As a hardware guy, it's fascinating for me to observe the truly real-time nature of management of Internet companies , and the amenability of the data to real-time analysis. For example, a retail web business continually tweaks its Web site -- everything from background color to placement of the buttons and core code -- but they do it in an endless series of micro tweaks, then measure the changes in site metrics to determine whether it was the right tweak or not. I saw the extrapolation of this metrics-based management in a company that sells specialty products to women when they tested a series of "hardware" initiatives in the same way they would test Web. In this case, they tried selling their product through real-world parties with interested women customers and measured the effects of different party structures and formats across multiple events to determine the optimum. When I heard the results and analysis and ultimate conclusions, I was struck with the simple question of "why not just ask someone who has done Tupperware parties or Avon, how they do it, and why?" Especially when some of the conclusions on the formats that didn't work were based on fairly obvious problems like, "if the women in the group didn't know each other, we didn't sell as muchâ¦" Fairly obvious, even to a non-Avon user.
Contrasting Internet, software, and hardware highlights to me a fundamental difference and issue in the area of product management, which is always the Achilles' heel of startups. Despite the experiences of VCs, most startups still suffer badly from poor product management -- it's not a well understood discipline, and confused with marketing and sales as much as marcomm is confused with true marketing. Being one of the classic failure points, I am always surprised when investors don't want to spend $ on a solid product management expert before "developing the technology" -- how on earth will you know what to develop? In any hardware company -- take your pick from semi to medical device -- the product is always late because the software isn’t finished. :) The software guys can't really test until the hardware is done. For the hardware guys, the product management and vision had better be perfect. Otherwise, the chip that took over a year to design, tape out, and FAB may work great but have the wrong interface or attributes, or control software that simply doesn't suit the customer -- and another spin takes at least 6 months more. :) These issues are somewhat amenable to analysis (not as in a consumer Web site), but largely the work of a very thoughtful and usually somewhat visionary product manager who understands the customer, the technology, the market, and the likely changes. Good startup CEOs know the fatal error of asking the customer what they want and building the product based solely on that (you get the same product you have today, but at a lower price with more features).
Let's face it, software might be trying to eat the world but if you really know what you are doing, you do it in hardware. :)