I’m a nuisance when it comes to system design because I can’t get things right first time. I feel guilty, of course, because asking that things should be done again, or done differently, costs us time and money, and I really ought to be able to get it right at the start, if only I thought about it more carefully.

But, better that I should raise issues now rather than later when it’s too late.

Last Friday I was reviewing the new browser-capable version of invoicing that we’re developing for Version 4.9 of time@work, due out in September (or so). Our invoicing module is complex (I should say ‘sophisticated’ really – it’s the right salesman’s euphemism) since our system is multi-company, and invoicing allows recalculation, discounting, exclusion, addition, rounding and all sorts of other features. Producing a .Net browser version has taken months, but we’re nearly done.

It all began some months ago when I sat down with our systems analyst and looked at the mock-up he’d produced. This followed some hours of discussion over the previous days. The mock-up allowed us to click through the invoicing process, from one dummy screen to another, ‘trying out’ how invoicing might look. Our invoicing process can involve about six different pages, and we’d thought very carefully about what fields should be where, and how to reduce the end-user’s clicks and make the whole process as intuitive as possible..

But, despite that, a mock-up isn’t real, so, to our developers’ dismay, when I looked at the system on Friday, with real data and real programming logic behind the screens, it didn’t seem so easy to use as we’d thought it might be, and it was obvious to me that we’d made a few mistakes.

Years ago, when I worked for Coopers & Lybrand Consulting Services in Budapest I went on a ‘Summit D’ course which was all about systems analysis, specification and development. I was taught the typical ‘cascade’ method. Each phase must be completed before the next phase is begun. Once systems analysis is complete, a requirements document must be produced that the end-users will sign off as correct, then a system specification, and then you build the actual system. Not much room for revision. When the system is complete, it’s take it or leave it, and ‘this is what you signed off on!‘

But we’re human and our imagination is limited. We need to touch and feel things before we know they are right. So, nowadays, I much prefer the ‘agile’ method of system development or system implementation. You engage in repeated demonstrations and mock-ups and at every stage you allow for revision and successive approximation to what is really needed. The end-user sees what he’s getting long before the system comes out of the laboratory. Of course he’s never HAPPY but he’s less UNHAPPY this way!

True, without careful control this can be an expensive method, but it’s certainly cheaper than delivering a big, expensive, system to users who find it totally unusable.

So, though I feel guilty from time to time at contradicting myself and asking for change, I know that it’s inevitable, and it’s the best way I know. So, forgive me, designers and programmers. And, after all, it’s not as if you don’t occasionally make a few tiny mistakes yourselves.

Here’s a happy user: