Beruflich Dokumente
Kultur Dokumente
Kalyan Talluri
13 January 2020
About myself
• Kalyan Talluri
• kalyan.talluri@imperial.ac.uk
• Room No:AC 3 L1, Room No. 3103
• Tel: 2318-7375
• Professor of Operations Management, Imperial College Business School,
London
• Director, MSc in Business Analytics
• Background
• Ph.d Operations Research (MIT)
• Worked at an airline (Scheduling, Revenue Management)
• Taught at various B-schools, UPF (Economics Department) in Barcelona
• Research interests: Pricing and Revenue Management, Transportation,
Service Analytics
PPRO Part 1
1
Prep notes that I expect you to read on your own.
About this course
This is a data & quant course, but . . . not just about numbers:
• Typical data we deal with:
• Operational data: Current and historical sales with timings, capacities
• Variable (marginal) costs2
• Industry/competitior data
• Other co-variates (special days future conditions, current conditions)
• Models
• Customer behavior
• Firm objectives and processes
• Industry (competition)
• Objectives: Profit, revenue, break-even, penetration, short-term/long-term
• Output:
• Design of products: mechanisms, conditions (eg: auctions or
posted-prices? refunds, cancellation penalties? . . . , in general design of
conditions of sale)
• Price/prices → Numbers potentially changing over time
• Assortments (with fixed prices) or allocations of a fixed inventory to the
products (airlines, hotels)
2
As we shall see, fixed or sunk costs are usually irrelevant for us operationally
Data for pricing
This is the central source, but we use many auxiliary sources in practice (weather,
economy, competitor prices, etc.)
What this course is about . . .
Need a simple model of how a customer purchases (assume customer buys at most
one unit). Think auctions!
3
Single laptop per group with latest Adobe Flash, 11.5—(a) you have to Allow Flash (b) allow images (c)
allow pop-ups. In Chrome, click on the button next to the URL
Once you register
Main reasons
• Heterogenous customers
• different valuations (however, often is private information)
• different purchase preferences
• Supply constraints
• Valuations changing over time
• Seasonality (peak vs. off-peak)
• Product life-cycle
• New products
• Competitor pricing
What is RM? Two flavors
HW1 individual
• Go through notes in the pricing concepts; Solve the sample exercises at the end
• Read and answer the Vertigo case in case packet
HW1 group
• Data but essentially a qualitative exercise—design RM products
• If time permits I will ask a few teams to discuss in class for 5 minutes each (can
give feedback and comments)
Last class: Two flavors of Differentiated
Pricing
Examples
• Round-trip tickets are much cheaper (20 to 30 %) if your trip crosses a
Saturday night
• Business customers (consultants, high WTP) fly to client-base on Monday
and return by Friday
• Leisure customers (low WTP) holiday combining weekends
• Note: Sold in parallel; customers self-select
• Flexible tickets cost 2 to 5 times inflexible tickets (no changes, cancellation fees)
Market
Segment 1
A person can potentially
belong to multiple segments
Segment 2
Person 1 Person 1
Person 1
Person 1
Person 2 Segment 3
Person 2
Segmentation is an art
Purchase behavior
Channels used
Flexibility
Information about their own usage
Urgency/need
Loyalty
Behavioural Geographic
Lifestyle
Dislikes
Social
Likes
Psychographic Demographic
What to do if the product cannot be
changed?
Be “creative”: Revenue Management
(RM) segmentation
Dilution and Fences
Dilution: High-WTP customers buy cheaper products
We need “Fences”
• Mechanisms to prevent dilution
• Distinguishing feature of RM products
• Absolutely essential to prevent dilution
Question: What are the fences in this typical (non-RM) product line?
Concept: Incentive Compatible Pricing
Net-utility model of customer
max[v1 − p1 , v2 − p2 ]
(If this is ≥ 0)
Designing product menus: Incentive
Compatibility
A key economic concept to keep in mind in product design with multiple products
Understanding customer segments and our products, features and their prices is a joint
exercise!. Difference in prices cannot be greater than difference in values
Perfect and imperfect segmentation
Imperfect Segmentation
• Dilution A fraction of the high segment buys the lower-priced product
• Buy-up We shut off the lower-priced product—a fraction of the low segment
buys up
No segmentation based on self-selection is “perfect” — however, if the restrictions are
strong, we can assume perfect segmentation
15-min Exercise
Design and match
Based on the description, (a) identify segments (b) create products (c) price (d) match
SEGMENTS PROUCTS
Features
Public channels
Compares Name:
No affiliations Label: Transient Price:
Package
Business
Below are the two main avenues. Challenge for for both is to identify “like-for-like”
and determining (a) when to draw an alarm (b) if so finding the root-cause.
• Internal reports and metrics comparing against previous periods/markets.
• Bench-marking reports: compare against industry and competitors.
Key metrics used to control at an aggregate level
• Occupancy: % of capacity being filled (at various levels of operations).
Alternate industry names: Load-factor or Yield
• Average revenue per available unit of capacity. Alternate industry names:
Revenue per seat, RevPAR (Revenue per Available Room)
• Average rate: Average price of each unit sold. Alternate industry names: ADR
(Average Daily Rate).
• Competitor index4 : market-share, yield compared to the competition etc.
Various industry names.
Plus a host of others: website views, searches and metrics.
4
Market research firms supply these numbers in many industries. They are often self-reported on a
participatory basis—you get info on the competitor set if you tell us your own performance.
Operational control: Real-time control of
sales of RM products
Products are sold via third-party channels, prices are set by algorithms, so we have to
be in control
• Control system (PMS, Res system) allows automation (rules and processes
coded up)
• Decouples the Forecasting + Optimization stages with the actual Sales Control:
Cannot forecast/optimize after every sale (takes 30 to 90 mins to run the
methods) so sales are controlled by some optimized controls, and the
optimization happens occasionally (say during the night)
• Popular control methodologies (Note: Prices are fixed – used only for
accept/deny decisions)
• Bid-price controls
Decision rule
Sell if [Price ≥ Expected marginal value of the last remaining inventory
unit]
No, otherwise
• Nested allocation controls/Nested protections
Understanding Nested limits
Nested: inventory available to the cheaper product is available to
the more expensive product
• Consider the following Nested Booking Limits for three classes and a capacity of
30; fares Class 1: $100, Class 2: $75, Class 3: $50
Class 1 30
Class 2 18
Class 3 8
• Say the booking requests come as follows during the day:
333333333322222111111111111
3322222222222222222222211
33223322233222222222333111111