Understanding Sample Usage and Sampling as a Promotion Tool: State of Industry Practice and Current Research - Innovation and Marketing in the Pharmaceutical Industry - page 515

would have a positive effect on her future prescription, as demonstrated in existing
studies on the effect of sampling in the pharmaceutical industry. On the other hand,
if the main objective is to provide fi nancial assistance or subsidy to an indigent
patient, a null effect of free sample dispensing on a physician's future prescription
of the same brand should be observed. This is similar to the cannibalization effect
as shown in Bawa and Shoemaker ( 2004 ). They fi nd that in general physicians
are more likely to dispense samples to patients who are newly diagnosed, have an
ongoing diagnosis but were prescribed a different drug on the previous visit, or do
not have any insurance coverage. However, the tendency to dispense samples to
each of the above-mentioned types of patients differs considerably across physi-
cians. As for the long run effects, they fi nd that free sample dispensing will induce
future prescriptions if the samples are dispensed to new patients. In addition, they
do not fi nd a signifi cant effect of free samples on future prescription decisions if
dispensed to patients without any insurance coverage.

17.5 Industry Research in Practice

The above section reviewed academic research on pharmaceutical sampling. In
practice, to gain competitive advantages, pharmaceutical companies also conduct
extensive studies on sampling. In general, these studies can be grouped into three
categories: (1) sample modeling at aggregate level; (2) sampling modeling at disag-
gregate level; and (3) sample allocation models. In this section, we provide an over-
view of these three types of studies.

17.5.1 Sample Modeling at Aggregate Level

Aggregate modeling using brand level data is one common way of conducting
sample analysis in industry practices. The objective is to understand and assess
adequacy and effectiveness of sample resource allocation at brand level from a stra-
tegic perspective. One approach is to evaluate how a brand's TWRx/NWRx volume
or market share is infl uenced by sample volume in addition to detailing volume.
Normally a multivariate time series model approach is employed for such analysis
using a brand's own promotion and sales data. This type of analysis is usually con-
strained by the length of the time series. Another aggregate model approach utilizes
competitor promotion and sales data to build a representative brand model by pool-
ing brand level data from all competitors. This approach requires a cross-sectional
time series model by using detailing and sample audit data such as IMS Health IPS
data. Since the mid-1970s, some practitioners started using the seeming unrelated
regressions (SUR) approach to build a brand level promotion response model,
which allows analysis across multiple therapeutic classes. A similar approach is also