Escalation Archetype

Thinking Tools3.3 Mapping the System Around the ProblemAlso known as Arms Race Dynamic

Described by Peter M. Senge (The Fifth Discipline, 1990) and Daniel H. Kim and Virginia Anderson (Systems Archetype Basics, 1998).

Describes two parties who each answer a perceived threat with a countermove that raises the threat to the other, so the contest ratchets upward.

Two retailers each answer the other's discount, and the price war climbs

Each side's balancing loop only tries to close the gap the other opened; together they form one reinforcing loop. At load both answer in full and the discounts climb for two years. Play it, then pull our answer down to half or to zero (a one-sided step back) and watch both curves flatten. Illustrative.

Select a loop to trace it, or a box to read it.B1, We close the gap. When they get ahead, we cut to catch up, which closes their lead. On its own this loop settles. One negative (−) link: an odd count, so the loop balances.B2, They close the gap. When we get ahead, they cut to catch up, which closes our lead. On its own this loop settles too. One negative (−) link: an odd count, so the loop balances.R1, Price war. Our cut opens a lead, their answer opens theirs, and our answer to that opens ours again: two loops that each settle make one that climbs. No negative (−) links: an even count, so the loop reinforces.Our discount. Points off list price at our stores. We aim to be one point cheaper than the rival, and close half the gap each month.Their discount. Points off list price at the rival's stores. The rival aims to be one point cheaper than us, and moves just as fast.Our price lead. How far our discount runs ahead of theirs: the threat the rival sees.Their price lead. How far their discount runs ahead of ours: the threat we see.Our margin. Gross margin, 30% before any discount; every point of discount comes off it.
100%
100%

Our discount, points

0: 0.01: 1.02: 1.53: 2.04: 2.55: 3.06: 3.57: 4.08: 4.59: 5.010: 5.511: 6.012: 6.513: 7.014: 7.515: 8.016: 8.517: 9.018: 9.519: 10.020: 10.521: 11.022: 11.523: 12.024: 12.50.05.010.001234567891011121314151617181920212223240: 0.0: 1.02: 1.5: 2.04: 2.5: 3.06: 3.5: 4.08: 4.5: 5.010: 5.5: 6.012: 6.5: 7.014: 7.5: 8.016: 8.5: 9.018: 9.5: 10.020: 10.5: 11.022: 11.5: 12.024: 12.50.05.010.0024681012141618202224

Their discount, points

0: 1.01: 1.02: 1.53: 2.04: 2.55: 3.06: 3.57: 4.08: 4.59: 5.010: 5.511: 6.012: 6.513: 7.014: 7.515: 8.016: 8.517: 9.018: 9.519: 10.020: 10.521: 11.022: 11.523: 12.024: 12.50.05.010.001234567891011121314151617181920212223240: 1.0: 1.02: 1.5: 2.04: 2.5: 3.06: 3.5: 4.08: 4.5: 5.010: 5.5: 6.012: 6.5: 7.014: 7.5: 8.016: 8.5: 9.018: 9.5: 10.020: 10.5: 11.022: 11.5: 12.024: 12.50.05.010.0024681012141618202224

Month 24 of 24

Our discount, points
12.5
Their discount, points
12.5
Our gross margin, %
17.5

B1, We close the gap
When they get ahead, we cut to catch up, which closes their lead. On its own this loop settles. One negative (−) link: an odd count, so the loop balances.
B2, They close the gap
When we get ahead, they cut to catch up, which closes our lead. On its own this loop settles too. One negative (−) link: an odd count, so the loop balances.
R1, Price war
Our cut opens a lead, their answer opens theirs, and our answer to that opens ours again: two loops that each settle make one that climbs. No negative (−) links: an even count, so the loop reinforces.
Our discount
Points off list price at our stores. We aim to be one point cheaper than the rival, and close half the gap each month.
Their discount
Points off list price at the rival's stores. The rival aims to be one point cheaper than us, and moves just as fast.
Our price lead
How far our discount runs ahead of theirs: the threat the rival sees.
Their price lead
How far their discount runs ahead of ours: the threat we see.
Our margin
Gross margin, 30% before any discount; every point of discount comes off it.

An example

Two home-improvement chains shared a metro area. One spring the rival took a point off its prices on paint and tools. The first chain's pricing team, measured on share, set out to be a point cheaper; the rival set out to be a point cheaper than that. Each month each side cut halfway toward its target, and each side's team explained its cut the same way: we are only responding.

After eight months the first chain's finance director drew the two loops on a whiteboard. Each team's rule made sense alone: if they get ahead, catch up. Together the two rules made one loop with no brake in it, and it had taken four and a half points off gross margin with no change in share. The way out had to come from one side or a shared limit. The chain announced a published everyday price on its top two hundred items, matched any lower local price on request, and stopped answering promotions. The rival's cuts slowed within two months, and neither chain's discount has climbed since.

Why it works

Each side in an escalation runs a balancing loop: when the other gets ahead, act to close the gap. A balancing loop on its own settles. The trouble is that each side's corrective move is the other side's threat, so the two balancing loops join into one reinforcing loop, and the contest climbs with neither side intending it. Both are telling the truth when they say they are only responding.

The structure also shows the ways out. The loop needs both sides to answer in full; if either answers by less, the climb slows and stops, even if the other side keeps going for a while. So a one-sided step back works, if it is announced plainly enough not to read as weakness. The other ways out change what each side measures itself against: a shared limit, a binding agreement, or a goal that does not depend on the rival.

The thinking behind it

Peter M. Senge Peter M. Senge, a senior lecturer at MIT Sloan who brought system dynamics to managers, named escalation among the system archetypes in the appendix of The Fifth Discipline (1990). The archetypes are recurring feedback structures that turn up in very different settings, so recognizing one in a price war, an arms race, or a budget fight tells a manager how the situation is likely to unfold and where it can be changed. The Fieldbook that followed (1994) collects practical tools for using them.

Daniel H. Kim Daniel H. Kim and Virginia Anderson turned the archetypes into practical diagnostic templates in Systems Archetype Basics: From Story to Structure (1998), a Pegasus Communications workbook that teaches each one through worked stories and causal loop diagrams. Kim's earlier Systems Archetypes I (1992) set out how to diagnose a systemic issue with them and design an intervention. Their escalation template is the two linked balancing loops drawn here.

How to run it

  1. Name the two parties and the contest. Two chains and their discounts, two departments and their budget requests, two teams and their headcount.
  2. Draw each side's balancing loop. What each sees as the threat, and the move it makes to close it.
  3. Join the loops. Show that each side's move is the other's threat; the combined loop has no brake.
  4. Plot both sides over time. Confirm the exchange is climbing, and what it has cost both.
  5. Pick a way out. A one-sided step back, a binding agreement, or a shared limit, chosen by what each side can live with.
  6. Announce a step back plainly. Say what you will stop doing and why, so the other side reads it as a policy and not as weakness.
  7. Change what each side measures. Swap a goal relative to the rival for one that is not, such as margin or customer retention.

Where it goes wrong

The step back is quiet. One side stops answering without saying so, the other reads it as a lull, and the next cut restarts the climb. Make the change public and give it a reason that does not depend on the rival.

The parties are not really responding to each other. Two rising curves can have separate causes, such as rising costs or a shrinking market, and a step back then fixes nothing. Check that each side's moves follow the other's in time.

The measure stays the same. A pricing team still paid on share against the rival will find a new way to compete on price. Change the measure along with the policy.

When to use something else

  • Nash Equilibrium. When the contest can be written as a payoff table and the question is which outcome each side's best reply leads to.
  • Tit for Tat. When the parties meet again and again and the question is which strategy for answering the other builds cooperation.
  • Accidental Adversaries. When the two parties are allies whose local fixes hurt each other, not rivals.

Where it sits in SOLVE eX

Go to the source

  • The Fifth Discipline, 1990
  • Systems Archetype Basics, 1998

Back to the catalog