Best practice to ensure inputting of data in production

[qualprod]

Hello everyone.

I´m in the phase of implementing a process, where data will be necessary to be inputted in computers by operators.
The Issue here is how reliable will be the data? how can I be sure is done in the right manner, right quantity, at the right time?.
The data to be inputted is: time spent in machines in manufacturing of several activities, compliments of work orders, % progress of jobs, # defects, name of operators,
time/hours they enter in production, time they left, time down of machines, time spent in maintenance, and so on.
The idea is to have a base of all this data, in order to make quotations for customers, and based on this, calculating the more precise profits.
Question, in your experience; is this approach reliable for the purpose?, because I doubt operators will input the right data.
or what have you done to ensure this? maybe with a "police" behind them, on their shoulders?
The pros on this approach, I guess is the the data inputted is "live" updated in real time, but the cons, I think is the reliability and the time spent by operators to fill out the data .
On the other hand I think in a scenario, where with some frequency ( during 1 month, and three times a year ) to capture this data, have this data analized and have
information as a base for quotations.
Pros, less time spent by operators at registering data, thus more productivity, cons, no data in real time, no too much data to be based on.
At inputting data, we all know that at minimum, will have to register some basic data, but in the first approach, the idea is to capture too much information as
to have more data to have more precise information to take decisions.
Please feed me back of your thoughts about this.

Thank you so much.

 

[Jen Kirley]

What I have seen for this is production software that intakes bar code scans for each process step, first at its beginning and then at its end, then figures the time and renders calculations for production optimization value.

An alternative might be to have the production personnel manually input start and finish times for their process steps into an Excel form, but I would absolutely not ask them to do anything more than that so as to ensure consistency in the calculations.

Do you already have the software or Excel template? I saw a great many offers in a Google search, but could not be sure that what I saw was aafe for download or just a commercial offering. I would have preferred to see something from a university Industrial Engineering program, but did not see any yet.

 

[toniriazor]

What the data will be used for ? That's the most important question. Why you will rely on it ? Depends on those factors there are several approaches to improve data input by operators.

 

[Ninja]

When I did this in a manufacturing group, I built the system 3 times.

First time, what I thought would work...then had one single person I trusted use it and comment.
Second time was based on that input (with some MASSIVE changes based on ease of use).
Third time was after explaining how this capturing of data would directly affect profitability...and thus directly affect the chance of larger raises next year. This final build was based on feedback from the whole group...how to make it easier, faster, less intrusive.

Most start and stop times turned into a single button click.
At least 5 additional computers were added for ease of access.
After using it for a month, only one person struggled to keep the data up to date...and we had a good chat about it one on one.
That system is still in use...

Every piece of data you can automate capture of will be immensely helpful.

 

[silentmonkey]

One of the biggest peeves of data analysis is dirty data i.e. data that isn't uniform. Having 1 or 2 dedicated people collecting this data could help keep the data uniform provided they are well-trained; a work instruction or process document dedicated to data collection would also help. Otherwise, you could use software such as Excel to restrict the type of data input into the system.

It would help the operators if the collection of data was incorporated into the Production process flow. Make it a natural part of the process. As [Jen Kirley] mentioned, using bar code scanners at the start and end of each process step can help capture the data as part of the workflow without operators having to stop work and input data into a computer system manually. This applies when dealing with defects and rework if items are scanned in and out of quarantine or your rework area at each stage. Similarly, giving operators an ID tag for them to scan in and out of the Production area, before and after using machinery will give you all the necessary data without having your operators sit in front of the computer entering the data.

You would make sure each scan is time-stamped so you can analyse the data later.

 

[optomist1]

"Every piece of data you can automate capture of will be immensely helpful." Ninja is spot on, whether data input or the actual manufacturing process, where ever possible minimize human input, or touch labor as some call it....i.e. use a DC nut driver for all fasteners...automatically captures torque and angle data....same can be said for AOI...

 

[Techster]

We are a machine shop who are using bar code scanners in our production to have the amount of time spent for a specific activity on the machine run time, idle time, different kind of alarms, tool change, measuring time. It is all collected by Cimco and put in to a calculation too see if we earn money, but also too see if the we are improving.
It has taken us minimum one man in one year to make the system work plus the money paid to Cimco to make adjustments.
But now it is very beneficial for us to see the things we do either doesn't work or improve the total economy on the product.

Prior to this system we only knew the cycle time and the amount of parts we needed to manufacture.

Maybe you don't need the data available too other systems right now, but it could improve a lot of other system to have this data live and avalaible.
As an example your planning could come a lot more reliable.

 

[qualprod]

We are a machine shop who are using bar code scanners in our production to have the amount of time spent for a specific activity on the machine run time, idle time, different kind of alarms, tool change, measuring time. It is all collected by Cimco and put in to a calculation too see if we earn money, but also too see if the we are improving.
It has taken us minimum one man in one year to make the system work plus the money paid to Cimco to make adjustments.
But now it is very beneficial for us to see the things we do either doesn't work or improve the total economy on the product.

Prior to this system we only knew the cycle time and the amount of parts we needed to manufacture.

Maybe you don't need the data available too other systems right now, but it could improve a lot of other system to have this data live and avalaible.
As an example your planning could come a lot more reliable.

Thanks techster, and others who shared Comments

 

[Silex7]

I understand deeply your point. I had a similar situation when I wanted operators at the production floor to register data for maintenance record with a plenty of details in order to monitor the machines breakdown and provide effective corrective actions. I made simple paper check sheet and instructed everyone about filling the form , I can tell you they stick to this sheet nearly 1 week , but after that, no one was actually filling it.

After this miserable failure, the lessons I've learned are:
#1. Automate the process (if possible)
#2. Use the least possible data reporting method to receive your required data
So,
I would say like other fellows pointed out here, go for a more 'auto procedures' and less 'human activities'. If you can not automate the process , make the least and the easiest possible way to obtain your data , and cooperate with their supervisors to ensure the reporting procedure as a part of their daily task operation ( you can say something like one shift can not hand to the other one, until they confirm and deliver their reporting activity). This can not be ensure data delivered is 100% true and accurate, but with a short term site patrol ( mentioning objective evidence found: M/C#... Operator name#:.... not following the guidelines rules,..etc.) they will start strictly to stick to the data , and then the task will be -unconsciously- a part of their regular daily activities.

 

[Steve Prevette]

There are some statistical tools that can help. From a data falsification/error detection perspective, there are tools like the Benford distribution that can detect unusual results Benford's law - Wikipedia

Another good tool, both from what to do with the data, and detecting data problems is to analyze the data in Statistical Process Control. If you get an unusual result, that can be flagged for investigation. One possible answer is data error. Dr. Deming relates the story of a shoe factory with high percent defects. But when plotting the defect rates on a p-chart control chart, it was obvious something was wrong with the data (hugging the average). Come to find out the inspector was falsifying the data to protect people's jobs. If the signal is real, it gives you a chance to determine what happened and how to correct it (if bad) and how to exploit it (if good).