Red Teaming

Thinking Tools5.3 Validating Impact and ExecutionAlso known as Red Team Review

From United States military and intelligence war-gaming, set out for organizations by Micah Zenko (Red Team, 2015) and Bryce G. Hoffman (Red Teaming, 2017).

Gives an independent team the mandate to attack a plan, defense, or set of assumptions, and turns what it finds into revisions before anything is committed.

Seven findings from a red team, sorted into fixes and accepted risks

A red team spent two weeks attacking a bakery chain's plan to open twelve stores in a new city. The first finding is placed as an example; select each other finding to read the attack, then the box it belongs in: how likely the attack is to work, and how much damage it would do, decide what the plan does about it. Illustrative.

Items (select one, then a box)
  • Not placedPlacedFix before committingHedge itBudget for itRecord and move on
  • Not placedPlacedFix before committingHedge itBudget for itRecord and move on
  • Not placedPlacedFix before committingHedge itBudget for itRecord and move on
  • Not placedPlacedFix before committingHedge itBudget for itRecord and move on
  • Not placedPlacedFix before committingHedge itBudget for itRecord and move on
  • Not placedPlacedFix before committingHedge itBudget for itRecord and move on
  • Not placedPlacedFix before committingHedge itBudget for itRecord and move on

Damage if it works

SeriousSurvivable
Hedge it
Fix before committing
Record and move on
Budget for it
UnlikelyLikely

Chance the attack works

Sales per store. The attack: the forecast copies home-city sales per store, but six sites sit in office districts that empty on Fridays.Not placed yet.Sales per store. The attack: the forecast copies home-city sales per store, but six sites sit in office districts that empty on Fridays.Fix before committing: Likely and serious: change the plan before any money is committed.The team cut the forecast by a fifth and dropped the two weakest sites.Sales per store. The attack: the forecast copies home-city sales per store, but six sites sit in office districts that empty on Fridays.Hedge it: Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.Sales per store. The attack: the forecast copies home-city sales per store, but six sites sit in office districts that empty on Fridays.Budget for it: Likely but survivable: accept it, put the cost in the budget, and name who watches it.Sales per store. The attack: the forecast copies home-city sales per store, but six sites sit in office districts that empty on Fridays.Record and move on: Unlikely and survivable: write it in the risk log and change nothing.Permit timeline. The attack: health permits in the new city take five months, and the plan allows six weeks.Not placed yet.Permit timeline. The attack: health permits in the new city take five months, and the plan allows six weeks.Fix before committing: Likely and serious: change the plan before any money is committed.Close to certain, and it delays every opening: the launch moved to spring.Permit timeline. The attack: health permits in the new city take five months, and the plan allows six weeks.Hedge it: Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.Permit timeline. The attack: health permits in the new city take five months, and the plan allows six weeks.Budget for it: Likely but survivable: accept it, put the cost in the budget, and name who watches it.Permit timeline. The attack: health permits in the new city take five months, and the plan allows six weeks.Record and move on: Unlikely and survivable: write it in the risk log and change nothing.One flour mill. The attack: both regions depend on one mill and one truck route over a pass that closes in heavy snow.Not placed yet.One flour mill. The attack: both regions depend on one mill and one truck route over a pass that closes in heavy snow.Fix before committing: Likely and serious: change the plan before any money is committed.One flour mill. The attack: both regions depend on one mill and one truck route over a pass that closes in heavy snow.Hedge it: Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.Rare in any given week, but it would stop every store: a second mill signed on as backup.One flour mill. The attack: both regions depend on one mill and one truck route over a pass that closes in heavy snow.Budget for it: Likely but survivable: accept it, put the cost in the budget, and name who watches it.One flour mill. The attack: both regions depend on one mill and one truck route over a pass that closes in heavy snow.Record and move on: Unlikely and survivable: write it in the risk log and change nothing.Rival's opening discount. The attack: the local chain will cut its prices for a month near each new store.Not placed yet.Rival's opening discount. The attack: the local chain will cut its prices for a month near each new store.Fix before committing: Likely and serious: change the plan before any money is committed.Rival's opening discount. The attack: the local chain will cut its prices for a month near each new store.Hedge it: Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.Rival's opening discount. The attack: the local chain will cut its prices for a month near each new store.Budget for it: Likely but survivable: accept it, put the cost in the budget, and name who watches it.Expected, and a month of thin margin is survivable: put in the budget and left alone.Rival's opening discount. The attack: the local chain will cut its prices for a month near each new store.Record and move on: Unlikely and survivable: write it in the risk log and change nothing.Overnight bakers. The attack: overnight bakers are scarce there, and the plan pays home-city wages.Not placed yet.Overnight bakers. The attack: overnight bakers are scarce there, and the plan pays home-city wages.Fix before committing: Likely and serious: change the plan before any money is committed.Overnight bakers. The attack: overnight bakers are scarce there, and the plan pays home-city wages.Hedge it: Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.Overnight bakers. The attack: overnight bakers are scarce there, and the plan pays home-city wages.Budget for it: Likely but survivable: accept it, put the cost in the budget, and name who watches it.Likely, and it costs money but not the launch: the wage line went up 15 percent.Overnight bakers. The attack: overnight bakers are scarce there, and the plan pays home-city wages.Record and move on: Unlikely and survivable: write it in the risk log and change nothing.Landlord exit clauses. The attack: two leases let the landlord end them if the block is redeveloped.Not placed yet.Landlord exit clauses. The attack: two leases let the landlord end them if the block is redeveloped.Fix before committing: Likely and serious: change the plan before any money is committed.Landlord exit clauses. The attack: two leases let the landlord end them if the block is redeveloped.Hedge it: Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.Landlord exit clauses. The attack: two leases let the landlord end them if the block is redeveloped.Budget for it: Likely but survivable: accept it, put the cost in the budget, and name who watches it.Landlord exit clauses. The attack: two leases let the landlord end them if the block is redeveloped.Record and move on: Unlikely and survivable: write it in the risk log and change nothing.Unlikely within five years, and two stores could move: logged, no change.A copied menu. The attack: the rival could copy the three best-selling pastries within a season.Not placed yet.A copied menu. The attack: the rival could copy the three best-selling pastries within a season.Fix before committing: Likely and serious: change the plan before any money is committed.A copied menu. The attack: the rival could copy the three best-selling pastries within a season.Hedge it: Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.A copied menu. The attack: the rival could copy the three best-selling pastries within a season.Budget for it: Likely but survivable: accept it, put the cost in the budget, and name who watches it.A copied menu. The attack: the rival could copy the three best-selling pastries within a season.Record and move on: Unlikely and survivable: write it in the risk log and change nothing.Possible but not likely to hurt much, since the stores sell on freshness: logged.

Your sort matched 0 of 7.Your sort matched 1 of 7.Your sort matched 2 of 7.Your sort matched 3 of 7.Your sort matched 4 of 7.Your sort matched 5 of 7.Your sort matched 6 of 7.Your sort matched 7 of 7.

Fix before committing
Likely and serious: change the plan before any money is committed.
Hedge it
Unlikely but serious: add a backup, buy insurance, or set a point at which to stop.
Budget for it
Likely but survivable: accept it, put the cost in the budget, and name who watches it.
Record and move on
Unlikely and survivable: write it in the risk log and change nothing.

The sort

  • Sales per store: Fix before committing. The team cut the forecast by a fifth and dropped the two weakest sites.
  • Permit timeline: Fix before committing. Close to certain, and it delays every opening: the launch moved to spring.
  • One flour mill: Hedge it. Rare in any given week, but it would stop every store: a second mill signed on as backup.
  • Rival's opening discount: Budget for it. Expected, and a month of thin margin is survivable: put in the budget and left alone.
  • Overnight bakers: Budget for it. Likely, and it costs money but not the launch: the wage line went up 15 percent.
  • Landlord exit clauses: Record and move on. Unlikely within five years, and two stores could move: logged, no change.
  • A copied menu: Record and move on. Possible but not likely to hurt much, since the stores sell on freshness: logged.

An example

A bakery chain with forty stores in one city planned to open twelve more in a city three hours away, and the plan's authors were about to ask the board for the money. The chief executive appointed a red team of four: a finance manager, a district manager who had opened stores before, a supply planner, and an outside adviser who had once run a rival chain. They had two weeks, could talk to landlords, suppliers, and the city's permit office, and were to report only to her.

They came back with seven attacks, each shown as a way the plan would break. The forecast copied home-city sales per store into office districts that empty on Fridays. Health permits took five months, against six weeks in the plan. Every store depended on one flour mill over a mountain pass. The authors sorted the findings with the red team in the room: two fixes before committing, one hedge, two costs to budget, and two risks to log. The launch moved to spring with ten stores and a second mill.

Why it works

The people who built a plan are the worst placed to break it. They know its logic too well to see around it, they have staked their reputations on it, and any review they run tends to confirm what they already believe. A red team works by changing the job: its members are rewarded for finding the failure, they owe the plan nothing, and they are told to show how it breaks, which forces specific attacks where a review would offer general doubts. The independence matters as much as the talent, which is why the team reports to a sponsor above the plan's authors. The last step matters most. Sorting every finding into a fix, a hedge, a budgeted cost, or an accepted risk turns criticism into a revised plan, and it lets the authors accept a risk on the record.

The thinking behind it

Micah Zenko Micah Zenko, a political scientist who studies national security, surveyed the practice in Red Team: How to Succeed by Thinking Like the Enemy (2015) across the military, intelligence, homeland security, and business. He describes red teaming as assigning a group to take the adversary's view, test an organization's plans and defenses, and challenge its assumptions, and he looks closely at the conditions under which such teams are listened to, which is why this tool puts the sponsor and the mandate first.

Bryce G. Hoffman Bryce G. Hoffman, a former Detroit News journalist, completed the US Army's Red Team Leader course and adapted what it taught for companies in Red Teaming: How Your Business Can Conquer the Competition by Challenging Everything (2017). His book presents military red teaming as a set of structured methods for challenging a plan before it is carried out, which is the form a business can run in two weeks with its own people.

How to run it

  1. Define the target. Name the plan, the decision it supports, and what is in and out of bounds for attack.
  2. Appoint a separate team. Choose three to six people with the expertise to find real weaknesses and no stake in the plan.
  3. Give them a sponsor and a mandate. Have them report to someone senior to the plan's authors, with permission to find what they find.
  4. Brief them, but not too much. Share enough to attack the plan without walking them through the authors' reasoning.
  5. Set a fixed time. Give one to four weeks, with rules on whom they may contact and what they may not touch.
  6. Ask for attacks with evidence. Each finding should say how the plan breaks, what would show it, and what it would cost.
  7. Sort every finding. With both teams present, mark each as a fix, a hedge, a budgeted cost, or an accepted risk, and record who decided.
  8. Revise and track. Change the plan, follow each fix to done, and call the red team back after any major revision.

Where it goes wrong

The first failure is a captured team. Briefed in depth by the authors, the red team starts to reason inside the plan and finds only small problems; if every finding is minor, ask whether they were told too much. The second is a team with no protection. When findings embarrass a senior sponsor of the plan, they get softened or buried, so the team reports to someone who can hear them. The third is theater: a long list of findings and no change to the plan. Close every finding with a decision on the record, including the risks the plan accepts and the name of whoever accepted them.

When to use something else

  • Premortem. When there is no time for a separate team and the plan's own authors need to imagine its failure in an hour.
  • Murder Board. When the test is a presentation or a proposal that must survive hostile questioning from a panel.
  • Steelmanning. When the work is understanding an opposing view at its strongest before anyone attacks a plan.

Where it sits in SOLVE eX

Go to the source

  • Red Team: How to Succeed by Thinking Like the Enemy, 2015
  • Red Teaming: How Your Business Can Conquer the Competition by Challenging Everything, 2017

Back to the catalog