dynamicpricingin ridesharingplatforms · relatedwork...
TRANSCRIPT
![Page 1: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/1.jpg)
Dynamic Pricing inRidesharing Platforms
A Queueing Approach
Sid Banerjee | Ramesh Johari | Carlos RiquelmeCornell | Stanford | [email protected]
With thanks to Chris Pouliot, Chris Sholley,and Lyft Data Science
20 November 2015
1 / 39
![Page 2: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/2.jpg)
Ridesharing and Pricing
2 / 39
![Page 3: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/3.jpg)
Ridesharing platforms
Examples of major platforms: Lyft, Uber, Sidecar
3 / 39
![Page 4: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/4.jpg)
This talk: Pricing and ridesharing
Ridesharing is somewhat unique among online platforms:
The platform sets the transaction price.
Our goal: Understand optimal pricing strategy.
4 / 39
![Page 5: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/5.jpg)
Our contributions
1. Amodel that combines:▶ Strategic behavior of passengers and drivers▶ Pricing behavior of the platform▶ Queueing behavior of the system
2. What are the advantages of dynamic pricing over staticpricing?
▶ Static: Constant over several hour periods▶ Dynamic: Pricing changes in response to system state;
"surge", "prime time"
5 / 39
![Page 6: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/6.jpg)
Related work
Our work sits at a nexus between several different lines ofresearch:1. Matching queues (cf. [Adan and Weiss 2012])2. Strategic queueingmodels (cf. [Naor 1969])3. Two-sided platforms (cf. [Rochet and Tirole 2003, 2006])4. Revenuemanagement (cf. [Talluri and van Ryzin 2006])5. Large-scale matchingmarkets (cf. [Azevedo and Budish
2013])6. Mean field equilibrium (cf. [Weintraub et al. 2008])
6 / 39
![Page 7: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/7.jpg)
Model
7 / 39
![Page 8: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/8.jpg)
Two types: Strategic and queueing
We need a strategic model that captures:1. Platform pricing2. Passenger incentives3. Driver incentives
We need a queueingmodel that captures:1. Driver time spent idling vs. driving2. Ride requests blocked vs. served
8 / 39
![Page 9: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/9.jpg)
Preliminaries
1. Focus on a block of time (e.g., several hours)over which arrival rates are roughly stable
2. Focus on a single region (e.g., a single city neighborhood)▶ For technical simplicity▶ Insights generalize to networks of regions
3. Focus on throughput: rate of completed rides▶ For technical simplicity▶ Same results for profit, when system is supply-limited▶ Similar numerical results for welfare; theory ongoing
9 / 39
![Page 10: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/10.jpg)
Strategic modeling: Platform pricing
Platforms:▶ Earn a (fixed) fraction γ of
every dollar spent (e.g., 20%)▶ Need both drivers (supply)
and passengers (demand)▶ Use pricing to align the two sides
Load-dependent pricing:If # of available drivers = A, then price offered to ride = P(A)
10 / 39
![Page 11: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/11.jpg)
Strategic model: Platform pricing
In practice:▶ Platforms charge a time-
and distance-dependentbase price
▶ Platformsmanipulateprice through amultiplier
▶ Base price typicallyis not varied
In our model:price≡multiplier.
11 / 39
![Page 12: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/12.jpg)
Strategic model: Passengers
How do passengers enter?▶ Passenger≡ one ride request▶ Sees instantaneous ride price▶ Enters if price< reservation value V▶ V ∼ FV, i.i.d. across ride requests
µ0 = exogenous rate of "app opens".µ = actual rate of rides requested.
Then when A available drivers present:
µ = µ0FV(P(A)).
12 / 39
![Page 13: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/13.jpg)
Strategic model: DriversHow do drivers enter?
▶ Sensitive to expected earnings over the block▶ Choose to enter if:
reservation earnings rate C ×expected total time in system< expected earnings while in system
▶ C ∼ FC, i.i.d. across driversΛ0 = exogenous rate of driver arrival.λ = actual rate at which drivers enter.Then:
λ = Λ0FC
(expected earnings in systemexpected time in system
)
13 / 39
![Page 14: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/14.jpg)
Queueing model
1. Drivers enter at rate λ.2. When A drivers available,
ride requests arrive at rate µ(A).3. If a driver is available, ride is served; else blocked.4. Rides last exponential time, mean τ .5. After ride completion:
▶ With probability qexit: Driver signs out▶ With probability 1− qexit: Driver becomes available
14 / 39
![Page 15: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/15.jpg)
Queueing model: Steady state
Jackson network of two queues: M/M(n)/1 and M/M/∞
=⇒ product-form steady state distribution π.
15 / 39
![Page 16: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/16.jpg)
Putting it together: EquilibriumGiven pricing policy P(·),system equilibrium is (λ, µ, π, ι, η) such that:1. π is the steady state distribution, given λ and µ
2. η is the expected earnings per ride, given P(·) and π
3. ι is the expected idle time per ride, given π and λ
4. λ is the entry rate of drivers, given ι and η:
λ = Λ0FC
(η
ι+ τ
)5. µ(A) is the arrival rate of ride requests when A drivers
are available, given P(·):
µ = µ0FV(P(A)).
16 / 39
![Page 17: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/17.jpg)
Putting it together: Equilibrium
If price increases when number of available driversdecreases:
▶ Equilibria always exist under appropriatecontinuity of FC, FV.
▶ Equilibria are unique under reasonable conditions
17 / 39
![Page 18: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/18.jpg)
Large Market Limit
18 / 39
![Page 19: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/19.jpg)
The challenge
▶ To understand optimal pricing, we need to characterizesystem equilibria.
▶ In particular, need sensitivity of equilibria to changes inpricing policy.
▶ Our approach: asymptotics to simplify analysis.
19 / 39
![Page 20: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/20.jpg)
Large market asymptotics
Consider a sequence of systems indexed by n.▶ In n'th system, exogenous arrival rates are nΛ0, nµ0.▶ In n'th system, pricing policy is Pn(·).▶ In each system, this gives rise to a system equilibrium.
We analyze pricing by looking at asymptotics of equilibria.
20 / 39
![Page 21: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/21.jpg)
Static Pricing
21 / 39
![Page 22: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/22.jpg)
What is static pricing?
Static pricing means: price policy is constant.Let P(A) = p for all A.
TheoremLet rn(p) denote the equilibrium rate ofcompleted rides in the n'th system. Then:
rn(p) → r̂(p) ≜ min{Λ0FC(γp/τ)/qexit, µ0FV(p)}.
Throughput =min { available supply, available demand }
22 / 39
![Page 23: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/23.jpg)
Static pricing: Illustration
23 / 39
![Page 24: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/24.jpg)
Static pricing: Interpretation
Note that at any price, queueing system is always stable:▶ When supply < demand:
Drivers become fully saturated▶ When supply > demand:
Drivers forecast high idle times and don't enterBalance price pbal: Price where supply= demand
CorollaryThe optimal static price is pbal.
24 / 39
![Page 25: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/25.jpg)
Dynamic pricing
25 / 39
![Page 26: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/26.jpg)
What is dynamic pricing?
Meant to capture "surge" (Uber)and "prime time" (Lyft) pricing strategies.
We focus on threshold pricing:▶ Threshold θ
▶ High price ph charged whenavailable drivers< θ
▶ Low price pℓ < ph charged whenavailable drivers> θ
26 / 39
![Page 27: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/27.jpg)
Dynamic pricing: Numerical investigation
▶ Fix one price,and varythe other price.
▶ Compare tostatic pricing.
n = 1
27 / 39
![Page 28: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/28.jpg)
Dynamic pricing: Numerical investigation
▶ Fix one price,and varythe other price.
▶ Compare tostatic pricing.
n = 10
27 / 39
![Page 29: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/29.jpg)
Dynamic pricing: Numerical investigation
▶ Fix one price,and varythe other price.
▶ Compare tostatic pricing.
n = 100
27 / 39
![Page 30: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/30.jpg)
Dynamic pricing: Numerical investigation
▶ Fix one price,and varythe other price.
▶ Compare tostatic pricing.
n = 1000
27 / 39
![Page 31: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/31.jpg)
Dynamic pricing: Numerical investigation
▶ Fix one price,and varythe other price.
▶ Compare tostatic pricing.
n → ∞
27 / 39
![Page 32: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/32.jpg)
Dynamic pricing: Numerical investigation
▶ Fix one price,and varythe other price.
▶ Compare tostatic pricing.
n → ∞
27 / 39
![Page 33: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/33.jpg)
Optimal dynamic pricing
TheoremLet r∗n be the rate of completed rides in the n'th system, usingthe optimal static price.
Let r∗∗n be the rate of completed rides in the n'th system, usingthe optimal threshold pricing strategy.
Then if FV has monotone hazard rate,
r∗n − r∗∗nn → 0 as n → ∞.
28 / 39
![Page 34: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/34.jpg)
Optimal dynamic pricing
In other words:
In the fluid limit, no dynamic pricing policyyields higher throughput than optimal static pricing.
This result is reminiscent of similar results in the classicalrevenuemanagement literature (e.g., [Gallego and van Ryzin,1994]).
The main differences arise due to the presence of a two sidedmarket.
29 / 39
![Page 35: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/35.jpg)
Optimal dynamic pricing
In other words:
In the fluid limit, no dynamic pricing policyyields higher throughput than optimal static pricing.
This result is reminiscent of similar results in the classicalrevenuemanagement literature (e.g., [Gallego and van Ryzin,1994]).
The main differences arise due to the presence of a two sidedmarket.
29 / 39
![Page 36: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/36.jpg)
Proof sketch
Under threshold pricing:▶ Drivers are sensitive to two quantities:
idle time, and price.▶ Show that optimal θ∗n → ∞,
but chosen so that idle time→ 0 as n → ∞.▶ In this limit, drivers are sensitive to
the average price per ride:
pavg = πhph + πℓpℓ,
where πh, πℓ are≈ probabilities of beingbelow or above θ, respectively.
▶ If pavg decreases, fewer drivers will enter.
30 / 39
![Page 37: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/37.jpg)
Proof sketch (cont'd)
We note that:1. If pℓ < ph ≤ pbal, then pavg = ph.2. If pbal ≤ pℓ < ph, then pavg = pℓ.3. If pℓ < pbal < ph, then πℓ > 0, πh > 0.
In first two cases, de facto static pricing.
31 / 39
![Page 38: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/38.jpg)
Proof sketch (cont'd)
We explore the third case.Suppose that we start with pℓ < ph = pbal (so pavg = ph).
Now increase ph:▶ Before πℓ = 0, but now πℓ > 0, so some customers pay
pℓ; this lowers pavg.▶ ph higher, so customers arriving when A < θ pay more;
this increases pavg.When FV is MHR, we show that the first effect dominates thesecond, so throughput falls.
32 / 39
![Page 39: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/39.jpg)
Robustness
33 / 39
![Page 40: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/40.jpg)
The value of dynamic pricing
How does dynamic pricing help?▶ When system parameters are known,
performance does not exceed static pricing.▶ When system parameters are unknown,
dynamic pricing naturally "learns" them.
34 / 39
![Page 41: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/41.jpg)
Robustness: IllustrationWhat happens to static pricing in a demand shock?
35 / 39
![Page 42: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/42.jpg)
Robustness: IllustrationWhat happens to dynamic pricing in a demand shock?
36 / 39
![Page 43: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/43.jpg)
Robustness: Dynamic pricing
We can formally establish the observation in the previousillustration:
▶ Suppose FC is logconcave, and µ(1)0 < µ
(2)0 are fixed.
▶ Let p(1)bal,n, p
(2)bal,n = optimal static prices in the n'th system.
▶ Let r(1)n , r(2)n = optimal throughput in the n'th system.▶ Suppose now the true µ0 ∈ [µ
(1)0 , µ
(2)0 ].
▶ Using both prices p(1)bal,n, p
(2)bal,n is robust:
▶ There exists a sequence of threshold pricing policieswith throughput at any such µ0 (in the fluid scaling)≥ the linear interpolation of r(1)n and r(2)n .
(Same holds w.r.t.Λ0.)
37 / 39
![Page 44: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/44.jpg)
Conclusion
38 / 39
![Page 45: DynamicPricingin RidesharingPlatforms · Relatedwork Ourworksitsatanexusbetweenseveraldifferentlinesof research: 1. Matchingqueues(cf.[AdanandWeiss2012]) 2. Strategicqueueingmodels(cf.[Naor1969])](https://reader033.vdocument.in/reader033/viewer/2022042307/5ed37bb6847f87317f77be6c/html5/thumbnails/45.jpg)
Platform optimization
This work is an example of platform optimization:Requires understanding both operations and economics.Other topics under investigation:1. Network modeling (multiple regions):
Our main insights generalize2. Effect of pricing on aggregate welfare3. Modeling driver heat maps4. Fee structure: changing the percentage5. Effect of changing the matching algorithm
39 / 39