Please enable JavaScript.

Coggle requires JavaScript to display documents.

Fundamentals (Time Series Analysis (Time series components (Models (model…

- - - - mean, no other pattern present its constant value
    - - Rate of growth, persistent movement in one direction, typically linear but can be exponential, quadratic
    - - repeated cycle around a known and fixed period, hourly,weekly...,can be caused by natural or man-made forces
    - - Remainder of variability after other components, irregular and unpredictable variations, noise
    - - Periodic movement not of a fixed period, duration can be of different lengths, most often tied to longer term business cycles or economic conditions
    - - model depends on how seasonality impacts trend and or/ level.
      - Level Model: **X(t)=a+e(t)**
      - Trend Model: X(t)=a+bt+e(t)
      - Mix Level-Seasonality Model: X(t)=aF(t)+e(t)
      - Mix level-Trend-Seasonality Model: X(t)=(a+bt)F(t)+e(t)
- - - - 5P: Product & Package/Pricing/Promotion/Place
      - what should we do to shape and create demand for our product
    - - What should we expect demand to be given the demand plan in place
      - Demand forecasting
        
        Strategic, tactical, operational
        Considers internal & external factors (不要把capabilities放入考虑)
        Baseline, unbiased & unconstrained
      - Forecasting levels
        
        Strategic Year/Years business/capacity/investment planning
        
        Tactical Quarterly brand plans/budgeting/sales/ manpower planning
        Months/weeks short-term capacity/master/inventory planning
        
        Operational Days/hours transportation/production/inventory plannning
    - - Balances demand & supply
        Sales & operational planning (S&OP)
        Bridges both side of a firm
  - - - Demand is essentially a continuous variable
      - Every estimate has a error band
      - Forecasts are highly dis-aggregated, typically SKU-Location-Time
      - Things happens...
      - what can we do?
        
        Don't fixate on the point value
        
        use range forecasts
        
        Capture error of forecasts, for future forecasting
        
        Use buffer capacity or stock
      - Use range &
        Check forecast error
    - - Aggregate by SKU, Location, Time
        
        Coefficient of Variation
        (CV)
        
        CV(aggregate)=CV(ind)/(n^1/2),某个店的某个SKU的CV除以DC服务的全部店的数量开根号
        
        大中小咖啡，及其盖子。Aggregate by SKU，如果采用统一盖子，其CV变小。
        经验：在不同的产品上采用通用件可以降低此通用件的CV
        
        Aggregate by Time，咖啡按周或月forecast, 则smooth out/ balance out,更准确, DV更小;
        Forecast with longer time buckets have better accuracy; the time bucket used should match the situation.
        
        Aggregate by Location：三个三明治店，合并为一个DC, CV变小
      - Risk Pooling reduces CV
    - - Postpone customization until as late as possible
      - 三明治店例子，离需求发生的时间点(Time Horizon)越短需求越准确
  - - - Judgemental
        
        Jury of experts
        
        Delphi techniques
        
        Sales force surveys
      - Experimental
        
        Focus group sessions
        
        Test marketing
        
        Customer surveys
    - - Casal/Relational
        
        Leading indicators
        
        Input-Output Models
        
        Econometric Models
      - Time series
        
        Past predicts the future
        
        Identify patterns
        
        "Black Box" Approach
- - - - 3 types
        
        Make to Stock 预做: push-push-push-pull
        
        Make to Order 现做: push-push-pull-pull
        
        Engineer to Order 定做: push-pull-pull-pull
        
        Buy Raw Material-Prepare components-Final assy-Sell product
        
        The last two types give more variety, MTO more efficient
  - - - Homogenous: Within the segment should be similar
      - Heterogeneous: Across segments should be very different
      - Critical Mass: should be big enough to make it worthwhile
      - Pragmatic: dimensions should be useful and commuicable
    - - Geography, Demographic, Channel segmentation
      - Strategic importance, Service level, Order size/volume
      - Lead time, Purchase history, Sales trends
    - - Physical characteristics (values, size, density, value, etc.
      - Demand Characteristics (Sales volume, volatility, sales duration, etc.
      - Supply characteristics (availability, location, reliability, etc.)
    - - True grocery example: 10% SKU account for 50% of volume sold, 50% SKU account for 10% volume sold
      - Truckload lane example:
        50% of volume is handles by 3% of the lanes!
        3% of volume is handled by 43% of the lanes
      - Power law: Y=a.X(k次), Y is cumulative percentage, X is percentage of product,
        掌握计算，用log10,有两组数据即可计算X和Y
        exceptionally common in social and physical systems
        The Critical few versus the Trivial many
        e.g. Pareto Law, 20/80 law, Zipf Law....
        How to treat them differently?!
      - Distribution of many phenomena across a population follow a Power Law relationship
      - ABC Analysis
        or ABCDEF.....
        
        Class B: the middleshare
        many moderate impact items (some times most)
        automated control w/ management by exception
        rules can be made for A (but usually too many exceptions)
        
        Class C: the trivial many
        Many if not most of the items that make up minor impact
        control system should be as simple as possible
        reduce wasted management time and attention
        group into common regions, suppliers, end users
        
        Class A: the important few
        very few high impact items are included
        require most managerial attention and review
        expect many exceptions to be made
        
        1.These are arbitrary classification
        
        Items don't stop at c
        
        计算，例：80%比重内(A)，15%比重内(B)，5%比重(C)
        sum(A:B) 用法
        
        ABC as horizontal axis, and add demand variability (usually stand deviation of mean), treat with different techniques, different frequency reviews
    - - Supply Chain Portfolio
        例：HP printer
        move from innov. to func.
        supply chain matrix change
  - - - Log-normal distribution Λ(μ,σ)
        not common, but good for supply chain
      - Normal distribution N(μ,σ)
        most commonly used
      - Poisson distribution P(λ)
        discrete, integers>=0, used for low valued distribution
        通过λ(即某段时间内观测到的随机事件的频率，区别于正态)，来预测某段时间(t)内该事件发生n次的概率。
        Excel: POISSON（x0,λ,0）概率密度；POISSON（x0,λ,1）概率累积分布，x<=x0的概率；