# Useful statistical functions for business — for use by a newbie

I am a database developer working for a sales and manufacturing business. I am mostly ignorant about statistics. We need useful metrics. Our managers are tuned to accounting, and the vagaries of production tend to confound us. We do very little measuring of our production, and what we have is poorly formed. I should note we are a "job shop", not a "flow shop" -- we do a lot of engineer-to-order work, so the usual MRP standards are often hard to apply.

Some tradition business metrics are known to us; for example "inventory turn-over rate." But we are unable to convert those to useful information. I believe our inability to qualify data statistically is a big reason why.

Of course we perform averaging all the time. Rolling averages (smoothing data over time using a 3-week rolling average, for example) is a helpful extension. Recently I discovered how to apply standard deviation to labor costing with wonderful benefits.

Now that I understand (A) averaging, (B) rolling averages, and (C) standard deviation, what are the next useful functions or functions I should seek to learn?

I would love to have your insights on "business intelligence," I mean defining and using metrics. But it's the use of stats to get from raw data to usable information that I'm really after. No matter, give me whatever you've got.

• You could learn lot of things but for useful suggestions you should ideally let us know what kinds of questions you like to ask and what kind of answers do you want to those questions. Some exemplars would really help as otherwise your question is too broad. – user28 Aug 10 '10 at 22:57
• My thought was that the tools I'm seeking are generic. Standard Deviation is a generic tool. What's another one like it? The basic goal is to remove noise from data. A bit later I will provide more about why Inventory Turn Rate is difficult for us and why I think more stats knowledge would help. – Smandoli Aug 10 '10 at 23:15
• After learning the basics, it would be quite helpful if you use some tool i.e. Matlab, R, Excel to view your results in a more human understandable format. – DumbCoder Aug 11 '10 at 12:01

There are lots of generic tools. You should probably start from these:

Foundational items to understand statistics:

This should probably keep you busy depending on how much background you already have.

• Linear Regression is precisely what I'm looking for to deal with one problem (forecasting sales from current customer inquiries). It looks difficult, but I will read more. I can't tell about Comparing Means. Thank you. – Smandoli Aug 10 '10 at 23:52
• "Forecasting sales from customer inquiries" is the type of example that you should have mentioned in the question as that enables a better response. Another tool that helps in this context is correlation coefficient (See: en.wikipedia.org/wiki/…). This concept is closely related to linear regression. – user28 Aug 11 '10 at 0:37
• Likewise, you should indicate whether you are interested in individual sales--will this customer buy--or aggregate sales--how much will we sell this quarter--as those will require different models. – Brett Aug 11 '10 at 3:22

I would suggest, if you can afford the time, to follow two online (taped) courses, one in probability and another one in statistics. I think it is the best way to get some basic knowledge that will help you move forward.

• Probability course
• Statistics course (This is lesson one. It goes up to Lesson 64. Unfortunately, I haven't found the links gathered together in one page. I think also that lessons 21,26 and 39 are missing but I don't remember having any problem whatsoever in following the course). You can also find the handouts for this class posted here.

I'm not sure if you're still around looking at this post, but I'd like to reiterate that regression could be a very useful tool for you. Regression is a way to understand how one or many variables influence another, and so it has tons of applications in business. I work at an online retailer, and we're always using regression (sometimes logistic, sometimes linear) to predict all kinds of things, from response rates of catalogs to open rates of emails to average sales per customer. The other cool thing about regression is that you can build fairly complex models by transforming variables (i.e. looking at sales^2 instead of just sales), and "multiplying" variables (also called cross-terms). Shoot me an email if you're still around and you need help on this, I could give you some good example models and help you understand what may be useful for your business.

2 great answers so far and I promise I'll accept one of them. But I wanted to post this link as an answer too: summary of how to compare two numbers

Yes, for me merely comparing two numbers had some questions, so that shows you the level I'm at. While it won't be relevant to most, here is the VBA function I made based on that info ... well, not based that info exactly, I needed something simpler:

Private Function fxDifference(sng_A As Single, sng_B As Single) As Variant
''When calling this function, sng_B must not be zero (Error div/zero)
fxDifference = (sng_A - sng_B)
fxDifference = fxDifference / sng_B
fxDifference = Round(fxDifference * 100, 1)
End Function


I suppose when I said 'generic' tools, I meant 'really, really basic' ...

• PS -- the real reason I included the code is to prompt responses from anyone in the programming realm. This function is probably found (and more elegantly) in some 'VBA stats snippets' resource ... I should search on that. – Smandoli Aug 16 '10 at 17:18