Beat the clock

With fierce competition in the channel, dealers know that timing is everything - one late product and sales could suffer greatly. So how can dealers make good on their promises?

Software marketing people love playing pin the tail on the donkey,s everything - one late product and sales could suffer greatly. So how can dealers make good on their promises? but they do have their own version. It's currently at version 1.02 because there were bugs in versions 1 and 1.01.

Instead of using a donkey's anterior appendage, players take pins representing the forthcoming releases of their products and stick them blindly into next year's calendar. PR representatives then use the results for press releases and marketing brochures, listing the salient features of these new packages and their expected launch dates.

This, at least, is how it often appears to the rest of the industry as software seems to rank second only to railway tunnels, bypasses and millennium memorials in its apparent inability to meet any pre-announced deadline.

The analogy with major construction projects is not incidental. A modern operating system or application is complex - there are 14 million lines of code in Windows 95 and tens of millions in NT 5. A simple package like Symantec's WinFax takes about 15 years to write. A bigger job like rewriting Microsoft Office for the Mac takes almost 10 times as long.

The complexity of testing such a monster, especially on hundreds of permutations of hardware and software configurations, is huge - Microsoft says it employs more people to test its software than to write it. 'Software development will never be an exact science,' says Jonathan Hulse, Microsoft Mac product manager for Office 98. 'It is an ongoing, living process. Configurations, operating systems and browsers all change.

'This is such a rapid, fluid industry, with so many dependencies, that it is difficult to set a date 12 months in advance. Yet this is what many customers demand.'

A big corporate, with thousands of PCs running bet-the-business applications, needs a detailed road map as far in advance as possible.

This leaves software houses in a cleft stick - pre-announce and be criticised for not delivering, or spring a last-minute surprise. Andrew Lees, Microsoft director of product marketing and internet, says: 'There's a lot of pressure from customers to give details of future products for planning purposes.

In most cases, by giving estimates we create a rod for our own backs.

We're damned if we do, damned if we don't.'

And there are other pressures. Microsoft had to reveal more than it would have liked of Windows 95 after details started leaking out in the press. A combination of pre-announcements and over-enthusiastic press coverage got the blame for some of the apparent slippage in new releases.

Outside Microsoft, the consensus is that NT 5 should have been launched by now. Lees argues that it is not late and that Microsoft sent out the first beta version as intended in September 1997. The second beta is expected - but not promised, says Lees - in second quarter 1998 and the finished product is expected to ship sometime in 1999.

'Most people get their information about Microsoft third-hand because we don't have a perfect communication mechanism,' says Lees. 'Selling cheap software means you can't have the IBM model of sales people visiting every customer.'

But Lees admits that Microsoft has completely changed its blueprint for NT 5. In response to the mass market it has added features to make the product appeal to the widest possible audience, and Lees says it has taken aboard customer concerns about cost of ownership.

All this takes time to implement and has had a knock-on effect on Windows 98. Its launch was originally scheduled for early 1998 but has been postponed until later in the year, partly because Microsoft wants to add some of NT 5's cost-reducing features to Windows 98. The testing overhead is so large it would have been impossible to launch the original product in January and an improved version in the summer.

Spec changes, often suggested by users during beta-testing, were responsible for the delay to Windows 95, which was scheduled for early 1995 but not shipped until August.

There is a plus side to the delay, though - it allowed for more thorough testing. 'We haven't had to do a bug fix for 95 - it's amazing,' says Lees.

The booby prize of the current Microsoft stable goes to Office for the Mac. The next version was originally intended to be Office 97, launched within three months of the PC version, which came out on time in January 1997.

Its new title, Office 98, says it all - availability is scheduled for this winter, so we can expect to see it before 21 March 1998.

The reason for the delay is quite simple, according to Hulse: 'When we reached the critical deadline, roughly when Office 97 for Windows was launched, the product wasn't ready.' This was because Microsoft had changed its working methods, setting up a dedicated Apple development team to tailor the software rather than have the original developer's port over the Windows version.

The implication, although Hulse does not admit this, is that since Microsoft invested $150 million in Apple, it wanted to make a better job of the software than usual.

When you're the market leader, of course, you can safely assume your customers will wait. And the plus side of being so late is that it has been possible to include new features in Office 98.

QuarkXPress users are in a similar position. Version 4 has only recently arrived on dealers' shelves, despite being announced over a year ago and widely reviewed in beta form.

For months, Mac owners could not understand why they couldn't buy the software. Developers complain that so many things are outside their control - coping with the demands of different operating systems and hardware components like graphics cards; writing for standards like GSM or ISDN, which are themselves fluid and subject to change; and the ever-changing focus of the internet and the technologies that surround it.

Apple subsidiary Claris had to delay the launch of its HomePage 3 Web page creator from October to December. This was partly because of the complexity of linking it to Filemaker Pro 4, but also because of the need to add features which would help it compete with rival products and meet the ever-changing needs of the internet.

The Web was indirectly to blame for delaying ClarisWorks Office by a few weeks - apparently there was a mix-up over third-party software from internet service providers. Sometimes though, software houses openly admit that delays are caused by bugs.

Corel released the 16-bit version of WordPerfect Suite 7 three months late in the US - July. The UK version should have been ready shortly afterwards. But there were problems with installation and networking, and Corel responded to customer reaction by delaying the international versions until the problems were fixed.

The UK code should be finished by the time you read this - unless the testers unearth any more bugs.

Dealers and resellers take the excuses with a pinch of realism. 'Most software manufacturers announce products far in advance of availability, for competitive gain,' says Steve Caunce, software marketing manager at Computacenter. 'If someone makes an announcement, then a month later their competitors make a strategic announcement about a product that is just around the corner, it causes confusion in the minds of customers and can stall the market - even if it arrives on time.' If a release is several months late, customers may ignore it and skip straight to the next version.

Caunce estimates that this could cost a large reseller #250,000 for sales, installation and services for a corporate customer with 2,000 PCs. But he admits that customers sometimes confuse strategic road maps with product announcements, and sometimes push software houses hard for advance information.

Big customers tend to know the score, anyway. Pete Springfield, technology manager at Compel, says: 'No responsible corporate IT group would plan any kind of project around products which aren't here today and proven.

They don't take much notice of beta dates and pre-release announcements.'

Springfield's customers are more concerned about quality than dates.

'The problem is that software manufacturers don't finish one product before they start another,' he says. 'A lot of people would rather see longer term stability than lots of new technology being rolled in.'

Software releases also tend to address problems people do not realise they have, he continues. 'Applications tend to come out in new versions every 12 to 18 months, which is just too often for many users.'

This is a view shared by Tim Felmingham, managing director of Apple dealer Axiom. 'Our customers' perception is that there's new software coming at them all the time,' he says. 'They are surprised that yet another new version has come out, rather than complaining that it's late.'

Those hardest hit by late delivery are mail-order dealers, who have to produce catalogues and advertisements up to three months in advance. Including a product which doesn't ship aggravates users, and omitting one which does costs sales.

'It's a lose, lose, lose situation,' says the software manager for one catalogue company. 'Even the manufacturers are losing business because of it. If it's because beta testing has thrown up bugs or feature changes, fair enough. But sometimes it's because the packaging isn't ready in time for photography, which is just plain stupid.

'The UK subsidiaries of the big software companies are very good at sharing what they know, but often they are not told soon enough by their US parents who don't seem to understand the mail-order market.'

There is also the legal aspect to consider. Byte Computer Superstores produces advertising material six to eight weeks in advance. Steve Rigby, Byte Computer operations director, says: 'Technically speaking, from the Advertising Standards Authority's point of view, you must not advertise something that you don't have in stock.'

The one thing that software houses, dealers and customers do agree on is that it's better to be late and clean than on time and buggy. The PC software industry is still haunted by the ghost of Ashton-Tate's dBase IV, launched around 1989, which was so riddled with bugs that it threw away one of the strongest market-leading positions in the industry. The company never really recovered and was eventually taken over by Borland.

And finally, the good news. If you thought late software was a recent problem, it isn't. Microsoft proudly informed us that, for the first decade of the company's existence, none of its applications was released on time or even within a week of its due date. The first to meet its deadline was Excel 3, launched in 1991.

Since then, quite a few of Microsoft's masterpieces have met their deadlines.

So, then, isn't it mean to always point out the ones that didn't?