When you're holding Excel, everything looks like a spreadsheet

Every few months I have a conversation with my dad where he asks me to help out with something at the tiny property business that he runs, and it's almost the same conversation every time. It goes like this:

Dad: Blah blah blah insurance blah blah blah ground rents blah blah blah so I just need your help with fixing the data in Excel so we can do a mail merge.
Me: But I don't know how to use Excel.
Dad: What do you mean, you don't know how to use Excel?
Me: I mean that I don't know how to use it. It's big and ugly and complicated and I get lost every time.
Dad: But that's ridiculous. I don't believe you. Everyone knows how to use Excel. You've got a bloody computer science degree. How can you not know how to use Excel?
Me: I've just never needed to use it.
Dad: Well, it's about time that you learnt.

It's no good, though. I learn about one new thing in Excel every year. This year I learnt about AutoFilters. Trouble is, I'll probably have forgotten that by next year. At work we've just started using Joel Spolsky's method of scheduling which relies on Excel. Half the time it's easy, because it's just editing numbers in individual cells. The other half of the time I need to make a major structural change of some kind, like adding a row, and Excel doesn't like it when I do that. (At least, it doesn't like it when I try and do it more than once. I dunno.) As for things like applying formulae, pfft. I realise that if you have even half a clue then you can paint magical formulae all over your sheets with easy swishes of the mouse like a left-brain Rembrandt, but half a clue is still several times more clue than I have when it comes to Excel.

To be honest, I get a kind of old-fashioned-Luddite thrill from it. Earlier today I was trying to fix something on J-Colo so I was tailing logs and wandering through Mailman and qmail and piping things to wc -l and generally acting in a way that suggested that, even though I'm not a proper sysadmin, I could probably play one on TV. And I can switch effortlessly back to Windows, which has been my main computing platform since... ooh... 1993. I'm ready to grab any Windows app by the horns - you just try and stop me! If you're fronting and stepping to my mad W1ndo$e skillz you won't get very far - unless, that is, you throw Excel at me, at which point I'll don a flat cap and click things randomly and whimper in pain, then mutter something about how we didn't need computers for these things in my day, and finally wander over to the nearest marketing bloke and tug his sleeve whilst wearing my best sad-puppy eyes.

If you're in a similar boat to mine and you're looking for a scapegoat, I suggest Bob Frankston. (I'd suggest Dan Bricklin as well, except that two years ago he wrote about my Yiddish piece and my fragile, easily-puffed ego is still glowing enough to mention it here. So I like him.) In fact, if you're looking for a scapegoat for the Personal Computer Industry as a whole, then Frankston's among the prime candidates. Back in the late 70's, Frankston and Bricklin created the first spreadsheet app, Visicalc, based on an idea that Bricklin had when he was doing his MBA.

Back then, spreadsheets were in common use for financial calculations, but were done entirely on paper; changing one of the initial estimate values meant a new sheet of paper and an hour with a calculator. Visicalc initially ran on the Apple ][ and was then ported to a whole load of other machines, but it was on the IBM PC that it made the biggest impact. There are wonderful stories of salesmen demoing Visicalc to accountants who had no idea what a PC was or why they would want one. The salesman would bring up an example spreadsheet, ensure that the accountant recognised what he was looking at, and then change one of the values, causing the change to ripple through the rest of the sheet. The reaction to this was almost uniform: the accountant's mouth would hang open for a few seconds, then he'd pull himself together and pull out his chequebook. Selling the PC to the business world was that easy. The spreadsheet was the biggest of the legendary killer apps, in the true meaning of the term: an application so utterly vital, so revolutionary, that it can single-handedly sell the platform on which it runs.

So the need to handle spreadsheets gave birth to spreadsheet applications... but these days, 90% of the time, people are using them for something else. My dad uses Excel as a lightweight database, storing the details of all the ground rents he manages. I've seen Sean drag it out when he wants a random data munging job done quickly. Chris and Dom have a mate who composes letters in it. During this rant I had yesterday about mocking up UI using Visio someone said that they sometimes used Excel for that purpose as well. For crying out loud, this Japanese bloke has even made it play Pac-Man.

As Joel explains here, the turning point came around Excel 5.0, which had to compete with Lotus's new app called Improv, designed to keep them at the top of the spreadsheet market. The reason that Excel won and Improv failed was that Improv was designed to make it much easier to do financial spreadsheets, but Lotus hadn't realised that people were now using spreadsheet apps for a whole load of other tasks. The Excel team saw this and exploited it, adding a load of features for non-spreadsheet uses such as managing lists. Microsoft's savvy about these things extends to shipping an ODBC connector for Excel so that you can treat an Excel sheet as a SQL database; "Sure, you should ideally be using Access or SQL Server for that, but if it's what you want to do..." (And a year down the line: "Since you're handling all your data over ODBC already, you'd find it a lot faster if you moved to SQL Server...")

Excel's pervasiveness is not just as an application, but as a UI model. In our application we have grid controls that can do smart column filtering. I thought that explaining this to users would be a nightmare, until near-identical functionality was pointed out to me in Excel, which all our target userbase are already using on a daily basis. On a related note, VIPS is the Excel paradigm wrapped around an image processing application, and apparently it even makes sense.

All of this speaks volumes about users' habits when it comes to dealing with new problems, preferring to use the tools they already know which don't fit quite as well as the tools built for the job but which require extra learning. Some techies throw their hands up in despair when seeing people use Excel as a database and scrabble to teach these people SQL. Frequently, these are the same kind of techies that are quite happy to keep large databases in text files on their UNIX systems, dealing with them entirely through vi and the CLI. As such, they should be taking heart: both groups are reducing different problems to a common data format and toolset. Admittedly one of them is proprietary and the other requires the use of arcane commands like cut and sed but fundamentally they're remarkably similar. This kind of raw data-centric thinking is core to the UNIX philosophy - more than that, it's core to most models of computing. Have software designers, in trying to provide a different, bespoke interface for each task, missed a trick here?

Of course, I'm sure this isn't news to anyone other than me. You've all been using Excel happily for years, haven't you? And there I was, thinking I knew about computers. Doubtless, this is the kind of blind spot that will see me begging for spare change within a couple of years. Feh.