I have doubted some statistics given out as savings that various techniques have achieved (See my post of Sept. 14, 2005 on 20% for e-billing and Aug. 5, 2005 for returns on other techniques.) A statistical notion, regression to the mean, may explain some of the alleged savings.
Isn’t it probable that law departments which have been assailed by large spending increases are more likely to put in place some cost-saving measure? And isn’t it likewise probable that extreme surges in spending are commonly followed by a drop off in spending – if only because the average expense rate returns to be the norm? To give a prosaic example: after several days of unseasonably hot weather, it would be wise to predict that the temperature will drop.
“Regression to mean” explains the likelihood that a high extreme will be followed by a drop, so that typical behavior returns and the average holds. Hence cost-afflicted law departments, even had they not implemented budgets or bill review or e-billing or competitive bidding or whatever, are more likely than not to report lower spending in the year or two after a spike. Their metrics regressed (went back) to the mean (the typical average). It doesn’t necessarily mean that the technique can take all the credit.