Effectiveness Checks - How Frequently Do You Check?

[Zero_yield]

Hello everyone,

The organization I work for often defaults to "no recurrences in 12 months" as an effectiveness check (and it's mostly a reasonable path to follow). However, that's a long time period to be monitoring a process. We can and do sometimes close ECs early if we have a recurrence, but we don't necessarily have a process beyond our procedure saying that we should.

How often do you check whether your CAPAs were effective?

Some possibilities:
1. Whenever closing a deviation, check it against open ECs.
2. Periodically review open ECs against open nonconformances on some sort of set interval (monthly? quarterly?).
3. Require EC owners to monitor their affected processes and report on them at some sort of set interval.
4. Some other monitoring process?

Thanks,
Zero Yield

 

[Tagin]

Typically, I would base effectiveness duration on activity frequency or rate, not on calendar dates alone.

For example, let's say the CAPA was about a defective product from a supplier. If we typically receive 5 batches of that product per month, I might set effectiveness for 10 batches received, so the duration ends up being roughly 2 months. But if we typically receive only 1 batch ever other month, I might choose 3 batches, with a corresponding timeframe of ~6mo.

One can imagine similarly that effectiveness durations for processes related to produced product, or other internal processes, can be guided by the frequency or rate of those processes.

This is not exactly scientific, and perhaps there are statistical guidelines that make such decisions more justifiable, but I imagine such statistics would still be based on speculative rates of recurrence.

Edit: Criticality would also factor into effectiveness duration decisions.

 

[Zero_yield]

I didn't get into it for the first post, but the organization I work for has a pretty mature quality system. Our nonconformances are usually something that's showing up on < 1 in 1,000,000 parts, or a glitch that only occurred once on a piece of equipment in >10 years. It's hard to even say "we're confident this action will eliminate / reduce the probability of recurrence" when the rate is so low to begin with, which usually leads to extended EC times. We want there to be at least some chance of a recurrence happening.

 

[Tagin]

I didn't get into it for the first post, but the organization I work for has a pretty mature quality system. Our nonconformances are usually something that's showing up on < 1 in 1,000,000 parts, or a glitch that only occurred once on a piece of equipment in >10 years. It's hard to even say "we're confident this action will eliminate / reduce the probability of recurrence" when the rate is so low to begin with, which usually leads to extended EC times. We want there to be at least some chance of a recurrence happening.

Interesting scenario. I wonder if you can use your historical information to assist in determining EC times. If you have an NC that occurs 1 in 1,000,000 parts, then how many of these parts do you use/produce in 12mo? If, say, you produce 10,000 in 12mo, then wouldn't you have to have an EC duration of 100 years? Of course, the idea is actually that you gain some confidence as your sample size grows, so you would not have to wait 100 years. But I'm not very good at statistics, so I don't know what that "confidence" curve (or line) might look like. But perhaps this could inform the extent to which your EC durations are helpful or not.

 

[Zero_yield]

Interesting scenario. I wonder if you can use your historical information to assist in determining EC times. If you have an NC that occurs 1 in 1,000,000 parts, then how many of these parts do you use/produce in 12mo? If, say, you produce 10,000 in 12mo, then wouldn't you have to have an EC duration of 100 years? Of course, the idea is actually that you gain some confidence as your sample size grows, so you would not have to wait 100 years. But I'm not very good at statistics, so I don't know what that "confidence" curve (or line) might look like. But perhaps this could inform the extent to which your EC durations are helpful or not.

For one of our product lines, we produce up to a million a month, so the infrequency of the defect is still manageable.

For other issues, I've investigated nonconformances where every SME on the product line (some of them having been with the company for 30+ years) has never seen it before. Those usually crop up around a change to the process (so we can estimate the EC around similar process changes), or around a manual process.