ML Systems Lab Open interactive version →
Foundational 22 min read Pricingobjective designcontribution marginunit economics

Revenue vs Margin: Choosing the Objective

The single most consequential pricing decision is what you are optimizing — top-line revenue, contribution margin, or long-term customer value

Before any elasticity math, someone has to answer a question that feels like a formality but decides everything downstream: *what are we maximizing?* Revenue and profit peak at different prices, and a model that quietly optimizes the wrong one will look successful while destroying value.


The two optimal prices are different, and the gap is the marginal cost. Revenue is P·Q, maximized where marginal revenue = 0, which happens exactly at |ε| = 1 — the unit-elastic point. Profit is (P − c)·Q, maximized where marginal revenue = marginal cost (MR = MC). These coincide *only when c = 0*. With any positive unit cost, the profit-maximizing price is strictly higher than the revenue-maximizing price (you don't want to sell cheap units that barely clear cost). So "we grew revenue" and "we grew profit" can be the results of opposite price moves.


Contribution margin is the identity that exposes the trap. Contribution = (P − c) × Q. A price cut can raise Q enough to grow revenue P·Q while the per-unit margin (P − c) collapses — so revenue is up but contribution is down. Example: c = 6. At P = 10, margin 4 × 1000 units = 4,000. Cut to P = 8, sell 1,400 units: revenue jumps 10,000 → 11,200 (looks great), but contribution 2 × 1,400 = 2,800 — you sold 40% more and made 30% *less* money. Revenue is a vanity target when c is meaningful.


So objective choice is a strategy decision, not a math default. Maximizing revenue/share makes sense in a land-grab (network effects, near-zero marginal cost, winner-take-most). Maximizing contribution makes sense in a harvest phase or a cost-heavy business. And often the true objective is long-term customer value — a low intro price that loses margin now but raises retention and lifetime value. This is the pricing analog of RecSys value-model weighting: the *weights on the objective* are a product decision the model then optimizes faithfully. Pick them wrong and every downstream number is precisely optimized toward the wrong destination.

Key points

Takeaway

The most consequential pricing choice is the objective itself: revenue P·Q peaks at |ε| = 1 while profit (P − c)·Q peaks at MR = MC, and they coincide only at zero marginal cost — so with real costs the two optimal prices differ. A price cut that grows revenue can shrink contribution (P − c)×Q, so revenue is a vanity metric whenever c matters. Picking revenue vs margin vs LTV is a strategy decision the model then optimizes faithfully.

Recap

Check your understanding

Q1. Unit cost c = $6. Price cut from $10 to $8 raises volume from 1,000 to 1,400 units. Select the two correct statements about this outcome.

Q2. Under what condition does the revenue-maximizing price equal the profit-maximizing price?

Q3. A subscription startup deliberately prices below the contribution-maximizing point in its first two years. What objective is this most consistent with, and is it necessarily irrational?

Try it interactively

ML Systems Lab is a free interview-prep platform for ML engineers — work through the full interactive module, quizzes, and drills.

Open ML Systems Lab →