How to do basic statistical analysis?

[semidevil]

My job requires me to update excel reports and also send out analysis and summaries. My concern is, I don't really have a lot of experience with statistics, numbers analysis, or have enough background to know what type of tools are available for me.

For example, let's say I send out a weekly report that tells how a company is doing in terms of sales. There is the total $$ generated, total items sold, and it can be filtered down to the state level. My ideas of an analysis includes: comparing how this week is compared to last week, this month vs last month, average price, this year vs last year etc etc...and maybe drilling down by state to identify who is doing best or worst.

Are there any other methods of analysis that I can use, just so I have more options in my review? I know i"m being vague, and it is because I'm a complete newbie and looking for general advice and ideas.

 

[Simon Bridge]

Are there any other methods of analysis that I can use, just so I have more options in my review?

Other than what? You have not described any actual methods, just the end results.

It sounds like you need a statistical analysis primer and a introductory excell howto book.

 

[semidevil]

Right. So I guess I'm looking at where is the best place to start. Books or website recommendations will be appreciated

 

[Simon Bridge]

With nothing else to go on:
https://www.amazon.com/dp/1118464311/?tag=pfamazon01-20

 

[pbuk]

This kind of reporting does not generally involve any statistical analysis - in the business world it is called variance analysis.

With retail sales it is common to make various adjustments so that comparisons can be made on a "like-for-like" basis - so adjusting for stores that are newly opened or closed, the number of days (perhaps counting weekend days separately) in a month etc.

Also, variances are often split into "volume" (the amount of product sold), "price" (the price of the products sold) and "mix" (selling products at different price levels) - so if last year sales for a week were £2,000,000 made up of £1,000,000 food at £1 per item and £1,000,000 alcoholic drinks at £10 per item and there was 20,000 sqft of retail floor, and this year sales for the same week were £2,500,000 made up of £1,100,000 food at £1.20 per item and £1,400,000 alcoholic drinks at £11 per item and there was 20,400 sqft of retail floor... (this is why accountants use spreadsheets!)

 

[WWGD]

This is one I have liked.

http://www.barnesandnoble.com/w/statistics-for-dummies-deborah-rumsey/1116928612?ean=9780470911082

 

[thelema418]

I think for analysis you first need a conceptual framework about "what matters." It seems to me that you have started this process. As you work with the data and problems that you need to solve for work, this framework will become more refined. If you are making claims about stores performing "better"or "worse" than each other, you must have a clear conceptual framework about what this means so that it can be articulated to others.

Words of caution: I worked on retail projects for many years, and it is important to understand how the point of sale software logs transactions and does backend calculations. These may not follow intuition. For example, when looking at Quantity-Sold and Dollars-Sold, you are not taking into account Quantity-Returns, Dollars-Returns, Quantity-Loss, and Dollars-Loss. In retail situations that reward sales performance, it is possible for managers to fraudulently make sales and then process returns. While this results in a zero net transaction, it increments the Quantity-Sold and Dollars-Sold values. So, if those are the only variables you are looking at, you would incorrectly believe more sales had occurred. The same is true for Dollars-Sold and Cost-Sold. (And cost can get funny depending on how the computer calculates it.)

Another factor is inventory. Quantity-Sold depends on Quantity-on-Hand. This could help identify stores that suffer an under or over stock of an item. Managers sometimes think the item they are selling the most of is their best product, but sometimes they fail to realize the most popular product is selling off the shelf faster than the supply.

 

[FactChecker]

I second what MrAnchovy said. It's more important to present the data in a logical way that non-statisticians are comfortable with, than to get too tangled up in statistical theory. Present the data on a "per-store" or "per-sale" basis, or whatever they are used to. Ensure that the comparisons are on a fair "apples-to-apples" basis, which is very important. It would be good to understand the mean and variance of data, but don't get too concerned about deeper theories.

I just read thelema's answer. That is a lot of good practical advice.

 

[WWGD]

I third Mr Anchovy: (Outside of rarefied research environments) Statistics is a means to an end, not the end itself: theory and real world applications should agree.