Guest entry from one of our consultants:
"I’ve recently been following the Elsmar.com Forum for manufacturing improvement and there’s a post on there asking what to consider for an OEE data capture system. Having spent some time responding I’ve decided to be lean and use the same content on my blog! Let me know what you think!
Oh a quick thought (I’m writing this after finishing everything else!) – here’s an idea of what an OEE system should be: “A good data capture system is simply a robust resource allocation tool”. Whatever you do it should lead directly to people doing something differently as a result of using the data. Now you can probably ignore everything else I’ve written below…and feel free to read on!
When choosing OEE system, my advice would be in the following areas:
1. Identify where the constraint is in your process
2. Identify the measure that not only tells you the extent of the constraint, but what the contributing factors to loss are
3. Understand the metric, automate it and train people thoroughly. Get their buy in and support.
4. Establish a robust management review methodology based on the metric – hence the need for automation; an automated process frees up your management team to fix the losses not spend all their time calculating them.
1. Identify the constraint:
If i may offer some advice it would be to help identify what your objective for recording the data might be, when choosing an OEE system. Whatever you choose to record i would recommend that it measures the constraint in your manufacturing process. If your constraint is a mechanical one – i.e. to produce more produce you just need to run a machine more/get less stops then OEE is a great measure for you. If however, your constraint is relieved by hiring more people, or more generally your constraint is labour based then i would steer you towards more of a man hour / tonne type metric. By the way an XL800 System can measure both very easily – how would you like a real time £/tonne measure displayed on the factory floor?
2. The correct measure:
The value to OEE is not that you get an OEE number. You’ve made what you’ve made – there’s little point in reviewing it. I visit so many sites that can tell me their OEE but can’t tell me where their losses are. The value to measuring OEE is in the categorisation of loss. If you know the loss you can apply the right tool to fix.
E.g. for OEE:
3 Loss: Quality, Performance, Availability
6 Loss: Speed, minor stops, major stops, quality in process, quality on startup, planned downtime.
If you have a planned downtime loss apply smed techniques. If you have a minor stop loss apply kaizan blitz techniques. The value is not the OEE number – it’s the collection and categorisation of the loss that counts!
3. Understand and train:
There are 2 main schools of thought on the collection of OEE data. One is the manual school that says it’s better for operators to collect so that they understand and you get the ‘real losses’. The other is the automated school that says it’s better to get the correct data and then work out the real losses later.
An automated system only ever tells you symptoms for your downtime – its diagnostics are only so good as the signals you give it. That said, how often has an operator identified the real root cause for a stop – 9.9 times out of 10 they’ll note down a symptom.
My personal belief is that it’s better to use an automated system that captures your losses accurately…and then use management process and review to drill down. Manual data systems like the ones you’ve listed take a lot of work to maintain and i have yet to find one that’s accurate to >60% simply due to the nature of human data capture.
Therefore i sit firmly in the second school and am capable of installing a system that correctly identifies loss on canning/bottling/packing lines running at >30,000units/hour to accuracies of over 98%.
The message here is again back to objectives – what do you want to achieve. Whichever root you choose involve your teams. I remember installing a system in a bottling plant in which the operator came up to me and said “this is crap, all crap. I’ll show you – i’ll find all the problems with your system”. It was brilliant! He was the best snagger i’ve ever met. I just took all his feedback and fixed it all. After a month he had to admit the system was good because he’d commissioned it for me! Now stop him from training everyone else on ‘his system’!
4. Robust process:
A system is ONLY so good as how it’s used. You could spend £10 or £180,000 on a system. If you don’t use it you’ll get the same result and the same payback.
We often support people to establish robust internal processes for using the data, documenting the actions that arise from interrogating the data, and then a management review process for driving change. Typically this looks as follows:
1. 24hr daily reviews – looking at 24hr data. Objective is to identify what actions are still open, see if we have any reoccurring issues, assign resource where needed.
2. 4hr Short Interval Control – a regular review (initially at 4 hr intervals, moving to 2) with front line management and engineers. 1st objective is to identify greatest loss from last window of time and ensure closed off. 2nd objective is to identify what needs to be done differently in the next window of time based on the data currently available.
3. A variety of strategic review – looking at trended data for maintenance, engineering, planning, forecasting etc.
Remember – your level of payback is directly related to how you use the data.
So what OEE system works? Whichever one you commit to using fully. My advice would be to look beyond a software download as I think you may struggle to use it fully simply as it’s reliant on manual data collection. That said, I’ve created and implemented several of these in the past on sites performing at between 25-40% OEE and got great results.
Also if you’re running at <45% my belief is that you don’t need to spend a fortune collecting data because you’re line is down for 4 hours in every 8. People know where the issues are because they’re in them! But when you head up into the 65%+ territory you’ll struggle to continue improving without robust automated data capture."