marginal contribution

You learned from Game Theory that outcomes depend on what other players do, not just your own moves. Marginal contribution makes that dependence precise and measurable.

Define a value function v that takes any group (called a coalition) and returns the total dollar value that group produces together. The marginal contribution of player i to coalition S is:

v(S ∪ {i}) - v(S)

That's it. The value of the group with player i, minus the value without them.

This is closely related to marginal value, which you already know as the dollar improvement from relaxing a constraint by one unit. Marginal contribution applies the same 'what changes at the margin' logic - but instead of relaxing a constraint, you're adding a player to a group. The constraint version asks 'what if I had one more unit of capacity?' The coalition version asks 'what if I had this specific person, team, or resource in the mix?'

The critical insight: marginal contribution is not a fixed property of the player. It's a property of the player relative to the group they're joining. A senior engineer's marginal contribution to a team of three juniors might be enormous. Her marginal contribution to a team that already has two seniors might be modest.

Operators face marginal-contribution questions constantly, usually disguised as Budget or Allocation fights:

• •Headcount justification: 'What's the marginal contribution of a 5th sales rep?' Your CFO isn't asking for a vibes answer. She wants v(team of 5) - v(team of 4) in Revenue dollars.
• •cost sharing: Three product lines share one data platform. How much of the platform's overhead should each line bear? Splitting evenly ignores that some lines use it far more - and that removing any one line barely changes infrastructure cost.
• •P&L ownership: When Revenue comes from the combined efforts of sales, marketing, and product, each function's 'contribution' depends on the other two. Marginal contribution gives you a principled way to decompose joint output.
• •resource allocation: If you have one more engineer to deploy, the team where their marginal contribution is highest is where they should go. This is exactly the logic behind Efficient Allocation.

Without marginal contribution, Allocation degrades into politics. The loudest voice gets the Budget. With it, you can anchor the conversation in measurable impact.

Step 1: Define the value function

Pick a measurable output - Revenue, Profit, Throughput, whatever your P&L cares about. Then measure (or estimate) v(S) for every relevant coalition.

For a small team of three channels - Direct (D), Partnerships (P), Self-Serve (S):

Notice the standalone values sum to $1,600K, but the full coalition produces $2,000K. That $400K gap is surplus - value that only exists because of combinations.

Step 2: Compute marginal contributions

Pick a player and compute their marginal contribution to every possible coalition they could join.

Direct (D):

• •Joining {}: $800K - $0 = $800K
• •Joining {P}: $1,500K - $500K = $1,000K
• •Joining {S}: $1,000K - $300K = $700K
• •Joining {P, S}: $2,000K - $700K = $1,300K

Direct's marginal contribution ranges from $700K to $1,300K depending on context. When partnerships already exist, direct is worth more because it closes the pipeline that partners generate.

Self-Serve (S):

• •Joining {}: $300K
• •Joining {D}: $200K
• •Joining {P}: $200K
• •Joining {D, P}: $500K

Self-serve's marginal contribution is lowest when joining a single channel ($200K) but jumps to $500K when both other channels exist - because it handles the Long Tail of small accounts that direct and partner teams don't pursue, freeing them for larger deals.

Step 3: Use the marginal contributions

Now you can answer real questions:

• •Should we cut self-serve to save overhead? You'd lose $500K in the full-coalition scenario. Compare that to the cost of running it.
• •Where does the next Budget dollar go? To whichever coalition gap has the highest marginal contribution per dollar spent.
• •How do we fairly split shared costs? Marginal contribution gives you the raw inputs. The Shapley value averages them across all orderings to produce a single fair number per player.

Use marginal contribution when:

1. 1)Joint output can't be cleanly attributed - Revenue comes from multiple teams, channels, or product lines working together. Standalone performance doesn't capture the full picture.
2. 2)You're making add-or-remove decisions - Hiring one more person, adding a product line, cutting a channel. You need to know the incremental impact, not the average.
3. 3)You're designing cost sharing or incentives - If you split costs or Commissions without accounting for marginal contribution, you tend to over-reward low-contributors and under-reward high-contributors. This creates bad incentives that erode your best performers' motivation.
4. 4)You suspect diminishing returns - Marginal contribution naturally reveals when the next unit of a resource adds less than the previous one. If the 5th engineer's marginal contribution is half the 4th's, you have a quantitative case for investing elsewhere.

Don't use it when:

• •Output is fully separable (each person's work is independent). Just measure individual output directly.
• •You can't estimate v(S) for the relevant coalitions. The math is simple, but garbage inputs produce garbage outputs.

Marginal contribution is the raw material for the Shapley value, which averages marginal contributions across every possible ordering to produce a single fair Allocation per player. It connects directly to Efficient Allocation (deploy resources where their marginal contribution is highest) and cost sharing (split overhead proportional to marginal contribution rather than headcount). When marginal contribution shrinks as coalitions grow, you've found diminishing returns - the natural stopping point for Capital Investment.