"Best of breed" thinking is for sheep, not software

Major international software vendors have done very well for themselves peddling not just their products, but also the idea that sticking with the big brands is the safest strategy: "Nobody ever got fired for going with the Gartner Magic Quadrant".

But the safest option for the person doing the software buying is often a poor solution for the company. When you buy one-size-fits-all software it very often fails to fit either your exact requirements or your way of doing things. If, as is often the case, it's your business processes that give you the edge over the competition, why on earth would you change them to fit the needs of a software package?

The question is even more appropriate when you consider that a "best of breed" enterprise system will probably cost millions to install, never mind the annual licence fees and the armies of expensive consultants you're going to need just to keep it ticking over.

Sure, all the monster packages come with extensive modification and customisation options, at an additional cost. But the amount of work required to make bits and pieces of different systems talk to each other effectively is almost always massively under-rated, and the results are hardly ever ideal.

Conventional wisdom has it that the longer a software system has been around, and the bigger its user base, the better. The truth is just the opposite. The longer a system has been around, the more likely it is to be built on very old foundations - and it can never take a radical leap into the present because it needs to preserve an upgrade path for all those thousands of clients.

By contrast, a system built from the ground up today can take advantage of massive improvements that have happened over the past 20 years in processor power, memory and software development methodologies. The result can be something that's thousands of times faster and more fit for purpose, at a fraction of the cost.

Despite these advantages many companies remain reluctant to commission their own software, for reasons that owe more to perception than to reality.

First is the perception that custom software development is more expensive. This may have been true ten years ago, but it's far from the case now. The R40 million to R50 million it will cost the average South African corporate just in consultancy fees to install "best of breed" software would buy thousands of hours of development time and top-drawer expertise.

Second is the perception that custom software development is more risky because you're dependent on a smaller pool of skills. This is certainly the case if your supplier is a one- or two-person company, where an accident, illness or emigration could leave you high and dry. But a larger company which is in the business of software production, with good development methodologies and sound quality assurance, shouldn't have any of these problems.

There have also been some high-profile disasters, for example the rocky introduction of eNatis at the national department of transport, that have soured perceptions of custom software development. But apart from the fact that disasters by their very nature get talked about amore than problem-free implementations, it's worth noting that eNatis is now working seamlessly. The new Sars e-filing system is another locally developed system that's functioning extremely well.

Shrinkwrapped software isn't any quicker to implement than custom development in many cases. The inevitable tweaks and modifications can extend the implementation time for big software packages by months, calling for scarce skills in older programming languages. Custom builds, on the other hand, can draw on current development toolsets that are built for speed and productivity: you can get a lot more for a lot less. Custom software development houses may also be able to draw on experience of similar systems, shortening the development cycle even further.

In summary, if you're looking to upgrade or replace your enterprise systems it's worth considering the custom option very, very seriously. From what we've seen, going with the Gartner Magic Quadrant has about the same success rate as relying on the advice of professional stock pickers - and we all know what that's led to.