market potential
an estimate of the maximum possible sales of a good or service for an entire industry during a stated time period

sales potential
refers to the maximum market share that a particular firm can achieve under ideal conditions

sales forecast
an estimate of the dollar or unit sales for a specific future period under a proposed marketing plan or program for an individual firm

NOTE: a forecast is what is realistically expected, not what is hoped or desired

How might you predict demand for:

  • wind generators in 2025
  • Ford Escorts next year
  • size 3a rivets made by Acme Mfg.
  • refrigerators in two years
  • Wendy's menu strips made by VFSign Co.


Controllable Factors
those that are under control of the firm

  • pricing
  • distribution
  • promotion
  • product characteristics
  • product mix
  • account policies
  • choice of customers
  • etc.
Uncontrollable Factors
environmental elements over which the firm has little, if any, direct control
  • economy, interest rates, inflation
  • public policy, government regulation
  • political conditions
  • market factors, changing demographics
  • competitors, competitor actions
  • supplies, supplier actions
  • industry trends
  • etc.


Three Basic Approaches:

  • Judgmental / Qualitative
  • Relational
  • Analytical / Quantitative

Judgmental / Qualitative Techniques

  • subjective; based on a hunch, intuition
  • assume that somebody knows the answer and ask them
  • experience based
  • subjective: might result in bias

Judgmental / Qualitative Techniques

  • Jury of Executive Opinion
  • Sales Force Composite
  • User's Expectation
  • Delphi Techniques
  • Scenario Method

Judgmental / Qualitative Techniques

Useful for:

  • long range forecasting
    • e.g., where technological, political, etc. factors play a significant role
  • when data is limited or non-existent
    • e.g., new product launch

    Relational Techniques

    • assume cause and effect, and cause can be used to predict sales
    • if you know one variable, you can forecast the other

    Relational Techniques

    • leading indicators 
      • e.g., housing starts suggest refrigerator sales
      • e.g., births suggest college enrollments
    • regression techniques
      • assume a straight line; cannot account non-linear sales
      • in some cases, assumes a causal relationship between time and sales (don't repeat this one on a stats exam!)
      • use a ruler for "eyeball regression"

    Analytical / Quantitative Techniques

    time series approaches

    • assume that historical data can be used to predict future demand
    • all we look at is historical data over time used to reduce the element of subjectivity
    used to describe a time series that is not flat 

    used to describe a time series as flat 

    Analytical / Quantitative Techniques

    Four Approaches:

    • naive
    • cumulative mean
    • moving average
    • exponential smoothing
    NOTE: cumulative mean is mentioned to develop insights into these methods and is generally not a method that is used in practice.

    Idea behind what we will be doing:

    • we want to smooth the data
    • we want to find the pattern in the noise


    St+1 = St
    • cumulative mean looks at all data
    • naive approach looks at no data past the present
    • forecast for the next period is the same for the last period
    • works best when data follows a "random "walk" or is very noisy
    • best in the short run, not so good in the long run
    • does not work with data that is trended or has a clear pattern
    • assumes high volatility


            S1 + S2 + . . . + St
    St+1 =  --------------------
    • assumes that all data are equally relevant
    • never throw anything out
    • not frequently used


    period  sales    forecast
       1    16,250     --
       2    17,000   16,250
       3    20,000   16,625
       4    16,000   17,750
       5    15,000   17,312
       6    17,250   16,850
       7    18,000   16,917
       8    20,000   17,071
       9      --     17,438


            St + St-1 + St-2 + . . . + St-(N+1)
    St+1 =  -----------------------------------
    • we want to try to "average out" the forecast to cancel out noise
    • looks for some sort of trend up or down; attempts to smooth out the trend, but always lags behind the trend
    small N
    the forecast will quickly respond to changes, but we lose the "averaging out" effect which cancels out noise

    large N
    we get good averaging out of noise, but poor response; sluggish

    N is usually chosen by trial and error. Whatever has worked the best in predicting past data is presumed to be the best for predicting the next period. 


    Note: a "period" is some amount of time. It could be a year, a month, a week, an hour, or a millisecond.

    period  sales   period  forecast
       1    16,250    --    
       2    17,000    --    
       3    20,000  53,250  
       4    16,000  53,000  17,750  (53,250/3)
       5    15,000  51,000  17,667      .
       6    17,250  48,250  17,000      .
       7    18,000  50,250  16,080      .
       8    20,000  55,250  16,750      .
       9      --            18,417  (55,250/3)


    S^t+1 = aSt + (1-a)S^t
      a is a smoothing constant
    naive forecast acts as though only the most recent observation has any forecasting value; all prior observations are treated as worthless

    cumulative mean procedure ignores the age of the observation; all observations are treated as equally relevant, no matter how old the observation

    moving average acts as though the last N periods of data are equally useful but that all prior observations are worthless 

    It might seem reasonable that historical observations gradually lose their value rather than so abruptly as in the moving average.

    This idea leads to the concept of weighted moving averages. 


    • assumes that the most recent data is the most valuable
    • assumes that data gradually loses its value over time
    • similar to moving average except:
      • most recent sales are weighted more heavily
      • older sales weighted less
    S^t+1 = aSt + (1-a)S^t
      a is a smoothing constant
      large a
      • fast smoothing
      • heavy emphasis on new data
      • highly responsive but "nervous" to noise
      small a
      • slow smoothing
      • heavier reliance on older data
      • sluggish response but calm to noise
    The value of the smoothing constant is usually chosen by trial and error. Whatever has worked the best in predicting past data is presumed to be the best for predicting the next period. 



    a = .8
    period  sales   forecast
       1    16,250  
       2    17,000  16,250  (use naive to seed)
       3    20,000  16,850  (.8)(17,000) + (.2)(16,250)
       4    16,000  19,370  (.8)(20,000) + (.2)(16,850)
       5    15,000  16,774  (.8)(16,000) + (.2)(19,300)
       6    17,250  15,335               .
       7    18,000  16,867               .      
       8    20,000  17,773               .
       9      --    19,555  (.8)(20,000) + (.2)(17,773)

    | return to syllabus | return to homepage |