Lecture No. 7
Held on: Wednesday, August 16, 2000
Notes Prepared by: Mukund Kumar & Gaurav Mangla
Subjective forecasting methods
Most of the forecasting decisions are based on past data/ demand. But this is not very accurate- a more accurate method would be to conduct consumer surveys etc.
A
look at various methods employed:
1.
Sales
force composite
·
Estimate
to be made by the Manager ( it could be pessimistic, most likely, optimistic etc.), but the drawback is that this
method is highly dependant on the judgement of the person concerned and hence
unreliable, e.g. the person might under project and then later claim bonuses
based on the supposedly “better” performance! Thus, there is a risk of giving
inaccurate numbers.
·
Customer
surveys- these should include people from different spheres, and biases in the
questions should be avoided
2.
Jury
of Executive Opinion- this exercise involves getting together the executive
heads of the company, who will see the past trends, but are unlikely to analyze
them closely.
3.
Delphi method
This involves posing questions to the experts in isolation, get their feedback, analyze the same and arrive at points of consensus. The idea of posing questions in isolation is to encourage people to come up with unbiased ideas, which they might otherwise hesitate to come up with in presence of more domineering members. Finally, the pool of ideas is recirculated to achieve some common and agreeable ideas.
Examples
of business decisions that rely on forecasting
a.
Cost
and revenue forecast for tax planning
b.
Recruitment
planning by human resources
c.
Cash
flow management
d.
Production
and planning
-
Capacity
planning
-
Inventory
planning
-
Shopfloor
activity planning
Time period of a forecast
1. Short term forecasts (daily-weekly)
It requires a high level of detail
Methods used are for large numbers and should be
thus quick and inexpensive.
High frequency of use
Data storage requirements should be modest
2.
Medium
term (monthly-bimonthly)
-
Forecast
typically aggregated by product type
-
Details
are not needed
-
Extra
cost and effort can be employed
3.
Long
range plans
-
For
capacity, planning, technology etc.
-
Inherent
inaccuracy in forecast
-
Subjective
procedures needed in addition to quantitative techniques
-
Extrapolative
and causal methods used
Extrapolative methods
® used for various situations, i.e.
a. Horizontal trend- sale of product in mature phase of life cycle with random fluctuations around constant mean
b.
Trends-
used for continuously increasing/decreasing demand.
c.
Seasonalities
d.
Cyclical
effects- depends on the business cycle
Moving averages technique
It
assumes a horizontal trend, i.e.
F t+1
= Dt + D t-1 + …… D t-N+1
N
The tradeoff in N is that larger the value of N, the more we will be able to eliminate the random fluctuations, but equally, our trend will lag by a greater amount.
Weighted moving averages
F t+1 = C(t) D(t) + C(t-1) D(t-1) + ….. C(t-N) D(t-N)
The coefficient C(t-i) are determined by least squares of errors considerations, but the drawback with this method is that it requires large storage space.
Exponential smoothing methods
St = S t-1 + a(Dt – S t-1)
= (1-a) S t-1
+ a Dt
St = forecast at time t after observing demand at time t
St-1 = best estimate at time t-1
Typically a Î (0, 0.2)
St = a å (1-a)j D t-j + (1-a)t s0
Here, more weight is given to recent readings and less weight to previous readings.
Higher the value of a, more the weight towards previous readings.
The storage requirements of this method are much less.