A big bang adoption is an instant changeover, a “one-and-done” type of approach, in which everyone associated with a new system or process switches over en masse at a specific point in time. Big bang process improvements are useful; however nearly every person involved in planning and executing change avoids them like the plague. Practitioners avoid big bangs for a number of very specific reasons although at the root of these reasons is risk. They are risky because they have:
- Slower feedback. Big bang programs gather production feedback only once when the project is implemented. Because there is only one implementation, defects are discovered late in the project and not contained early when they can have less of an impact. Slower feedback increases risk; you have only one shot to get it right. Kim Pries, the Software Sensei, summed this reason for avoiding big bangs up by stating “Big bang is risky. Incremental allows some containment of error, although, like most things in life, it is not foolproof.”
- Too many moving parts. Because of their size, big bang projects are hard to control, therefore require more administrative overhead. The increased overhead is a tool for trying to control risk. A direct corollary to too many moving parts is that size of the project is directly related to increased risk. This is complicated by the late production feedback, which makes it difficult to recognize risks. Jeremy Berriault, of the QA Corner, summed up this issue by stating, “Big bang leads to too many variables and leads to change requests.Focus on one set (of changes) and build over time. Solely focus on one change, interface or file set. Then see where it goes from there.”
- Late recognition of value. Big bang projects deliver a significant amount of value late in the project. This makes Big Bang process improvement more apt to be canceled if management’s focus wavers (Deming call this constancy of purpose in his famous 14 Points). Arguably, late recognition of value is a specialized version of delayed feedback that increases cancellation risk. Dominique Bourget, also known as the Process Philosopher, humorously summed this issue up by stating, “It’s like losing weight… is it better to do a bit each day rather trying to lose it all on the last day”
The strength of big bang implementations, that everything happens at once, directly increases risk. Increased risk makes big bang approaches to change unpopular. Patrick Holden summarizes the case against big bang change implementation by making the case for incremental and continuous by stating:
“With complex systems or processes with multiple layers, components, services and especially people, then these improvements should be incremental. This incremental approach should take the opportunities, manage risk, overcome inertia or resistance to change, fail and fix, test and improve, create a momentum and demonstrate the benefits.”
In the end, big bang change implementation is useful, but is a riskier approach. Use it with your eyes wide open.