Metrics

In his 14 points, by saying “Eliminate management by objective. Eliminate management by numbers and numerical goals,” Deming directly attacks Peter Drucker, the most revered guru of American management and creator of Management-By-Objectives (MBO). By the time Deming contradicted him, Drucker was an American institution, as sacred as the constitution. Drucker has published a new book every two to three years from 1939 to 2005 and contributed more than 30 articles to the Harvard Business Review. He received numerous awards from several governments, including the US Presidential Medal of Freedom, and has streets and a business school named after him.

In the Lean world, MBO has been replaced by Hoshin Planning, which is subtly but radically different. Drucker’s influence in the US remains so strong, however, that my favorite book on Hoshin Planning, Pascal Dennis’s Getting the Right Things Done, actually pays homage to him. The title itself is Drucker’s definition of effectiveness and, inside, Dennis presents Hoshin Planning as a refinement of MBO. It is still easier today to get through to an audience of American managers by standing on Drucker’s shoulders than by debunking him.

But Deming was in his eighties when he wrote his 14 points, his success in Japan had finally given him an audience in the US, and he must have known that he was running out of time to make his points. He thought MBO was a bad idea  and he would not pussyfoot. 15 years later, Drucker himself came around to the same point of view and recognized that MBO had failed. What did Deming find wrong with it?

In The Practice of Management (1954), Drucker repeatedly asserts that it is essential for managers to have quantitative objectives to work towards. They should set these objectives themselves, consistent with the objectives of the business as a whole. But Drucker is short on how this could actually be done, how you identify proper metrics, how you set targets for each metric, and how you go about reaching these targets. It is not an accidental omission. In  Drucker’s world, managers can find the answers themselves. It is part of their job, and they are supposed to have the requisite talents.

Instead,  Deming sees arbitrary objectives, unrelated to the purpose of the business, set without any plan on how to achieve them or even a means of taking accurate measurements. Deming sees MBO as “an attempt to manage without knowledge of what to do.” The managers focus on outcomes without looking into the processes that produce them. They end up gaming metrics rather than improving  processes, and playing management whack-a-mole.

About managing by numbers, Deming notes that performance evaluations of groups or individuals fail to  differentiate between fluctuations and outliers. Whether you measure quantities produced or units sold, there will always be a ranking, but top performance only indicates a difference in effort or talent if it is high enough to be an outlier. Otherwise, it would be just as meaningful to rank people based on the number of heads in 50 coin tosses.

Deming makes the case that fluctuations should never be the basis for management decisions. If you sell a product through three dealers of equal talent and resources, with territories of equal sales potential, you will still observe differences in results that will enable you to rank the dealers, even though these differences are just the luck of the draw for this particular period.

What Deming does not go into is what happens when you repeat this kind of ranking over a sequence of periods. If the first period was a fair competition, the following are not, as the top performing dealer starts receiving preferential treatment, such as an award to post in the show room, priority access to new products and marketing materials, and special training in sales or customer support. All of these advantages make it more likely that the ranking will be the same in the next period, with a larger gap. Revisiting Pareto shows how iterations of a pure game of luck can lead to the top 20% of the players holding 80% of the chips when the outcome of each round determines the probabilities of success in the next round.

If one of the dealers sold not 5% more but three times as much as the other two, it would be difficult to explain it away as a fluctuation or a fluke. It would be most likely due to the use of a different and more effective approach, which would warrant special treatment.

What is perhaps most remarkable about the story of MBO is that its poor performance has made no dent in Drucker’s reputation as a business thinker, and that many companies continue to use it.