Data Pattern Pro

Mathematical programming / Linearization

Big-M Method

Turn an "if this binary is on, then this constraint applies" rule into a linear constraint MILP solvers can handle directly.

Note Intermediate

Short note. A compact reference, correct as far as it goes. It will be expanded into a full guide with a worked example and an interactive demonstration.

Problem

You need a constraint that only applies conditionally — e.g. “if a machine is running (y = 1), production x can be up to 500 units; if it’s off, x must be 0.” That’s a logical condition, and MILP solvers only understand linear inequalities.

When to use it

Any time a constraint should switch on or off based on a binary decision, and you don’t have (or don’t want to lean on) native indicator-constraint support in your solver.

Formulation

Introduce a binary y and a large constant M that safely bounds the variable it controls:

x <= M * y          # x can only be positive when y = 1
x >= 0
y in {0, 1}

For a two-sided condition (“if y = 1, then x >= 10”):

x >= 10 - M * (1 - y)

Code sketch

import pulp

prob = pulp.LpProblem("big_m_example", pulp.LpMaximize)
x = pulp.LpVariable("x", lowBound=0)
y = pulp.LpVariable("y", cat="Binary")

M = 500  # tightest valid upper bound on x, not "infinity"
prob += x <= M * y

Pitfalls

  • M too large wrecks the LP relaxation and slows the solver dramatically. Always derive the tightest valid bound from the problem data, not a round number pulled from the air.
  • M too small silently cuts off feasible solutions — worse than slow, it’s wrong.
  • If your solver exposes native indicator constraints (Gurobi’s addGenConstrIndicator, CPLEX’s add_indicator), prefer those — see the Indicator Constraints pattern. They skip the M-tuning problem entirely.