Fao Factor Or Anyone With A Good Working Knowledge Of Statistics.

[DTCwordfan]

Problem, client with approximately 6000 product/size/colour combinations, each necessitating a product line for a retail site.

To run a beta test of the site, how many live transactions would be required to reach, say, a 95% or 98% confidence level as to errors?

Many thanks - DTC

[factor-fiction]

Hi DTC- I've only just got back from work in the last 30 mins and will be off for a bath and bed ready for early start tomorrow. I'll have a look tomorrow but I think the sample size needed will depend on the expected number of errors and the variance, so data may need to be collected on these first.
There are some who are more up-to-date on things like this -jim360 etc-who may see it first

[jim360]

Thanks for the vote of confidence FF but I'm not really familiar with this particular problem so wouldn't know where to start... might need more information e.g. what constitutes an "error" in this context, and similarly what is a "success"?

[DTCwordfan]

thanks guys, success would be a successful transaction, failure would be the incompleteness of the order owing to a dud product line, the ingredients being product, quantity, perhaps colour, perhaps size, weight and price.

[jim360]

Hmmm... that's still too vague for my liking. It sounds, though, not unlike most experiences at online stores I have seen, you can apply filters to item choices in a number of ways by selecting a choice or a selection of choices out of this labels e.g. colour to the products in question. In which case this presumably amounts to a computing problem, and success or not would depend on the particular code and search/ sorting algorithms used. In principle if you were able to find an existing code that works, and understood how to populate the database correctly, then you wouldn't need much testing at all. Starting from scratch, on the other hand, and it could take a few trials to iron out bugs and such.

So I suppose the answer to your question is, "It depends on how you're doing it."

[DTCwordfan]

Typical transaction would be for 5 items (product/size/colour etc), range from 1 to 10.

[DTCwordfan]

Thanks Jim.....yes, it's a question of choose the item, then choose the size, and colour - then into a shopping bag......the weight being calculated for postage/Fedex purposes behind scenes.

I also want to see how the order forms appear to check that they are legible as tothe manufacturing side of boards and muffin cases and then the warehouse for all the accessories that they buy in. If you send me a throw-away, I can let you see the site before the trigger is pulled. It's quite a story in getting them sorted out!

My 'guesstimate' is that some 35 transactions of 5 to 6 items should give a pretty good live test. The family are going to do that with the owner crediting their bank accounts back - one way of raising funds!!!

[factor-fiction]

In that case I suggest you do an initial trial of maybe 30 and then give us info on the number of errors in each order. That way we can get an idea of the mean and variance for errors in the distribution. The normal distribtion approach to sampling could then be tried

[DTCwordfan]

Good idea and thanks, factor; that makes an awful lot of sense.