Context Before Authority
One note for Team of Teams: complexity kills prediction, so efficiency stops paying. McChrystal fused daily shared consciousness with pushed-down decision rights, and monthly raids went from eighteen to three hundred.
The Core Insight
On 30 September 2004, four Al Qaeda in Iraq vehicles converged on a sewage plant opening in Baghdad. Thirty-five children died. The blasts wounded 10 Americans and 140 Iraqis.
The task force hunting the bombers held every material edge. It had firepower, armor, surveillance, the hardest training in special operations history, and it hit two to ten targets a night. Al Qaeda in Iraq built bombs from propane tanks and expired Soviet mortars. By the end of 2004, attacks on American forces ran at 87 a day, and the task force was losing.
Stanley McChrystal took command in 2003, and this book is his hunt for the binding constraint. The military term is limfac, and the task force limfac was the mundane art of management. Arquilla counted some twenty of Al Qaeda's number threes killed over a decade, because everyone in a network is number three.
Most managers believe an organization wins by doing the known work faster and cheaper. McChrystal argues that efficiency is a bet on prediction, and in 2004 the bet stopped paying. His reading of that year: the problem was adaptability, not efficiency.
The Framework
Complicated systems have many parts joined in simple ways. One cog turns the next, so a change in one part gives a result you can predict. Complexity arrives when the interactions between parts multiply, and that density of linkages makes a system fluctuate and behave unpredictably.
A system is complex over a given time frame when so many interactions run that nobody forecasts outputs from inputs. Weather is complex over a month and complicated over a day. Lorenz restarted a weather simulation at MIT in 1961 with three decimals of a value the machine held to six, and the runs diverged.
Efficiency is the most output from the least input. You can only optimize that conversion by naming inputs and outputs far enough ahead to build the machine. Efficiency rests on prediction, so it fails wherever prediction fails.
Resilience is the capacity of a system to absorb disturbance and keep its basic function and structure. A 1953 storm tide killed more than 1,800 people in Zeeland, and the Dutch built the Delta Works over the next forty years. In 1993 and 1995 the Rhine flooded Zeeland from behind those defenses. The high levees narrowed the channels, so the water ran higher and faster. Hardening strengthens the parts, and resilience links elements so they can reconfigure.
- Small teams run on trust and shared purpose, and both connections are horizontal.
- Trust does not survive addition, so links between teams get built one person at a time.
- Shared consciousness is a set of central forums that hand every team the whole picture.
- Decision rights then move out to whoever sits closest to the problem.
- The order is strict, because authority without context produces havoc.
Key Ideas
Small Teams Run on Trust and Purpose
On 28 December 1978, United 173 held over its destination with a landing gear indicator problem. McBroom had more than 27,000 hours without an incident. The crew prepared for seventy minutes, five minutes past their reserve fuel, and the crash killed eight passengers and two crew. In 2009, US Airways 1549 lost both engines at barely two thousand feet, and that crew ditched in under four minutes with everyone alive. The 1978 crew was a command, and one man decided. The 2009 crew was a team.
SEAL training runs six months, and its output is superteams. Trainees get boat crews of five to eight on day one and keep them for six months, because nothing there survives alone. They travel with a swim buddy even to the dining hall. Of some 160 entrants around 90 drop, and only about one in ten of those quitters was beaten by the pace.
Hardship tests commitment. Ruiz screened for it by asking why a man came, and those who wanted to try the course dropped. Purpose is the second ingredient, and purpose is shared context. A man entering an unmapped storeroom knows how his teammates react if it starts a firefight.
Adding Teams Does Not Add Trust
In October 2003, a supply closet at Baghdad airport held a four-foot-high mound of sacks. Each sack held evidence from a raid, flown back with the mail and never opened for lack of translators. McChrystal drew an hourglass, covered the bottom half, and asked whether removing it changed anything at the front. The answer was no. He named these choke points blinks.
Teams do not scale by addition. Ranger platoons hold forty-two people and SEAL squads sixteen to twenty. Dunbar caps trust between 100 and 230. The links to manage rise at close to an exponential rate, and adding people to a late project makes it later.
The common failure state is a command of teams, where adaptive teams sit inside a rigid superstructure. It adapts inside each team and stays rigid as a whole. The fix is representation. Across seven thousand people, everyone needs to know someone on every team.
A Meeting Nobody Cancels Carries the Context
Shared consciousness is the book's name for one state. The whole force holds a working picture of the environment and of its own organization, and each team keeps its specialty. The goal was general awareness underneath specialized skill.
The whole Balad headquarters was designated top secret, so nobody had an excuse to withhold. About sixty people worked on its open plywood floor in 2004. His standing guidance was to share information until you are afraid it is illegal.
The Operations and Intelligence brief ran six days a week, was never canceled, and attendance was mandatory. The clock was fixed at 9:00 a.m. Eastern, which is 4:00 p.m. in Iraq. Operators woke late morning, trained, attended, then raided from dusk to dawn. Each speaker got four minutes and gave the update in the first sixty seconds. The rest of the slot was open conversation.
It grew to seven thousand people attending almost daily for up to two hours, across more than seventy locations. Thousands of them watched problems get solved live and stopped asking for permission. One partner agency reported nothing new every day for the first year. By 2005 those agencies raised their participation, because the task force gave them faster intelligence than any other source.
Trust Crosses Units Through People You Cannot Spare
Trust between units gets built by moving people. Late in 2003 the task force began embedding, which sends one person from one team to another for six months. Resistance was intense. Units sent their best people, because whoever went represented the unit. The host unit read one man as the whole unit behind him, so trust generalized from a sample of one.
Liaison duty was a dumping ground, a last tour before retirement. The task force sent its best commandos in civilian suits to embassies thousands of miles from the fight, on two rules. If giving the person up does not hurt, pick someone else. If you do not recognize the voice at 2:00 a.m., pick someone else.
Conway reached one embassy over objections, with no intelligence access and nothing to do. He took out the trash office by office every afternoon, and flew in Chick-fil-A for a colleague who liked it. After three months the ambassador asked about force protection, and the whole task force arrived at once.
A twenty-three-year-old analyst watched three men reach a compound raided the night before and run for their vehicle. She called an operator she knew by name sixty seconds after it sped away. Rotors turned at four minutes, the aircraft was airborne at eight, and the senior man was in custody forty-six minutes after the vehicle left. No supervisor planned any of it.
Decision Rights Belong Where the Eyes Are
Officers used to wake the commanding general to approve a daylight airstrike. He knew only what they told him a minute earlier, and his approval cost fleeting chances.
The book names the pattern the Perry Principle. Perry sailed for Japan in 1852 with orders that set priorities and left the method to him. Grant, in 1864, gave Meade the road for each corps and the place for the wagons. Meade commanded more than two hundred times Perry's force. Land and a postal service made control possible, and when leaders see they want to control. Instant communication made it worse, because reachable leaders held on to authority.
Routing decisions up and down assumes delay costs less than the errors made without a supervisor. By 2004 that assumption was false, so the rule became the 70 percent solution today over the 90 percent solution tomorrow. The estimate turned out backwards. Decisions came faster and better, because whoever decides owns the outcome and reads what no video carries.
If something supports the effort, and it is not immoral or illegal, do it. His most frequent question became what do you need. He decentralized until it made him uncomfortable, and the sweet spot sat on the brink of instability. The trade was Eyes On, Hands Off.
The old system, run tighter and faster, produced ten to eighteen raids a month. By 2006, under the new one, the figure hit three hundred. Personnel and funding rose little, the task force ran seventeen times faster, and a higher share of the raids found their targets.
The Gardener Outperforms the Chess Master
Chess pieces do not think and do not see the board. In 2004 the enemy moved several pieces at once and did not wait its turn. The replacement image is his mother's 1966 vegetable plot. The gardener grows nothing, and she maintains the conditions in which the plants grow.
The daily brief was his instrument, live to the whole force for an hour and a half. When he took off his reading glasses and rubbed his temple, several continents read the gesture. He greeted junior analysts eight levels down by first name, from a cheat sheet. Technology gave him the means to monitor everyone, and he used the same visibility to let the force monitor him.
Practical Applications
Put one forum on a fixed clock and never cancel it. Set the hour so the far side of the organization is awake. Give each team four minutes with the update in the first sixty seconds. Replace the how-many question with the why-do-you-think-that question.
Send your best person to the team you trust least, for six months. The test for the choice is whether giving them up hurts. If it does not, you picked the wrong person.
Push one class of decision down and write the rule as a boundary. If it supports the goal and is not immoral or illegal, the person closest decides and tells you after. Take the visibility trade in return, because the rule holds only while you can see.
Build the picture before you release the authority. Wide latitude with thin guidance is what the book blames for the 2008 crisis.
Who This Is For
Founders past their first team get the most from this. It aims at the moment when one team becomes three. It names the failure where every team is sharp and every handoff is blind.
Skip it while you are five people in one room. Shared consciousness is free at that size, and this machinery exists to buy back what scale takes away.
Read the source for what it is. This is a memoir of a special operations command with an effectively unlimited budget and total authority over its people. Nobody there quits, transfers, or negotiates a title, and the mission supplies purpose at no cost. McChrystal writes in the introduction that this is not a scientific study, and the figures are his own. The campaign result stays contested, and he notes that ISIS took Mosul in 2014.
Three parts copy cleanly: the fixed forum with a hard clock, the exchange of people you cannot spare, and a written boundary for decisions. Mandatory attendance is unavailable from people free to leave, and a purpose that recruits belief for free is rare outside a war.
The Decision
The test takes an hour this week, and it runs on your information flow. Pick the last decision that waited on you. Write the time the person closest to it knew, and the time somebody decided. Cover each layer in between and ask what the front line loses when it goes.
Then run Flynn's question on your reporting. Name your single best source, and treat every other one as worth less. He asked it at a unit claiming more than 250 sources, and the best one proved new and unproven.
Cut one layer that fails the test, then time the same kind of decision again next week. The gap between the two numbers is the only proof that matters.