Development of a DSS using Stratified PPSWR to
Estimate the Retail Sales and Market Share of a Company
Student Name: R. Ashwin Kumar
Monthly retail sales and market share
of a company was proposed to be estimated by observing these variables for
a sample of dealers of the company. Since the wholesale purchase patterns
of the universe of dealers of the company was available, it was reasoned
that exploiting this information in the sampling design would be more
beneficial, instead of a simple random sampling. In particular, it was
reasoned that dealers making large wholesale purchases would contribute
more to the retail sales (of all brands) and thus one should include such
dealers in the sample with larger probability. This gave rise to the idea
of PPSWR sampling design. But it was reasoned that the frequency of purchase
should also be factored into this equation (because dealers making large
wholesale purchases every month, is in a different category compared to dealers
making same kind of purchase, say, once in only three months) to stratify
the dealers. Thus the dealers were stratified according to their frequency of
wholesale purchase using cluster analysis, and their ``sizes'' were calculated
according to their average monthly wholesale purchase amounts, for implementing
the Stratified PPSWR sampling design to estimate the monthly retail sales and
market share. However since intrinsically the problem was dynamic in nature
with respect to the dealer profiles, this M.Tech project was conceived, which
supplemented the above theoretical work, which arose as a consulting project,
by developing a fully automated Decision Support System for maintenance and
upgradation of the dealer database, stratification of the dealers using
Hartigan's $k$-means clustering algorithm, optimal sample size allocation to
each stratum, drawing of the sample by accommodating various logistic
constraints, and finally providing the required monthly estimates.