exploring business models and dynamic pricing frameworks...
TRANSCRIPT
![Page 1: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/1.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Exploring Business Models and DynamicPricing Frameworks for SPOC Services
Zhengyang Song, Yongzheng Jia, and Wei Xu
Tsinghua University
August 24, 2018
![Page 2: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/2.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Outline
1 Introduction
2 Problem Formulation
3 Offline Combinatorial AuctionVCG MechanismVVCA Mechanism
4 Online Combinatorial AuctioniBundle
5 Conclusion
![Page 3: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/3.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
MOOC platforms
Coursera: 3133 courses
EdX: 2293 courses
XuetangX: 1507 courses
![Page 4: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/4.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
How do they generate revenue?
B2C (Business-to-Customer)
Verified Certificates
Specializations
Online Micro Masters
Advanced Placement
B2B (Business-to-Business)
sub-licensing MOOC contents
on-campus SPOC platforms
![Page 5: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/5.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
SPOC services
Figure:
![Page 6: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/6.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Why do we need an auction?
A Bundle of User’s Demand
MOOC contents
Teaching assistant services
SaaS services
Technical supports
However, resources are limited.
![Page 7: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/7.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Notations
[X ]: set {1, 2, . . . ,X}C : number of available courses
N: number of users
K : number of steps for negotiation
Bn,k : the bundle of user n for step k
vn,k : the valuation of user n for his k-th bundle
sn,k,c : number of enrollments for course c in bundle Bn,k
wn,k,c : operational cost for course c in bundle Bn,k
qc : enrollment capacity of course c
xn,k ∈ {0, 1}: whether bidder n wins his k-th bundle
pn,k : the price we charge for bidder n’s k-th bundle.
![Page 8: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/8.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Auction Mechanism Design
Allocation Rule
xn,k = A(Bn,k , vn,k ,R) =
1 Accept
0 Reject∀k ∈ [K ], n ∈ [N]
Pricing Rule
pn,k = P(Bn,k , vn,k ,R)
![Page 9: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/9.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Problem Formulation
maximize:∑
n∈[N],k∈[K ]
(pn,k − dn,k −∑c∈[C ]
ωn,k,c) · xn,k (1)
s.t. ∑k∈[K ]
xn,k ≤ 1, ∀n ∈ [N]; (2a)
∑k∈[K ]
∑n∈[N]
sn,k,c · xn,k ≤ qc , ∀c ∈ [C ]; (2b)
xn,k ∈ {0, 1}, ∀n ∈ [N], ∀k ∈ [K ]. (2c)
![Page 10: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/10.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
VCG Mechanism [PR03]
Allocation Rule:
max∑n∈[N]
∑k∈[K ]
vn,kxn,k
s.t. Constraints (2a) - (2c)
Payment Rule:
pi =∑j 6=i
∑k∈[K ]
vj ,k xj ,k −∑j 6=i
∑k∈[K ]
vj ,kxj ,k
wherexj ,k = arg max
xj,k
∑j 6=i
∑k∈[K ]
vj ,kxj ,k
![Page 11: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/11.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VCG
Bidding
vA,{P} = 5, vB,{Q} = 1, vC ,{P,Q} = 16
![Page 12: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/12.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VCG
Formulation
max 5 · xA,{P} + xB,{Q} + 16 · xC ,{P,Q} (3)
s.t.xA,{P} + xC ,{P,Q} ≤ 1 (4a)
xB,{Q} + xC ,{P,Q} ≤ 1 (4b)
xA,{P}, xB,{Q}, xC ,{P,Q} ∈ {0, 1} (4c)
Allocation
xA,{P} = xB,{Q} = 0, xC ,{P,Q} = 1
![Page 13: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/13.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VCG
Formulation without user C
max 5 · xA,{P} + xB,{Q} (5)
s.t.xA,{P} ≤ 1 (6a)
xB,{Q} ≤ 1 (6b)
xA,{P}, xB,{Q} ∈ {0, 1} (6c)
Allocation without C
xA,{P} = xB,{Q} = 1
![Page 14: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/14.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VCG
Pricing
Pc =(xA,{P} · vA,{P} + xB,{Q} · vB,{Q}
)−
(xA,{P} · vA,{P} + xB,{Q} · vB,{Q}
)= (1 · 5 + 1 · 1)− (0 · 5 + 0 · 1)
= 6
![Page 15: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/15.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Virtual Valuation Mechanism [LS04]
Allocation Rule:
max∑n∈[N]
∑k∈[K ]
(µnvn,kxn,k + λn,kxn,k)
s.t. Constraints (2a) - (2c)
where µ are positive, λn,k is for particular bidder n and bundlek .For example, to ensure bidder n never gets bundle k for a pricebelow p0, set λn,k = −p0.
![Page 16: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/16.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Virtual Valuation Mechanism
Payment Rule:
pi =1
µi
∑j 6=i
∑k∈[K ]
(µjvj ,k xj ,k + λj ,k xj ,k − µjvj ,kxj ,k − λj ,kxj ,k)
− 1
µi
∑k∈[K ]
λi ,kxi ,k
where
xj ,k = arg maxxj,k
∑j 6=i
∑k∈[K ]
µjvj ,kxj ,k + λj ,kxj ,k
![Page 17: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/17.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VVCA
Bidding
vA,{P} = 5, vB,{Q} = 1, vC ,{P,Q} = 16
![Page 18: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/18.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VVCA
Formulation
We assign the following λ, µ:
µC = 0.5, λB,{Q} = 1
Now the integer programming would become:
max 5 · xA,{P} + xB,{Q} + xB,{Q} + 0.5 · 16 · xC ,{P,Q}
s.t. Constraints (4a) - (4c)
Allocation
xA,{P} = xB,{Q} = 0, xC ,{P,Q} = 1
![Page 19: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/19.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VVCA
Formulation Without C
Without the presence of C , we have:
max 5 · xA,{P} + xB,{Q} + xB,{Q}
s.t. Constraints (6a) - (6c)
Allocation without C
xA,{P} = xB,{Q} = 1
![Page 20: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/20.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Example - VVCA
Pricing
p′C =
1
µC
(xA,{P} · vA,{P} + xB,{Q} · vB,{Q} + λB,{Q}xB,{Q} · vB,{Q}
)− 1
µC
(xA,{P} · vA,{P} + xB,{Q} · vB,{Q}λB,{Q}xB,{Q} · vB,{Q}
)=
1
0.5(1 · 5 + 1 · 1 + 1 · 1 · 1)− 1
0.5(0 · 5 + 0 · 1 + 1 · 0 · 1)
= 14
Thus the revenue of VVCA mechanism would be 14, which ismuch higher than the revenue of VCG mechanism, i.e., 6.
![Page 21: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/21.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Business Process in MOOC Industry
Algorithm 1: Negotiation between user n and the platform
1 Initialization: Set t = 1 and flag = 0. Suppose the currentstatus of resource capacity is R.
2 while t ≤ T do3 (a) User n submits his bids (Bn,k , vn,k) to the platform.4 (b) The platform calculates xn,k and pn,k , and sends the
response message to the user.5 (c) If accepted, then the negotiation succeeds, update R,
set flag = 1, and break. Else (i.e. rejected) thenegotiation continues with t = t + 1.
6 end7 If flag = 0, then the negotiation fails.
![Page 22: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/22.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
iBundle [PU00]
maintain ask prices and provisional allocationbid is competitive if it is not lower than ask pricebidder is competitive if he has at least one competitivebids
Algorithm
for each round, bidders submit bids on bundles
provisional allocation computed to maximize seller’srevenue
terminate if each competitive bidder receives a bundle inthe provisional allocation
o.w., ask prices are increased by a preset parameter,feedbacks are provided to bidders
on termination, provisional allocation becomes the finalallocation, the bidders pay their final bid prices.
![Page 23: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/23.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Conclusion
What we have done
Model formulation for SPOC services
Mechanisms for Offline combinatorial auction
Mechanisms for Online combinatorial auction
Future Work
Compare different mechanisms by simulation
Real Data Analysis of SPOC Services
![Page 24: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/24.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Reference
[LS04] Anton Likhodedov and Tuomas Sandholm. Methodsfor boosting revenue in combinatorial auctions. InAAAI, pages 232–237, 2004.
[PR03] Aleksandar Pekec and Michael H Rothkopf.Combinatorial auction design. Management Science,49(11):1485–1503, 2003.
[PU00] David C Parkes and Lyle H Ungar. Iterativecombinatorial auctions: Theory and practice.AAAI/IAAI, 7481, 2000.
![Page 25: Exploring Business Models and Dynamic Pricing Frameworks ...people.iiis.tsinghua.edu.cn/~weixu/Krvdro9c/dmmooc... · A;fPg v A;fPg + x B;fQg v B;fQg B;fQgx B;fQg v B;fQg = 1 0:5 (1](https://reader033.vdocuments.mx/reader033/viewer/2022052015/602c4a5b4393dd40903476dc/html5/thumbnails/25.jpg)
Introduction
ProblemFormulation
OfflineCombinatorialAuction
VCG Mechanism
VVCAMechanism
OnlineCombinatorialAuction
iBundle
Conclusion
References
Thanks!