Cool article! I never thought about the similarities between incentive pay for developers and top management in this way. I guess some schooling in game theory would have prevented a lot of suffering at a couple of major company.
I think the measurement problem also affects software estimates. A developer will naturally try to give a realistic estimate. However, when he knows that he will be held accountable (shouted at and spit on by the boss, forced to work until after midnight for many moons) when his estimate later turns out to be too optimistic, he will react by putting his self-interest first by both inflating his estimate a bit (left-shifting it more likely) and working slower in order not to beat his estimate by too much.
Solution for management: ensure that estimates made by developers do not affect them personally (affect, not effect, and ensure, not insure, sorry, just had to say that) and win their trust so that they realize that it's safe to tell management how long this word count feature REALLY takes.
If your programmers are always, like Microsoft, just beating their estimates, it's probably time for a little soul-searching...
Michiel de Mare
Monday, July 15, 2002
Wow. People focus on what gets the to the rewards. I believe this is known as a blinding flash of the obvious. Not faulting Joel for saying it, though, as one of the problems with leadership is ignoring the obvious.
The problem with leading large organizations is that people forget that they can handle leading a MAXIMUM of 7, and most effectively 4 other people. If you try to lead all 200 Customer Service Reps, you will fail. Your brain is not wired to do it., You do not have a sufficient set of eyes.
Lets look at Uncle Sam's layout in the Army: A unit is almost always subdivided into 3 Main subunits and a Headquarters element. A Commander/Leader has to Command and Control about 5 maneuver elements, as that HQ usually has 2 maneuver elements. He Has one-two people supporting him directly and Executive officer and a Senior Non-commisioned officer. But he has to lead these two people as well. So now he is a little more tapped out. Course, most smart COs put the HQ section directly under the XO reducing the burden a little bit.
The success or failure of this system comes from the Leader non-micromanaging. There is a positive and negative aspect to this. The easier part is to only give orders to the person heading up the subordinate unit, and to task organize so that one of these people is always incharge of a supporting element. This give the subordinate consitancy of command. That is the easy part. The hard part is to then give those subordinates the leeway to do the task their own way, and grade them solely on performance. This is the hardest part for a commander to do, becasue he remembers being a platoon leader, and hass that experienmce at his beck and call. He can solve the problems of a subordinate unit fairly easily. Good Commandders realize that they cannot do this or they remove their subordinates ability to lead.
Does this setup transition over to the civilian world? Well, the 7 person max part does. What it means is that no division within a company should be larger than 6 people + 1 manager. If it is the type of division that routinely has outside interaction, it needs to be smaller. Probably 4 is the target size, as that allows you room to add one or two as necessary, and if you lose one or two, you haven't massively killed the balance (1 leader / 1 lead).
So those massive call centers with 45 people on a shift need to be subdivided. 6 people per team * 6 teams will get you 36. Perhaps three main groups, with two levles underneath. This gives you a manageable group size and a promotion heirarchy. Just remember, those people up at the top of the ladder are going to spend more and more of their time supervising. You may not want to "Waste" money on this aditional overhead, but you'll end up with those phantom hangups.
Monday, July 15, 2002
This topic has been discussed quite a few times before in this forum. Organizational hierarchies are one thing, but performance measurement is something else entirely.
The military does do performance assessments (once upon a time, they were "fitness reports"), although they just go "up" (i.e., the individual doesn't review themselves, or their peers). These are (or were) subjective and not anonymous, and made a huge difference to one's career.
Monday, July 15, 2002
I have to disagree with the span of control that an individual can lead is optimum when limited to the short term memory limitation of "seven plus or minus two".
The most obvious part of this fallacy is the implication that what is most important for a leader to keep in mind is the "persons" of the team or of those above them. An effective leader keeps his goals in mind, not the bodies, names, roles, titles or "feelings" of people.
The success or failure of any "system" comes from the leader non-micromanaging. The micromanager, who is such because he keeps his team members shortcomings in mind, will be most capable of micromanaging with a staff span of "seven plus or minus two".
This low level of control span will only increase the number of hierarchy levels... increasing all of the bad things about hierarchies along with... such as communication compression.
Most low control span organizations are mediocre... and that is on their good days.
Tuesday, July 16, 2002
I second Joe AA's disagreement with Adam on seven subordinates being the optimal number.
I work in a group of about 20 people with only one boss. Most of the workers are quite senior, and need little if any management. Tell us what needs to be done, then quickly get out of the way. More management would only slow us down.
Tuesday, July 16, 2002
You have more management, you're just not aware of it. All of those senior people are informally "managing" you. If that one boss is smart, he's regularly getting information from the senior people about the junior people, operating through informal channels. The senior people are being relied on to help keep the junior people from drifting into the weeds.
There are also a (very limited) number of exceptional people who really can keep the enough short-term data in their heads at once to manage 20+ people. The rule-of-7 is there so a "normal" person can effectively manage.
Wednesday, July 17, 2002
Would it be useful here to consider that military management requirements are much more rigorous than for business? The military system has to remain valid while everyone is flat on the ground with bullets thrashing through the foliage.
Thursday, July 18, 2002
Fog Creek Home