) When given a supposed 1 gm wt. all readings recieved were 1 gm on the nose.
2) All points in range when you know that is not the case.

Yes, training is a big factor, as is time, no one caring etc.

Please share your examples, thoughts and how you conqured the problem.

We had similiar issues were I work. To correct this we had the data entrys that had errors put under the control of the Quality Department. Make sure your Quality Department does not report to management of the production area. Also we set up electronic forms (Infopath) that highlghts and will not allow bad data to be ebtered until corrected. (You may be able to this with access and excel also I am not sure). These actions have seemed to resolve the errorneous data for now. Also our employee manual states anyone found to be falsifying data can and will be terminated.

HTH,
Matt

I have generally found "as found" frontline data to be similar - of poor quality.

The problem is that there generally is a self fulfilling prophecy - no one is looking at the data, so no one cares about the data. When threatened with data analysis, the data owner usually states "give me a year to get the data cleaned up" - and you go back a year later and nothing has changed.

Best thing I have found to do is to analyze the "as found" data in a competent manner (for me, that implies SPC). Then show the data to the owners, warts and all, but in as non-threatening a manner as possible. Once the data is out there for folks to see, it will generally "cry out for action". And one of the first actions may be that the data is giving signals due to errors in the data.

Me too, Steve.

All good advice!