"Tackiling the Excel problem"

[Kyle123]

I came across this last night:

http://www.insightsquared.com/2013/0...excel-problem/

Whilst the blog is really a sales pitch for the services of the poster, I found myself agreeing with quite a few of the points he makes (though a number of them could equally be levelled at applications).

It would be interesting to hear others' take on it if you fancy a brief break from Excel

[snb]

The absolute lack of any logical reasoning in this 'article' is abhorring.
The lack of any systems analysis, data analysis, structure analysis has nothing to do with the quality of a software program.

[Kyle123]

I'm pretty sure the 'logical reasoning' behind the article is to flog the writer's services

That said, for the most part I agree with him, in my time I've seen a lot of truly terrible business critical spreadsheets - that's not top say good ones don't exist, they do, but IME they are hideously outweighed by the sheer volume of bad ones. I don't believe that it's generally the fault of the people who build them, indeed they usually start out filling a specific niche that is overlooked by general systems - and they often do this well; rather it is the feature/scope creep that goes along with them when managers see the improvement that a spreadsheet has made and ask for it to be expanded upon. Before you know it, entire departments/companies are dependant on spreadsheets that simply weren't designed for the role that they have ended up in, maintained by people who "did a workbook on the side" to help themselves/colleagues in some aspect of their role. You end up with the person who created it then working full time to support it and add features - often without any real knowledge of 'systems analysis, data analysis, structure analysis'. This makes it then extremely difficult to scale up / recover from should something go wrong (like the maintainer leaving / getting hit by a bus). The overriding point is that the versatility and accessibility of Excel ultimately end up being the downfall since anyone can throw together a workbook and add a few macros that follow no conventions, are not properly documented and have no sense of version control.

Maybe I'm just jaded in all this, I've recently had to deal with a Excel "system" composed of over half a million spreadsheets, copious amounts of undocumented, badly written code that people "can't get reports out of" - it's a fun one

Of course it's possible to level all the above at software applications also - lots of bad ones are out the, I've seen a lot of them; but the ease of access isn't the same so the people creating them (usually ) have had some form of training and understand application/data structure.

Again, ( because I reckon I'm definitely gonna get a roasting over this on an Excel forum ) I'm not saying good systems don't exist, of course they do, there are some really good workbooks out there - but the accessibility of Excel also create a huge amount of terrible ones.

I'm now going to go cower in a corner and await my kicking

[RHCPgergo]

I agree with snb. There are logical flaws in the article. I agree with some points, but not with the conclusion. The fact that way too many people claim to be advanced excel users in their resume (without actually being experts) doesn't mean that "Excel is far too complex". That being said, I think that large amounts of data should be stored in databases, not spreadsheets.

Obviously written in order to make their own product appear in better light.

[romperstomper]

There are some truths there, but the example quoted is not a basis for any of the conclusions as far as I can see. What system can cope with users simply inputting false information?

The general theory I think applies to any generic software - it's certainly something that has been said many times about MS Access: although it's actually a good piece of software in many respects, it's far too easy to build a crappy database with it. However, I think that's an inherent problem with flexible programs - if you allow users the freedom to tailor it to any situation, it's pretty much impossible to program logical constraints into it.

[Alf]

It seems that the GIGO effect is still alive and kicking!

Alf

[MrShorty]

A few thoughts:

Point 1) I would wholeheartedly agree with him when he says that Excel (or any spreadsheet for that matter) is like using a programming language. Like any programming language, there are good practices and bad practices. Like any programming language, one can create a robust, user friendly application that performs its task well. One can just as easily (if not more easily) create an application/program that is buggy, difficult to use, crashes, and does not do its task well.

What he calls Excel's "complexity" I might call its "flexibility." I notice that his target audience is fairly narrow -- apparently business managers who need better ways of analyzing their business's data so they can make good business decisions. A spreadsheet, like a programming language, is a general purpose tool that I can use for engineering, someone else can use for statistics, someone else uses it as a financial analysis tool, and so on. When you narrow your focus down to a single niche performing the few tasks associated with that niche, it readily seems possible to build an user friendly, robust application that is simpler to use than a generic, all-purpose application. As a spreadsheet Excel is a good spreadsheet. That does not mean someone cannot come along and build a specific application for a narrower audience that will perform those specific tasks much better than Excel can.

Point 2: I suppose one would have to see what kind of "flags" his application creates. The underlying problem here is older than Excel, older than spreadsheets in general, older than computers, probably even older than writing and arithmetic. The old saying goes, "There are lies, damned lies, and then there are statistics." People have always tried to massage the data to make their case look good and/or their opponents case look bad. I doubt there are any "magic bullets" that will identify bad data/bad statistics. If one wants an accurate picture of whatever data they are analyzing, there does need to be skills and routines in place to allow someone to identify real data and massaged data.