Just do it

Linux World points out how Google tests new ideas (via 37signals)...

Google is one of the few large companies that gets one fundamental rule of the Internet: Trying stuff is cheaper than deciding whether to try it. (Compare the cost of paying and feeding someone to do a few weeks of P* hacking to the full cost of the meetings that went into a big company decision.)

Don’t overplan something. Just do it half-assed to start with, then throw more people at it to fix it if it works. Worked for every successful Google project from AdWords to Google Maps.

Years ago I was consulting for a large technology company. They had an intranet measuring in the billions of pages and were looking to rein it in. One part of the effort was to create and maintain a global taxonomy to make things easier to find, but to do this, they needed a tool to manage that taxonomy. I was on a team of 8 people who were tasked with picking a product that would meet the client’s needs. These were all mid to senior level people, and I’d imagine the annual salary averaged over $100k each. In addition to this full time attention, we had the part time attention of dozens of others. Two project managers were assigned to keep the project on the rails. It wasn’t uncommon to have a meeting where 15 people attended. We also had two outside consultants who were experts in taxonomy management.

To choose the product, the eight of us worked full time for a year, analyzing requirements, interviewing stakeholders in various groups, documenting all the different pieces of software the taxonomy would be used in, and generally trying to figure out which software package best met the needs of the client.

Two weeks into the process we determined that there were four software packages that had any chance of being useful. Three of these packages cost around $20k and the other cost around $70k. Early on in the project planning process, a software engineer was tasked with eventually implementing a pilot project of whatever software we picked. His estimate of the task was that it would take 4 weeks to do.

We spent well over one million dollars to decide which $20k software package to buy so one person could spend a month creating a pilot project.