Price Elasticity of Demand

Why does elasticity matter to founders?#

It stops revenue-only pricing. A discount must be repaid out of a smaller unit margin, so the required volume lift is always larger than the discount percentage. The formula is:

required volume multiplier = old unit contribution / new unit contribution

It exposes segmentation you are currently averaging away. SMB and enterprise, power users and occasional users, and different geographies frequently sit on different demand curves. A single company-wide coefficient blends them into a number that describes no one. This is the argument for price fences and tiered packaging rather than one price.

It separates a price problem from a packaging problem. Buyers are often highly sensitive to total contract value and much less sensitive to a well-aligned usage meter, an annual commitment, or a narrower tier. Changing scope changes the product being measured — do not report that as pure price elasticity. See Pricing Metric and Value Metric.

It makes a pricing claim fundable. "Customers told us they'd pay more" is an interview finding. "We ran a randomized test across 40% of new signups, measured an arc elasticity of −1.5 ± 0.4 over a ±15% band, and contribution improved 9% with retention flat at 90 days" is evidence.

Worked example: an elastic price cut that destroys contribution#

A product sells 1,000 units at $100. Price is cut to $90 and demand rises to 1,180 units. Unit variable cost is $35.

Step 1 — Arc elasticity#

midpoint quantity = (1,000 + 1,180) / 2 = 1,090

midpoint price = ($100 + $90) / 2 = $95

% change in quantity = 180 / 1,090 = +16.51%

% change in price = −$10 / $95 = −10.53%

E = +16.51% / −10.53% = −1.57

|E| = 1.57 > 1, so demand is elastic over this interval. Note the sign: quantity moved opposite to price, exactly as expected.

Step 2 — Revenue rises, as elasticity predicts#

Step 3 — Contribution falls anyway#

Contribution fell $100. The break-even volume for this price cut is:

$65,000 / $55 = 1,181.82 → 1,182 units (at 1,181 units contribution is $64,955, still short)

Equivalently, required volume multiplier = $65 / $55 = 1.1818, a +18.18% lift. The test delivered +18.0%. A 10% discount needed an 18.2% volume gain and got 18.0% — the decision was a coin flip dressed up as a 6.2% revenue win.

Now add realism: if the extra volume costs $3/unit more to support, new contribution is 1,180 × $52 = $61,360, a 5.6% decline against the original $65,000.

Step 4 — What the inverse-elasticity rule says, and what it does not#

For a profit-maximizing monopolist facing a known, differentiable demand curve with constant marginal cost, the first-order condition MR = MC rearranges into the Lerner index:

(P − MC) / P = −1 / E = 1 / |E|

At $90 with MC = $35, the realized margin ratio is ($90 − $35) / $90 = 61.1%, which corresponds to |E| = 1.64. The measured elasticity is 1.57 — less elastic than the optimum requires, which says the $90 price sits below the rule's optimum. That is consistent with what the contribution table showed: the cut went the wrong way.

You can see the same thing through marginal revenue directly, using MR = P × (1 + 1/E):

MR = $90 × (1 + 1 / −1.57) = $32.63, which is below the $35 marginal cost. Selling the marginal unit at $90 loses money on the margin.

Now the caveats, which matter more than the arithmetic. This condition assumes a single-product monopolist, a known and smooth demand curve, constant marginal cost, no competitor reaction, and profit maximization as the objective. A startup typically has none of these: demand is estimated from one noisy two-point test, competitors respond, marginal cost steps with capacity, multiple products cannibalize each other, and the objective often includes learning, land-grab, or penetration. Treat MR = MC as a consistency check on a decision you reached another way — never as the decision.

What are the common mistakes?#

• Reading the sign wrong. "Our elasticity is 1.5" is ambiguous, and "elasticity improved to −2.0" usually means demand got more price-sensitive, not less. State the convention every time.
• Using revenue change as the quantity change. Revenue already embeds price; dividing revenue change by price change produces a number with no interpretation.
• Mixing list price with realized price. Discounts, credits, ramps, and multi-year terms change the price the customer actually faced. Elasticity must be computed on the price that drove the decision.
• Treating a historical price/volume correlation as an elasticity. Netflix is the clean counterexample: price up, volume up, because content, distribution, advertising, and the competitive set all moved too. Without a counterfactual you have a correlation, not a coefficient.
• Extrapolating outside the tested interval. An arc elasticity measured across a 10% band tells you almost nothing about a 2x price change, where new segments enter and budget-approval thresholds bind.
• Averaging incompatible segments. One coefficient over SMB plus enterprise plus a self-serve tier describes a customer who does not exist.

When does price elasticity break?#

Small samples and long sales cycles. Enterprise deals are infrequent, negotiated, and heterogeneous. A handful of closed-won records cannot identify a demand curve. Use customer-level scenarios, win/loss data, and discrete-choice research instead of manufacturing a coefficient.

Capacity constraints. Observed quantity is min(demand, capacity). If you sold out, your measured elasticity is a measurement of your supply, not your demand. Instrument attempted demand, waitlists, and lost deals.

Simultaneous changes. A price move shipped with a feature launch, a rebrand, a new channel, or a competitor's outage is not an elasticity observation. Prefer randomized or phased rollouts; where that is impossible, name the confounds explicitly.

Network and multi-sided effects. In a two-sided market, a price change on one side alters participation and therefore value on the other. One-sided elasticity is structurally incomplete there.

Publication and vendor bias in benchmarks. Corrected estimates in the gasoline literature were roughly half the published ones. Assume any external elasticity benchmark is biased away from zero, and discount accordingly.

Fairness and trust. Price tests touch real customers. Use coherent eligibility rules, avoid sensitive attributes, honor existing contracts, and never use a deceptive reference price. See Behavioral Economics for the boundary.