panoptic.com
Compare with the graph invented by Virginia Satir (Satir Change Model)
panoptic.com
Combine that truth with Incremental Development, and you have a Better Change Model:
(Side-by-side view: panoptic.com 
)
Note: All I did was repeat the graph into 5 steps. Notice how much less "Negative Performance" area the incremental version takes up than the big-leap version.
Technically, this is no different from anything Satir might have proposed. However, I've heard (and, embarrassingly, used) the Satir model as an excuse for being too lazy or stupid to subdivide a problem into smaller pieces.
This applies to much more than programming.
Stephen Jay Gould HasThisPattern in his book Full House (ISBN 0609801406) where he describes his ascension to pinball wizardry.
I don't understand how you get less negative performance? If each iteration as a negative performance region then you will get negative performance proportional to the number of iterations.
This argues against incremental development.
If you assume that it takes you the same length of time to reach the new status quo, you spend less time operating less effectively than the old status quo. The initial dip below the line is much shorter. You'll lose effectiveness on each iteration, but since you're still ahead of the original status quo, this is better.
I'd argue the other way. The overhead of each period of chaos makes the development time larger.
That's why I made the graph. Just imagine shading in the areas of the "Swan" that exist below the Time axis -- this is "Negative Performance". Which one occupies more, visually? Is that just a parlor trick, or is it an artifact that demonstrates why incremental solutions can be more readily accepted than big-commitment solutions?
Hmm, there are a lot of free variables here. That's not a problem if we consider the chart to be qualitative, but introducing a definite number of increments suddenly forces it to be quantitative. So, really how much does performance drop during the "chaos" stage? Is the drop equal in magnitude in every increment? Or is is equal in proportion? Or does it vary? Why?
Does the drop below the line labelled "time" really mean negative performance in the sense of going backwards, of undoing today what was done yesterday? Now this makes we wonder; are you meaning Incremental Development here? (read "development" as "work", if you like) Indeed, with purely Incremental Development, the chart might look as you suggest, but not with Iterative Development and more so not again with Evolutionary Development.
In these cases there has to be an undoing of previous work, which if micro-scale "productivity" is your measure of interest, does indeed mean negative performance. This is why those approaches scare many project managers. And indeed the reason for doing this is that those bits of undoing prepare you to move forward more quickly this time round. If you do mean Incremental Development, but without iteration or evolution, then it's not clear to me that any increment is going to involve the kind and severity of change that the Satir model addresses. Is the change this increment of moving on from one sub-system to another really likely to cause the severe reactions that Satir talks about, psychosomatic illnesses and all? Note that in a well functioning iterative/evolutionary team, the changes introduced on each iterations boundary also are not the kind of thing that Satir was talking about.
However, in the case of the really severe organisational change that this model is usually applied to, then you might have a point. You still need to bind those free variables, though.
The following comment split off from the above thread.
Unless you are assuming a "closed system." By closed system I mean the changes are not visible to the external world. Otherwise changes require propagation through the development organization, which can be the bulk of any effort.
Actually, that's part of it. If you have an organ that works wonderfully in a closed system, such as incrementally developing features, the organ doesn't do any good if the larger organism rejects it. So, the lesson to learn: Incrementally affect change on your organization. No one likes to be in the "Negative Performance" zone. Almost everyone wants credit for "Positive Performance". So serve it up in sipper cups and clink your glasses to ol' Ginny Satir.
Although the above graph may be technically true, it becomes meaningless as the time periods between changes become ever smaller. If your period of change is say, 4 hours, then when viewed at the scale of 1 month or more, progress is continually upward. This is the true significance of incremental development, exceedingly small cycles of change.
See original on c2.com