Know-How Exchange : Research/Metrics

I would let the boss have his say, but describe to him /her what you have found in your research. If it fails -- you did everything in your power to provide them with the knowledge of the best practice.

There's no magic behind a 95% confidence level. No one says you have to have that in all circumstances. You set the sample size based on the value of the information you retrieve from the test or survey and your plans for the information that results. 500 versus 1000 results in a lower confidence, yes. Maybe it's 80%. But, if the use for the information is being combined with other less confident information or the decision to launch the product has already been made because of other reasons, or a host of other points, then 500 for 80% confidence may be all that's needed.

So, in answer to your question, what would I do? I'd evaluate the situation and determine if the cost of an additional 500 samples outweighs the benefit of the additional confidence. If it does, I'd do the 500. If it doesn't, I'd make sure that the "boss" understands the opportunity he's losing by not investing and when he assures me he does, do it.

I hope this helps.

Wayde

Thanks for reminding me, Jeff. You have three dimensions involved here: Sample size, margin of error, and confidence. If you reduce sample size, you can increase your margin of error or reduce your confidence. Or both, actually. And both confidence and margin of error levels versus sample size trade-offs are still based on the value of the information versus the cost of sampling in money and time.

Wayde

Absolutely. I'd also notate the results to indicate the correct sample size should be 1000.

That is unless my job depended on the results. Other than that, I work for him/her.

Michael

You can take your savvy brain power, start your own business, and command your boss to use the sponge to dry his tears when you leave!

Much love, feeling silly today, no disrespect to your situation intended! (Wink)