Module 6 Flashcards
LO: What are unique forecasting and inventory management challenges for innovative products?
TBD
LO: How do the dynamics between production/distribution lead-time and demand forecast accuracy influence opportunities for inventory replenishment?
TBD
LO: What is the “newsvendor model” and how can we use it to make single period inventory decisions?
TBD
What is Fast strategy?
- MAKE-TO-STOCK (Example: Zara)
- competitive cost and continuous portfolio renewal (short time “from idea to market”) and affordable cost
- functional and short-lifecycle products
- Demand is “pushed” by a collection” forecast, where fast product development is a critical capability
- applied by “fashion creators” in industries such as apparel and beauty products
What supply chain strategy is this?
What is the Custom-Configured strategy?
- CONFIGURE/ASSEMBLE-TO-ORDER (Example: Room & Board furniture)
- focus on product configurability (customization with limited options for several features)
- not having to forecast for all products
- forecast is for workload before product decoupling, with extra capacity and design for easy assembly
What supply chain strategy is this?
What is the Agile Strategy?
- HYBRID: MAKE-TO-ORDER/MAKE-TO-STOCK (Example: Industrial Suppliers)
- job-shop style, responsiveness to unpredictable demand
- “exclusive” and short-lifecycle products
- workload has medium-size peaks and valleys
- less asset utilization
- extra capacity and common components for agility
What supply chain strategy is this?
What is the Flexible Strategy?
- CONFIGURE/DESIGN-PROCESSES-TO-ORDER (Examples:
- making, configuring, and designing to order
- price is irrelevant due to uniqueness
- capacity on standby, capacity pooling, reconfigurable
- low volume, super specialized, relationships are shorter-term
What supply chain strategy is this?
What things do we need to think about for forecasting for innovative products?
- options for demand forecasting
1) may need forecast coverage of the full demand
2) in other case, production lead time is quick, and can do rapid replenishment based on initial sales
What do we need for demand forecast of innovative products?
- Probabilistic forecast requires (not just a point prediction)
- Forecast accuracy improves with time (as sales history builds)
- Aligns forecasts and ordering decisions
- Initial/Pre-launch forecast
- Updated forecast (based on early sales)
What are the tools for developing pre-launch forecast?
- Market research (consumer based, A/B testing)
- Causal-based Forecast Engine
- Expert Opinion (e.g., Committee-based) Forecast Engine
- Like-SKU-based Forecast Engine
What is Causal-based Forecast Engine?
Causal Modeling uses independent explanatory variables to predict future demand.
May use statistical regression and/or machine learning techniques
1) external factors drive demand in a systemic, predictable way
2) data on these external factors are available and easily updated
What is Expert Opinion Forecast Engine?
Solicit information from an individual or panel of “experts” to inform the demand forecast.
Techniques differ by:
- types of “experts” used
- procedures for collecting information (individual vs consensus)
- steps for reaching final forecast (type of feedback between iterations, discussion vs no-discussion, number of iterations)
What was the key takeaway from the Expert opinion example for clothing items in the lecture?
When the Committee agrees (scores have lower std dev), they tend to be accurate (forecast error).
- high agreement = low error
- low agreement = high error
What is Like-SKU-based forecast engine?
Uses data on the demand history of like-SKUs to generate demand forecast.
Qualitative, in selecting the right Like-SKU
Helpful for:
- estimating lifecycle curve - timing of peak demand and rate of demand decline
- predicting impact of launch date and marketing plans
- rolling aggregate demand forecast to lower level (National -> region, region -> store, style -> size)
(basically scaling down to the specific situation)
What was the key takeaway of music industry forecast example in lecture?
It was an example of Like-SKU-based forecasting
Rap-music example:
26% of total demand in 1st week, 40% of total demand in 2nd week - RAPID growth in beginning, falls out quickly with long tail
Latin-music example:
smaller increase
LO: Once we have an initial, pre-launch, demand forecast, how should we plan to utilize early sales information?
TBD
What are some qualities of early sales?
- highly predictive
- expert committee average forecast error is 55%
- first 2 weeks of sales average forecast error is 8%
How is early sales used?
1) Initial forecast (gut feel) and read for a period time, quick reaction - low accuracy
2) read market (high demand vs low demand), but need to consider lead time. May need to get shorter LT to fit them in the demand for the season
3) Replenishment to carry through season
What do you if there is only one order opportunity?
no opportunity to replenish
Use the newsvendor model
What is the purpose of the newsvendor model?
A method to forecast items that can’t be replenished during the sales cycle (1 day in newsvendor). Might have cycle stock and safety stock.
What information does a newsvendor need?
- probabilistic demand forecast
- goal for product (maximize profit or specific service goal?)
- profits influenced by order decisions
- mismatch costs (cost of inv overage & cost of inv underage)
- customer goodwill loss etc
What is the probabilistic demand forecast?
time series plot of demand, plotted on demand history is a normal distribution
Normal distribution is used to approximate the demand.
- average demand
- std dev is variation
How do you make the single order decision?
Probabilistic Demand forecast
Performance Objective
Underlying Costs (or Target Measures)
How do you order to meet a service target for single orders?
Mean demand = 100
Std Dev demand = 20
in-stock rate = 0.90
1) How much should we order to support an in-stock rate of 90%?
2) If current policy is to set the order quantity at 120 units, what is the in-stock rate?
Probability of 90%, cumulative probability
P(D<=Q) = .90
1) Excel: =NORM.INV(cum prob, mean, std dev) =NORM.INV(0.9, 100, 20) = 125.65, round up to 126
Z table Z = 1.32 (at .9066) Q = Mean + std dev * Z = 100 + 20(1.32) = 126.4, round up to 127
2)
Q = M + SigmaZ
120 = 100 + 20(Z)
z = 1, P = 0.8413
=NORM.DIST(quantity, mean, std dev, 1)
=NORM.DIST(120,100,20,1)
= 0.8413, or 84.13%
How do you maximize profit in single order/period context?
Minimize market mediation costs
- cost of excess inventory: overage cost, Co: unit cost of overage
- cost of stock-out: underage cost, Cu: unit cost of underage
Increase Q until expected marginal underage and overage costs are equal:
P(D > Q)Cu = P(D <= Q)Co
Rearranged, largest Q that satisfies the rule
P(D<=Q) = Cu/(Cu+Co)
Fraction is the target in-stock rate that maximizes expected profit for the newsvendor.
What are the market mediation costs for single order?
- cost of excess inventory: overage cost, Co: unit cost of overage
- cost of stock-out: underage cost, Cu: unit cost of underage
How do you find the order quantity that maximizes profit?
Scenario 1:
Mean demand = 100
Std Dev demand = 20
newspaper cost = $2
newspaper price = $6
Rule” Choose Q* so that {(D<=Q*) = Cu/(Cu+Co)
Cu = $6 - $2 = $4 per unit Co = $2 per unit
P(D<=Q) = 4/(4+2) = 2/3
Q with in stock probability of 2/3 -> .67
Stats Table - .67 -> 0.44 Z score
Z = 0.44
Q = M + Z*Sigma
= 100 + 0.44(20)
= 108.8, 109 units
=NORM.INV (2/3, 100,20) = 108.6, 109 units
How do you find the order quantity that maximizes profit with credits?
Scenario 2:
Mean demand = 100
Std Dev demand = 20
initial cost of news paper = $3
newspaper price = $6
remaining inventory at end of day = $2 credited back
Cu = $6 - $3 = $3 Co = $3 - $2 = $1
P(D<=Q*) = 3/(3+1) = 3/4 -> .75
Stats Table - .7517 -> 0.68 Z Score
Z = 0.68
Q* = 100 + 20(0.68)
= 113.6, 114 units
=NORM.INV(3/4,100,20) = 113.5, 114 units
How do you maximize profit?
- determine the underage and overage costs
- compute the target in-stock rate that maximizes expected profit
- use order maximizing formula
What in-stock rate is associated with the profit maximizing prder quantity?
Profit Maximizing In-Stock Rate = Cu/(Cu+Co)
How do you compute the expected profit level (E[Profit]) for a given order quantity decision?
E[Profit] = (Price - Cost) x E[Sales] - Co x E[Leftover Inventory]
How do you computer E[Sales]? Is it the same as E[Demand]?
E[Sales] = E[Demand] - E[Lost Sales]
= mean - std dev*L(z)
L(z) is the loss function
No.
E[Sales] will be less than E[Demand]
Not all the demand will turn into sales
How do you compute E[Leftover Inv]?
E[Leftover Inventory] = Q - E[Sales]
What is the expected profit E[profit]
Scenario 1:
Mean demand = 100
Std Dev demand = 20
newspaper cost = $2
newspaper price = $6
Q* = 109 units Co = $2 Cu = $4 In-stock rate = 2/3 Z = 0.44
E[Profit] = (Price - Cost) x E[Sales] - Co x E[Leftover Inventory]
E[Sales] = E[Demand] - E[Lost Sales] = mean - sigma*L(z) for z = 0.44, L(z) = 0.217 = 100 - 20(0.217) = 95.66 (don't round)
E[Leftover Inv] = Q - E[Sales]
= 109 - 95.66
= 13.34
E[Profit] = (6 - 2) x 95.66 - 2 x 13.34
= $355.86
What are the performance measures for innovative products
What are their formulas?
How would you you describe them?
find in-stock rate, order quantity (Q)
E[Lost sales]: sigmaL(z)
E[Sales]: E[Demand] - E[Lost Sales] = mean - sigmaL(z)
E[Leftover inventory]: Q - E[Sales] = Q - (mean -sigma*L(z))
E[Profit]: E[Profit] = E[Profit] = (Price - Cost) x E[Sales] - Co x E[Leftover Inventory]
E[Lost sales]: average # of units demand exceeds the order quantity
E[Sales]: average # of units sold
E[Leftover inventory]: average # of units left over at the end of the season/period
E[Profit]: depends on situation
What decisions can the newsvendor model help to answer?
- single or initial production/order quantity decision for an innovative product
- repeated order quantity decisions for perishable products (newspaper, food)
- capacity decision for new plant
- Lead time quote