Marketplace Pricing Simulator

Elasticity estimation for pricing systems and marketplace supply-demand equilibrium modeling for Uber, DoorDash, and Airbnb-style platforms.

Use this simulator to connect demand curves, price elasticity, promotion impact, dynamic pricing, matching frictions, and market-clearing equilibrium in one place.

Elasticity estimation for pricing systems

1. Demand curves

For pricing systems, a practical starting point is a log-log demand model:

$$ \log Q_t = \alpha - \epsilon_d \log P_t + \beta , \mathrm{Promo}_t + X_t^\top \gamma + u_t $$

where $Q_t$ is demand, $P_t$ is consumer price, $\epsilon_d$ is the own-price elasticity, and $X_t$ includes controls such as geography, hour of week, weather, events, inventory, and competitor conditions.

In practice:

• Uber-style markets often estimate demand at the city-hour or zone-hour level and include rain, commute windows, airport flows, and special events.
• DoorDash-style markets often work at the market-hour or store-hour level and control for basket composition, restaurant availability, ETA, and fee mix.
• Airbnb-style markets often use listing-day or market-day panels with lead time, seasonality, local events, and occupancy controls.

The simulator uses a constant-elasticity demand curve:

$$ D(P, d) = D_0 \left(\frac{P(1-d)}{P_0}\right)^{-\epsilon_d} \left(1 + h \cdot \frac{d}{0.10}\right) (1 + s_D) $$

where $d$ is promotion depth, $h$ is the extra lift from a 10% promotion beyond the mechanical price cut, and $s_D$ is a demand shock.

2. Price elasticity

The point elasticity is

$$ \frac{\partial Q}{\partial P} \cdot \frac{P}{Q} $$

and in the log-log specification it is simply the coefficient on $\log P$.

Applied guidance:

• Use randomized price experiments when feasible. That is the cleanest way to separate willingness to pay from correlated market conditions.
• When experiments are impossible, use quasi-experimental variation such as tax changes, weather-driven supply shocks, or cost pass-through that moves price but not demand directly.
• Estimate elasticity by segment. Riders with urgent trips, diners during dinner rush, and travelers booking months ahead can have very different elasticities.

3. Promotion impact

Promotion impact is not just price elasticity in disguise. A coupon or subsidy can change ranking, salience, urgency, and conversion even after controlling for the net price. A useful regression is:

$$ \log Q_t = \alpha - \epsilon_d \log P_t + \theta , \mathrm{Discount}_t + \phi , \mathrm{Merchandising}_t + X_t^\top \gamma + u_t $$

What to watch:

• Randomize promos or keep holdout groups so you can measure incremental lift instead of gross redemptions.
• Separate short-run conversion lift from longer-run habit formation or cannibalization.
• For DoorDash-like systems, track whether promos pull forward orders from later time slots.
• For Airbnb-like systems, promotions can interact with lead time and occupancy, so estimate by booking window.

Marketplace supply-demand equilibrium modeling

Supply response

On marketplaces, supply responds to payout rather than the full consumer price. A simple supply curve is

$$ S(P) = S_0 \left(\frac{(1-\tau)P}{W_0}\right)^{\epsilon_s} (1 + s_S) $$

where $\tau$ is the platform take rate, $W_0$ is a reference payout, $\epsilon_s$ is the supply elasticity, and $s_S$ is a supply shock.

Adding supply-side incentives

For a simple per-unit incentive, the effective supplier payout becomes

$$ W(P, I_u) = (1-\tau)P + I_u $$

where $I_u$ is an extra payout per completed trip, order, or booked night.

For a threshold bonus or guaranteed-earnings regime, the simulator converts the program into an expected per-completed-unit equivalent:

$$ I_{\mathrm{eff}}(P) = \rho \left[q \frac{B}{T} + \max{0, G - (1-\tau)P}\right] $$

where $\rho$ is the eligible supplier share, $q$ is the expected attainment probability, $B$ is the quest bonus, $T$ is the threshold, and $G$ is the guaranteed payout floor.

The incentive-augmented supply curve is therefore

$$ S(P, I) = S_0 \left(\frac{(1-\tau)P + I_{\mathrm{eff}}(P)}{W_0}\right)^{\epsilon_s} (1 + s_S) $$

This is a reduced-form approximation. Real quest and guarantee programs are path-dependent, but converting them into expected per-unit equivalents makes the trade-off between consumer pricing and supplier incentives easier to reason about.

The simulator then applies a matching-efficiency term $m$ to reflect geographic mismatch, acceptance behavior, batching, and routing losses:

$$ \widetilde{S}(P, I) = m \cdot S(P, I) $$

Dynamic pricing

Dynamic pricing on platforms like Uber or DoorDash is usually trying to reduce excess demand, protect service quality, and improve fill rate. In reduced form, you can think of it as nudging price upward when demand exceeds effective supply:

$$ P_{t+1} = P_t \left[1 + \lambda \frac{D_t - \widetilde{S}_t}{\max{D_t, 1}}\right] $$

where $\lambda$ controls how aggressively the pricing system responds.

That is the simple dynamic mode in the simulator. It moves price from shortage pressure alone and treats surge as a fast feedback controller rather than an immediate jump to the market-clearing solution.

For the richer control mode, the simulator also targets service quality and a small supply buffer:

$$ \mu_{t+1} = \min!\left(\mu_{\max}, \max!\left(1, \mu_t \left[1 + \lambda \left(0.7(f^\star - f_t) + 0.3(b^\star - b_t)\right)\right]\right)\right) $$

where $\mu_t$ is the surge multiplier, $f_t$ is realized fill rate, $f^\star$ is the target fill rate, $b_t = \frac{\widetilde{S}_t - D_t}{\max{D_t, 1}}$ is the realized supply buffer, and $b^\star$ is the desired supply buffer.

The richer mode also allows surge-linked supplier incentives:

$$ I_t^{\text{dynamic}} = \min!\left(I_{\max}, \kappa \max!\left(\mu_t - 1, 0\right)\right) $$

where $\kappa$ is the slope of the extra supplier top-up and $I_{\max}$ caps the payment.

Because supply cannot jump instantly, the simulator uses a partial-adjustment rule:

$$ S_{t+1}^{\mathrm{realized}} = (1-\beta) S_t^{\mathrm{realized}} + \beta S_t^{\mathrm{target}} $$

where $\beta$ controls how much of the gap to target supply closes each interval.

Matching and equilibrium

Completed transactions are limited by the short side of the market:

$$ M_t = \min{D_t, \widetilde{S}_t} $$

and market-clearing equilibrium solves

$$ D(P^\ast, d) = \widetilde{S}(P^\ast, I) $$

for the equilibrium price $P^\ast$.

For the platform, gross revenue and incentive-adjusted net revenue are

$$ R_t^{\mathrm{gross}} = \tau P_t M_t, \qquad R_t^{\mathrm{net}} = \tau P_t M_t - C_t^{\mathrm{incentives}} $$

where the simulator approximates incentive cost as completed-volume times the effective per-unit incentive equivalent.

Interpretation by platform:

• Uber: equilibrium is about balancing rider requests and available driver capacity while keeping ETAs and cancellations under control.
• DoorDash: equilibrium combines consumer fees, promotions, dasher pay, and batching efficiency to keep order fulfillment healthy.
• Airbnb: equilibrium is slower moving because supply responds through host availability and listing participation rather than minute-level labor supply.

Marketplace examples

Practical workflow

1. Estimate demand elasticity using experiments or quasi-experimental price variation.
2. Estimate promotion lift separately from price effects.
3. Estimate supply response to payouts and incentives.
4. Calibrate matching efficiency using fill rate, ETA, or occupancy data.
5. Simulate equilibrium before shipping pricing or promo changes platform-wide.

This page is deliberately simple: it is a planning model for reasoning about direction and magnitude, not a replacement for production forecasting or causal identification.

References and further reading

These references inform the conceptual framing, formulas, and marketplace examples on this page. The simulator defaults are still illustrative rather than calibrated to any one paper or platform.

Platform economics and equilibrium

• Rochet, J.-C., and Tirole, J. (2003). Platform Competition in Two-Sided Markets. Journal of the European Economic Association.
• Armstrong, M. (2006). Competition in Two-Sided Markets. RAND Journal of Economics. Open-access version via UCL Discovery.

Demand curves, elasticity, and promotion impact

• Cohen, P., Hahn, R., Hall, J., Levitt, S., and Metcalfe, R. (2016). Using Big Data to Estimate Consumer Surplus: The Case of Uber. NBER Working Paper 22627.
• Bajari, P., Nekipelov, D., Ryan, S. P., and Yang, M. (2015). Demand Estimation with Machine Learning and Model Combination. NBER Working Paper 20955.
• Dubé, J.-P. H., Fang, Z., Fong, N., and Luo, X. (2016). Competitive Price Targeting with Smartphone Coupons. NBER Working Paper 22067.

Dynamic pricing, matching, and marketplace supply

• Castillo, J. C., Knoepfle, D., and Weyl, E. G. (2017). Surge Pricing Solves the Wild Goose Chase. Microsoft Research / ACM EC.
• Yan, C., Zhu, H., Korolko, N., and Woodard, D. (2020). Dynamic Pricing and Matching in Ride-Hailing Platforms. Uber Engineering summary with paper link.
• Chen, M. K., Chevalier, J. A., Rossi, P. E., and Oehlsen, E. (2017). The Value of Flexible Work: Evidence from Uber Drivers. NBER Working Paper 23296.
• DoorDash Engineering (2021). Managing Supply and Demand Balance Through Machine Learning.
• DoorDash Engineering (2026). Smarter promotions with causal machine learning.

Airbnb-style flexible supply and pricing

• Farronato, C., and Fradkin, A. (2018). The Welfare Effects of Peer Entry in the Accommodation Market: The Case of Airbnb. NBER Working Paper 24361.
• Airbnb Resource Center (2023). What is Smart Pricing?
• Airbnb Resource Center (2024). Using Airbnb pricing tools.