Wednesday, May 27, 2009

Burn Your Gantt Charts

Let's say you're in the ditch-digging business. Day in and day out, the team you manage digs ditches from one place to another. If you're managing a team like that, you probably want to allocate resources in a certain way: you have to know when the shovel-repairman will need to be at work, when the wheelbarrowers will need to do their wheelbarrowing, and how many ditch-diggers it will take to dig Ditch #713 by the end of the week.

If you're digging ditches, you can probably use a Gantt chart to plan the project. If you're doing anything that involves a spark of creativity, a confluence of talent, or a unique skill -- keep Gantt charts as far away as possible. Start your project off right by finding a big stack of illegible, impenetrable, utterly obscure and useless project management charts, and setting it on fire.

Gantt charts track predictable progress towards measurable goals. That's dandy, but more and more of the economy is based on unpredictable projects and unpredictable goals. Building a car, a house, or something else nobody wants to buy lately is a linear project with fairly interchangeable inputs. There's complexity, sure, but it's the kind of complexity you can bury under extra resources.

But building a new website, or an internal HR application, or a CRM system? That's a nonlinear task. That's a task that might spawn a brilliant new feature if your main software developer grabs a cup of coffee with the UI designer. A flash of insight might cut the necessary database coding time by 50%, or an unexpected flaw could mean that 5% of the features require 150% of the work.

The last thing you need in that case is great big sledgehammer of a Gantt chart to pound the fun and creativity out of your project. Instead of framing things in terms of Gantt-based projections, throw out the projections entirely: start a project by building the smallest application you can build that would make a positive difference in user's lives. Then, let them use it. Then, ask them what features they want next.

And when you're done, find the ashes of those Gantt charts from a couple paragraphs ago, and bury them. You can never be too safe.