May 1, 2016

How to navigate

F for Full Screen 全屏
W for Wide Screen 宽屏
^+ for Zoom in 放大
^- for Zoom out 缩小
^ stands for "Ctrl Key" 控制键
-> or <- Next or Previous slide 浏览 
## png 
##   2
## png 
##   2

Agenda

  • Importance of forecast
    • the best, and the worst scenario
  • What make a forecast effective
    • Purpose of forecast
    • Roles of forecast
    • Rule for effective forecast
  • Tools that help evaluate assumption
    • Effectiveness of sales channel
    • Market facts — purchasing patten, industrial growth etc
    • Opportunities and risks

Make sure the best scenario a real best, the worst a real worst

The cost when the best is not real best

Redemption In December 2011 “Due to overwhelming demand of hot product … redemption of some of our customers’ online orders. We are very sorry for the inconvenience this has caused…”

The cost when the worst is not real worst

Bankrupcy "…unable to move their inventory," … That backlog left the company paralyzed, unable to buy fresh product or pay off its existing debts. Circuit City still owes $118M to HP, plus $116M to Samsung and $60 million Sony

Accuracy is not enough

Even a broken clock is right twice a day

What portray Effective forecasting

Forecasting is an art

Looks at how HIDDEN CURRENTS in the present signal POSSIBLE CHANGES in direction for companies, societies, or the world at large.

quoted from Harvard Business review "Six rules for effective forecasting" by Paul Saffo

Roles of forecasting

  • Identify the full range of possibilities
  • Map uncertainties
  • Able to articulate and defend the logic
  • Account for the opportunities and risks it presents

Rules for distinguishing good sales forecast from bad

  1. Addresses the risk and opportunities
  2. Looks for the S curve
  3. Well addresses the effectiveness of sales channel and purchasing behavior
  4. Facts and assumptions are scrutinized
  5. Proposed actions address the risks or opportunities identified

Points 2 are quoted from Harvard Business review "Six rules for effective forecasting" by Paul Saffo

Build an effective revenue forecast

Notes to the data and the following slides

The following slides are to demonstrate an Iphone shipment forecast for the 4 quarters of 2016. All data are publicly available and downloadable from:

  1. Apple financial report www.apple.com

  2. Industrial or market public data www.statista.com or www.gartner.com

For the detail calculation of the data, please refer to Appendix

Caveat: All of this slide is prepared to demonstrate building/evaluating effective forecast applying the rules of distinguishing good forecast from bad. I have NO any implicit or explicit views on the forecasted numbers.

Effectiveness of sales channel

  • Understand customer purchasing behavior

  • Sales funnel conversion rate

Understand customer purchasing behavior

What impact customer choice, Sasung royalty is higher than Apple? Especially when customer service satisfactory Apple is much higher than Samsung.

Sales funnel conversion rate

Dummy data. Compare it to history, identify weakness at stages.

Identify Market facts

  • Rapid growth is over (see next slides)
  • Higher than industrial growth means overtaking other competitors
  • The limits of cost down (the learning curve and scale of economy limits)

The Rapid growth is over

Growth Rate reached peak at 2015, and goes down gradually.

Identify Assumption

  • Sustainability
  • Realisticity

Sustainability

  • Is it cost efficiency to spur short term growth curve, with investment, when in maturity stage?

  • Is the cost down potential support price competing, on the learning curve?

Realisticity

  • Competitor react and capacity
  • Alignment with strategy
  • Resources available
  • Milestones (or Key actions) support the number
  • Effectiveness of sales channel

To further analyze, an internal confidential information is needed.

Opportunity and Risk

  • Plan B for extraordinary scenario
  • How sensitivity of the underlying assumptions

Simulation

shipment.sim()

# Growth Rate ~ N(9%, 2%)
# Market share 14~18% at 15%,20%,30%,20%,15% probability

Plan B for the extraordinary scenario ?

Sensitivity analysis

## 
## Call:
## lm(formula = iphone.2016 ~ ms.sim + gr.sim, data = df.sen)
## 
## Coefficients:
## (Intercept)       ms.sim       gr.sim  
##      -20363        15523         2256

Every 1% increase in Smartphone industrial growth rate, it will contribute 15,523 units of iphone shipment; 1% in market share, 2,256 units.

Most Important: Keep eyes on the variance

Variance of actual may signal the validness of assumptions made before.

Appendix

Frequently used quantitative forecast techniques

  • Regression
  • Time series
  • Smoothing
  • Markov chain

Regression

Coefficients of linear model

## 
## Call:
## lm(formula = S ~ Q, data = S.df, subset = sel)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -11.660  -5.905  -2.796   5.627  23.748 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  -11.078      2.927  -3.785 0.000617 ***
## Q              1.931      0.136  14.197 1.31e-15 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.128 on 33 degrees of freedom
## Multiple R-squared:  0.8593, Adjusted R-squared:  0.855 
## F-statistic: 201.5 on 1 and 33 DF,  p-value: 1.309e-15

The residual shows a pattern

Variance is not constant, adding weights to the model

## 
## Call:
## lm(formula = S ~ Q, data = S.df, subset = sel, weights = 1/log(Q))
## 
## Weighted Residuals:
##    Min     1Q Median     3Q    Max 
## -5.810 -3.343 -2.296  2.974 13.399 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  -8.2632     2.1667  -3.814 0.000569 ***
## Q             1.8059     0.1168  15.461  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.719 on 33 degrees of freedom
## Multiple R-squared:  0.8787, Adjusted R-squared:  0.875 
## F-statistic:   239 on 1 and 33 DF,  p-value: < 2.2e-16

Predict Q1 and Q2,2016

Growth Curve

Time series - Exponential Smoothing

Decompose time series into trend, seasons and random

Forecasted shipment using exponential smoothing

Markov chain for forecasting market share

  • Key three brands Market share

  • Induce customer loyalty or brand shifting matrix

Key three brands market share

Convert market share to brand loyalties

smartphone.tp
##                Apple     Other    Samsung
## Apple.l1   0.3782093 0.2015956 0.42019508
## Other.l1   0.0374973 0.8663642 0.09613853
## Samsung.l1 0.2506479 0.2030633 0.54628883

Samsung's royalty is better than Apple. the "Other" is made up of more than 10 brands, the high aggregated loyalties are meaningless here.

The market survey seems to support our conclusion

Predicting next period market share

ms.s[,ncol(ms.s)] %*% smartphone.tp
##         Apple    Other  Samsung
## [1,] 13.96294 61.14233 24.89473

Purchasing behavior

Others

  • No any warranty is provided in this document.

  • If you are to redistribute in any form of this document, please contact xizhijie_xm@outlook.com

newpath
## [1] "C:/Users/Administrator/OneDrive/MyDocs/CMA/FCST"
setwd(oldpath)