`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`An Exploratory Study to Improve Sales Operations
`When Selling Multiple Prescription Drugs
`
`John C. Yi *
`Saint Joseph’s University, Philadelphia, PA, U.S.A.
`
`Ming Zhou
`Taeho Park
`San Jose State University, San Jose, CA, U.S.A.
`
`This paper explores the importance of integrating knowledge with quantitative modeling process to
`improve sales operations in multiple product selling situations in the pharmaceutical industry. A
`knowledge-based approach is proposed to minimize challenges in detailing multiple products to
`physicians who are more and more difficult accessing in recent years. The performance of this new
`approach is compared against the traditional approach via actual implementation by the firm that is
`sponsoring the research. Results based on three months of implementation indicate that the
`knowledge-based approach performs significantly better with increasing the number of responsive
`physicians by 71% and profit by 9%.
`
`*Corresponding Author. E-mail address: jyi@sju.edu
`
`I. INTRODUCTION
`
`The pharmaceutical industry has faced a
`number of challenges in the recent years, with
`many branded drugs going off patent without
`enough blockbuster drugs in the pipeline to
`replace them (PricewaterhouseCoopers, 2008). In
`addition, the industry has received a lot of
`negative press from both the government and
`consumers for the aggressive investment in to
`their sales and marketing efforts (Gagnon and
`Lexchin, 2008; Washington Post, 2002).
`Obviously, the industry needs to find a way to
`better utilize their sales and marketing spending
`to fend off some of these challenges.
`Sales
`force
`is
`the most expensive
`marketing
`investment
`that a pharmaceutical
`company can make. The primary function of
`sales force is to provide detailing to their target
`physicians. The target physicians are those who
`already prescribe or have potential to prescribe
`
`the firm’s prescription drugs; detailing involves
`pharmaceutical sales representatives visiting each
`of their physicians to disseminate the latest
`information on the firm’s prescription drugs that
`is meaningful to the physician’s specialty and the
`patients he or she is treating. The detailing is
`done with the goal of encouraging the physician
`to correctly prescribe the firm’s drugs for those
`patients who fit the diagnostic criteria, and given
`a
`similar
`treatment
`situation where
`two
`prescription drugs are equal in providing help to
`patients, the firm assumes that the sales rep’s
`selling capability would sway the physician to
`prescribe their product. With a heavy price tag of
`$150 to $200 per detail, companies put a
`significant effort into determining the right
`physicians to target, the order of the details, also
`known as detailing sequence, when multiple
`products are involved, and the frequency of
`details to the targeted physicians over time
`(Gagnon and Lexchin, 2008).
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`43
`
`Page 1 of 15
`
`ACRUX DDS PTY LTD. et al.
`
`EXHIBIT 1623(a)
`
`IPR Petition for
`
`U.S. Patent No. 7,214,506
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`benefits
`of
`evidence
`scientific
`objective
`(Robinson, 2001). Moreover, a more competitive
`detailing environment (LeadDiscovery, 2006);
`lack of new blockbuster drugs to gain physicians’
`attention (PricewaterhouseCoopers, 2008); and
`the
`increasing
`role of direct-to-consumer
`advertisements
`and
`electronic
`detailing
`(Davidson
`and
`Sivadas, 2004)
`all have
`contributed to the declining detailing impact. In
`fact, the average detailing duration dropped from
`five minutes in 1998 to less than one minute in
`2004 (Yi, 2008), signaling
`the physicians’
`declining interest in hearing from the reps.
`Many researchers have found evidence of
`high market share of detailing voice positively
`impacting the market share of detailing product
`(Jones, 1990; Shimp, 2000; Gonul, Carter,
`Petrova, and Srinivasan, 2001; Pesse, Erat, and
`Erat, 2006). As a result, pharmaceutical firms are
`committed to maximizing their share of voice
`within their resource constraint in an effort to
`increase sales; one way to increase the share of
`voice without adding more sales reps is to detail
`multiple products instead of single product.
`
`2.1. Share of Voice Computation
`
`This paper explores the importance of a
`knowledge-based approach in improving sales
`force operations in multiple product detailing
`situations by integrating domain knowledge with
`quantitative modeling process. The approach
`specifically targets to minimize major limitations
`of the traditional approach in planning for
`detailing multiple products. The result from this
`study is implemented and tested in a real-world
`environment to a sample of physicians in a
`territory to explore its performance against a
`control group of similar physician size and sales
`volume.
`The remainder of the paper is organized
`as follows: Section II gives an overview of the
`background of the pharmaceutical industry and
`sales operations related challenges taking place
`in the industry. Section III explains the data sets
`used for this study. Section IV describes a
`knowledge-based approach developed to derive a
`set of weights for planning detailing strategy, and
`Section V summarizes the plan’s performance
`based on actual implementation of the approach.
`Section VI
`discusses
`the
`approach
`and
`concluding remarks.
`
`II. BACKGROUND OF THE INDUSTRY
`
`Physician detailing is the primary means
`to market pharmaceutical drugs because in this
`market the physicians are the ones who decide
`the best treatment algorithm for their patients,
`who are
`the end users. This dynamic of
`promoting
`to physicians,
`is different from
`traditional marketing, which
`targets
`its
`promotional efforts directly to the end users;
`however, detailing is similar to other forms of
`promotion, used in traditional markets, in a sense
`that
`it
`is both a marketing
`tool and an
`informational source (Nelson, 1974).
`The detailing efforts have been losing its
`impact over the years due to significant changes
`in the selling. The primary change is from
`managed care organizations’ growing influence
`in regulating the use of drugs coupled with an
`increasing number of physicians seeking more
`
`To derive the share of detailing voice,
`physician detailing equivalent (PDE) weights for
`the product and the market are computed first;
`PDE is used by the industry to calculate total
`detailing efforts when detailing is done in
`multiple sequences, and the PDE weights reflect
`the relative detailing impact of each sequence.
`Equation (1) shows how PDEjkl, which denotes
`physician detailing equivalent for physician j, in
`time period k, for product l, is calculated:
` ¦
`i
`Dijkl is defined as the total number of details
`made in sequence i to physician j in time period
`k, for product l, while Wi defines the PDE weight
`for detailing sequence i. In addition, the weights
`play an instrumental role in computing share of
`voice in time period k, for product l, denoted as
`SOVkl, as shown in Eq. (2):
`
`PDE
`
`jkl
`
`(W
`i
`
`u
`
`D
`ijkl
`
`
`
`r) fo
`
`
`
` i,
`
`j,k,l
`
`(1)
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`44
`
`Page 2 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`PDE
`
`jkl
`
`¦
` ¦¦
`
`SOV
`
`kl
`
`j
`
`for
`
`
`
`k
`
`(2)
`
`PDE
`
`jkl
`
`l
`
`j
`
`2.2. Impact of Detailing Multiple Products to
`Share of Voice
`
`The traditional PDE weights, shown in
`Table 1, are provided by the firm sponsoring this
`research; the table shows that the full weight of
`one is assigned to the first detailing product
`independent of the number of products in the
`detailing portfolio. In other words, as long as the
`product is detailed in the first position, it will
`always carry the full detailing weight. Similarly,
`if the product is detailed in the second sequence,
`
`it will always have the PDE weight of 0.6.
`Finally, any product detailed
`in
`the
`third
`sequence or beyond will have the PDE weight of
`0.3. Clearly, the firms detailing multiple products
`will have higher share of voice with 1.9 PDE
`when a sales rep details three products to a
`physician in a single visit versus 1 PDE when a
`rep details a single product.
`According to the sponsoring firm, the
`origination of the PDE weights is based on
`primary market
`research
`to
`physicians.
`Interviews with sales operations professionals in
`other companies made possible by pre-existing
`professional contacts have validated that these
`values are similar across the industry.
`
`TABLE 1: SUMMARY OF TRADITIONALLY APPLIED PDE WEIGHTS FOR DETAILING
`SEQUENCE BASED ON THE NUMBER OF DETAILING PRODUCTS
`1st position
`2nd position
`3rd + position
`
`In single product detailing
`
`In two-product detailing
`
`In three+ product detailing
`
`1.00
`
`1.00
`
`1.00
`
`-
`
`0.60
`
`0.60
`
`-
`
`-
`
`0.30
`
`2.3. Limitations of Traditional Approach
`
`The traditional approach in utilizing PDE
`weights has two major limitations. The first
`limitation is that the approach always gives
`benefit
`to detailing more products versus
`detailing fewer products by a way of increasing
`SOV. This is a flawed assumption because it is
`hard enough to access physicians in recent years
`and when the access is granted, they are not
`allowing for more time if reps detail more
`products with average details lasting less than a
`minute (Yi, 2008). More likely, the detailing
`products will likely cannibalize the individual
`detailing
`impact due
`to
`the spreading of
`information in a fixed time. Thus, always giving
`SOV advantage
`to multi-product detailing
`strategy may mislead management in making
`sound sales operations decisions.
`
`the PDE weight for each
`Secondly,
`detailing sequence is constant for all physicians
`regardless of how well they respond to details.
`This
`is another flawed assumption because
`physicians and their patients’ needs are different;
`if the firms do not accommodate for these
`differences and neglect to provide individualized
`detailing
`strategy,
`a
`significant negative
`consequence
`such as
`suboptimal
`resource
`allocation
`and
`undesirable
`sales
`force
`performances are
`likely consequences
`(Yi,
`Anandalingam, and Sorrell, 2003).
`In spite of the significance of these
`weights have on sales operations decisions,
`surprisingly little is known about them via
`published research. In this paper, we propose a
`new approach to minimize the impact of the
`aforementioned limitations to sales performance,
`and investigate the feasibility and performance of
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`45
`
`Page 3 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`the approach when implemented to a small
`sample of physicians.
`
`III. DATA
`
`A pharmaceutical company with annual
`US sales over $2 billion sponsored this research
`on the condition that the company would receive
`the model and report of the findings while
`remaining anonymous and having a say in when
`to release the research publication. The firm
`provided (1) the detailing history and respective
`sales data for one of its territories in the
`Northeast region, comprising a total of 72
`physicians on its target list, and (2) team of
`domain experts and their time to help in this
`research on a $275 million prescription drug
`product. This drug was launched in late 1990s, is
`promoted by multiple sales forces in different
`detailing sequences, and competes against four
`branded products for market share. The product
`was selected for this research mainly due to the
`wealth of detailing data available.
`generally
`Pharmaceutical
`companies
`target physicians detail based on the volume of
`prescriptions they generated in both the drug
`class and the drug itself. The physicians were
`sorted in order of prescription volume in the
`disease class, and then they were grouped into 10
`equal segments, with the first decile representing
`the lowest prescribers and the 10th decile the
`highest; the higher-decile physicians received
`more detailing visits from the sales reps than did
`the lower-decile physicians.
`the
`To
`initiate
`this study and find
`direction of the research, we merged two sets of
`data, by physician identification number, to form
`the database. One data set contains the number of
`prescriptions
`that
`the physicians on
`the
`company’s target list wrote for the studied drug
`and
`its competitors. The second contains
`information about
`the sales reps’ detailing
`activity with the physicians. Two years’ worth of
`data, broken out into eight quarters from 1st
`quarter 2003 through 4th quarter 2004, were
`collected for the study; we used quarterly data
`
`because monthly data contained too much noise
`for the research.
`In addition, the company provided the
`competitive sales activity data at the territory
`level for the same period as that used for the data
`analysis.
`It
`captured
`information on
`all
`competing products marketed
`in
`the same
`therapeutic area of the company’s product: the
`competitors’ sales force structure; the number of
`sales reps detailing the drugs; and the detailing
`sequences of the products for each territory.
`There were concerns about data integrity of other
`promotional events, such as direct-to-consumer
`advertising,
`electronic
`detailing,
`journal
`advertising, and sponsored medical educational
`programs; these data points were excluded from
`this study.
`
`IV. NEW APPROACH: KNOWLEDGE-
`BASED APPROACH
`
`A knowledge-based approach is defined
`as one designed to extract and integrate the tacit
`and explicit knowledge within the organization
`and then to apply it as a vital component in the
`quantitative modeling process to improve the
`organization’s performance as well as gaining
`insights that can provide competitive advantage
`(Blattberg and Hoch, 1990). This paper proposes
`a knowledge-based approach at the physician
`level to explore whether or not limitations of the
`traditional approach can be alleviated while
`improving sales operations involving multiple
`products. The theoretical framework for this
`approach
`is
`founded on knowledge and
`micromarketing.
`
`4.1. Theoretical Framework for Knowledge
`and Micromarketing
`
`Knowledge is defined as the set of
`justified beliefs that enhance a firm’s capability
`to
`take effective action
`(Nonaka, 1994).
`Knowledge can largely be divided into two areas:
`tacit and explicit. Tacit knowledge refers to
`insights, intuitions, and hunches that are not
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`46
`
`Page 4 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`easily verbalized or communicated. This tacit
`knowledge is critical in decision making process
`because it is the primary source of problem
`definition and alternatives
`(Davenport and
`Prusak, 1998). On the other hand, explicit
`knowledge refers to that which can be formally
`expressed and collected as data, words, and
`software, therefore, be easily diffused throughout
`an organization (Davenport and Prusak, 1998).
`Researchers have found that converting tacit
`knowledge
`into
`explicit
`knowledge
`and
`integrating the two significantly enhances a
`company’s competitive position by improving
`organizational
`capability,
`competence,
`and
`performance
`(Brown and Duguid, 1998).
`Moreover, knowledge integration across different
`functions within a
`firm has demonstrated
`improvement in decision making quality and
`organizational performance (Blattberg and Hoch,
`1990; Cai, 2006; Liebowitz, 2008).
`that
`shown
`Recent
`studies
`have
`knowledge capture and management can be
`improved by integrating visualization into the
`modeling process, with visually agreed-upon
`knowledge being very successful in capturing
`and segmenting complex knowledge (Coffey,
`Hoffman, and Cañas, 2006; Strohmaier and
`Lindstaedt, 2007). Also,
`integrating domain
`experts’ knowledge with secondary data that can
`be used
`to derive visually
`agreed-upon
`promotional response patterns has proven to be
`an effective way
`to
`identifying responsive
`physicians,
`leading
`to derivation of more
`accurate response functions and, consequently,
`improvement in the quality of the detailing plan
`(Yi et al., 2003). Moreover,
`it has been
`demonstrated
`that
`the promotional response
`function parameters for individual physicians can
`improve
`its accuracy
`by calibrating
`the
`
`parameters to reveal responsiveness as defined
`by the experts (Yi, 2008).
`Based on these previous studies, this
`paper hypothesizes
`that optimally utilizing
`knowledge is critical to improvement of detailing
`planning. In addition, accurate PDE weights are
`those
`that
`visually
`reveal
`physicians’
`responsiveness by matching its pattern to the
`predetermined responsive patterns developed by
`domain
`experts,
`resulting
`in
`improved
`promotional
`functions and detailing plans.
`Moreover, since PDE weights are inputs to SOV
`computation as well as to detailing planning,
`improvement in the weights will also improve the
`qualities of SOV calculation as well as detailing
`planning. These benefits are expected to result in
`minimization of non-value-added costs, making
`the sales reps more effective and therefore
`increasing revenue.
`Micromarketing is tailoring marketing
`at
`the
`consumer
`level
`to better
`plans
`accommodate individual differences in responses
`to promotions (Leeflang and Wittink, 2000;
`Zhang and Krishnamurthi, 2004). In addition,
`similar
`to
`traditional consumers, physicians
`respond better to marketing messages tailored to
`their individual needs (Yi, 2008). Therefore,
`incorporating micromarketing as part of a
`knowledge-based approach is expected to be
`more effective than the traditionally targeting
`physicians at a macro level, and further increase
`the effectiveness of sales operations.
`
`4.2. Process Flow of the Knowledge-Based
`Approach
`
`The process flow of this approach is
`shown in Figure 1. This flow is developed to
`provide transparency to the proposed process.
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`47
`
`Page 5 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`1. Define & Agree
`on Patterns of
`Responsiveness
`
`2. Develop, Train,
`& Test Neural
`Nets (NN) Model
`
`3. Eight qtrs
`of Rx and
`PDE data for
`physician i
`
`4. NN Model
`Application
`
`7 Search for
`PDE
`weights
`
`5.
`Responsive?
`
`No
`
`Yes
`
`Yes
`
`6.
`First Entry?
`
`No
`
`8. Collect info of
`doctor i
`
`9. Optimization &
`Reporting
`
`FIGURE 1: PROCESS FLOW OF THE KNOWLEDGE-BASED APPROACH
`
`of
`definition
`the
`1,
`Step
`In
`responsiveness, based on the visual relationship
`between PDE and prescription volume over time,
`is constructed by working with a cross-functional
`team that includes representatives from Sales
`Operations, Sales, Marketing, Market Research,
`and
`Information Management. Each
`team
`member carries a title of manager or higher and
`at least three years of work experience in this
`brand as well as familiarity with the territories
`
`selected for the research. This cross functional
`team defined responsiveness based on two sets of
`rules and those not meeting these rules are
`defaulted as non-responsive. The two rules of
`responsiveness are: 1) synchronize movement for
`all eight quarters, and 2) allowing for a single
`quarter
`deviation
`from
`the
`synchronize
`movement property. Figure 2 illustrates examples
`of responsiveness based on these two rules.
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`48
`
`Page 6 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`Two rules
`
`Responsiveness examples
`
`1. Synchonized
`
`movement
`
`# of Rx
`
`PDE
`
`Time in
`quarters
`
`2. Single Quarter
`
`Deviation
`
`FIGURE 2: EXAMPLES OF PREDETERMINED PATTERNS OF PHYSICIAN
`RESPONSIVENESS TO DETAIL
`
`examples,
`non-responsiveness
`The
`defaulted from not meeting the aforementioned
`rules, demonstrate cases where there exists no or
`insufficient visible pattern of
`relationship
`
`between PDE and prescription volume over time
`and are shown in Figure 3. Clearly, detailing
`alone cannot explain these physicians prescribing
`behavior.
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`49
`
`Page 7 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`Nonresponsiveness examples
`
`Single classification
`
`# of Rx
`
`PDE
`
`Time in
`quarters
`
`FIGURE 3: EXAMPLES OF PREDETERMINED PATTERNS OF PHYSICIAN
`NONRESPONSIVENESS TO DETAIL
`
`In Step 2, a neural network (NN) model is
`developed to identify, from the target physician
`pool, individual physicians who are responsive to
`details. The main reason for using NN model in
`this study is because it automates otherwise
`manually
`intensive
`activity of
`classfying
`hundreds of physicians into two categories of
`responsiveness based on visual patterns between
`PDE and respective prescription volume for eight
`quarters developed
`in Step 1. In addition,
`strengths of NN models are the properties of
`adaptability, nonlinearity, fault tolerance, and
`input-output mapping (Jain and Vemuri, 1999;
`Kim, Lee, and Aguihotri, 1995). On the other
`hand, NN’s limitations are that its functionality is
`often perceived as black box,
`the model-
`
`development process is more art than science,
`and
`time consuming data-preparation step.
`(Livingstone, Manallack, and Tetko, 1997).
`Similar to the work done by Yi et al.
`(2003), this research uses a back-propagation
`network with 16 input nodes (8 quarters of PDE
`and 8 quarters of the respective prescription
`volume (TRx), 1 hidden layer containing 7
`neurons, and 1 binary output node (1 for
`responsive and 0 for nonresponsive physicians).
`The model was developed with 450 training
`samples with known results. With a predicted
`accuracy of 84%, the NN model compared
`favorably with the logistic regression model that
`produced a predicted accuracy of 53%, using Eq.
`(3).
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`50
`
`Page 8 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`u
`
`%100
`
`(3)
`
`º
`
`»»»» ¼
`
`Abs
`
`(
`
`Act
`
`
`
`i
`
`Pred
`
`)
`
`i
`
`n
`
`n
`
`¦
`
`i
`
`1
`
`
`
`ª
`
`«««« ¬
`
`1
`
`Predicted accuracy % =
`
`where
`
`actual output of physician i
`Acti
`Predi predicted value for physician i
`absolute value function
`Abs
`number of testing samples
`n
`
`In Step 3, the eight quarters of TRx and
`respective PDE data for physicians are prepared
`for Step 4, the NN model application. In Step 5,
`responsiveness of physicians is determined, with
`nonresponsive physicians’ data directed to Step 6
`and responsive physicians’ data to Step 8.
`All nonresponsive physicians entering Step 6 for
`the first time go through to Step 7, where a
`nonlinear mathematical model interface with the
`NN model searches for a set of PDE weights that
`reveal physicians’ responsiveness to detailing
`efforts; the nonlinear program interfacing with
`the NN model is shown here:
`
`Maximize
`s.t.
`
`
`
`NN(Rx
`
`,
`
`PDE
`
`jk
`
`jk
`
`)
`
`
`
`for k
`
`
`
`1 8,...,
`
`
`
`
`
`3
`
`PDE
`
`jk
`
`u
`( DW
`ij
`ijk
`
`
`
`)
`
`
`
`for k
`
`
`
`1 8,...,
`
`
`
`
`
` ¦
`
`i
`
`
`1
`i
`for
`
`d
`1
`Wij
`t
`WW
`
`,1
`ij
`i
`j
`
` variablesall
`
`
`
`
`
`21,
`
`for i
`
`
`t
`
`0
`
`(4)
`
`(5)
`
`(6)
`
`(7)
`
`(8)
`
`The objective function, given by (4),
`maximizes the number of responsive physicians
`in the first summation while maximizing the
`summation of
`the weights
`in
`the second
`summation. The first summation interfaces with
`the trained NN model by providing, to the model,
`the physician-level prescription data and the PDE
`data for all eight quarters, given by Rxjk and
`respectively,
`to
`determine
`the
`PDEjk,
`responsiveness of the targeted physicians.
`The first constraint, given by (5), defines
`PDE for each physician in each quarter. The set
`of PDE weights, for the ith position to physician j,
`Wij, is initialized to 1, 0.6, and 0.3 for detailing
`positions 1, 2, and 3+, respectively. Constraint
`(6) sets the upper limit for the weight to be one.
`Constraint (7) forces the weights of preceding
`detailing positions
`to be bigger
`than the
`subsequent ones
`to
`reflect
`the
`inverse
`relationship between the detailing time and the
`order in which a product is detailed as well as to
`limit the searching space for the nonlinear
`program. The last constraint, given by (8),
`defines
`the non-negativity condition for all
`variables. Figure 4 illustrates how this step works
`by having a physician visually fitting to the
`nonresponsive definition with the traditional set
`of PDE weights, but the optimization algorithm
`found a new set of PDE weights to make the
`physician fit the definition of responsiveness, and
`this new set of weights replaces the traditional
`weights for this physician, with the physician
`classified as responsive.
`
`Where NN(Rxjk, PDEjk) are trained neural
`network function, returning 1 if physician j is
`responsive and 0 if physician j is nonresponsive
`based on relationship between Rx written and
`PDE over eight quarters; PDEjk is the physician
`detail equivalent for physician j in quarter k; Rxjk
`is the total number of prescriptions written by
`physician j
`is the detailing
`in quarter k; Wij
`weight for the ith sequence for physician j; Dijk is
`the total number of details made from the ith
`sequence to physician j in quarter k.
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`51
`
`Page 9 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`Wl. 1
`Wl. 2
`Wl.3
`c ails in 1 I
`5
`3
`4
`5
`2
`3
`5
`5
`o
`c ails in 2
`3
`3
`2
`1
`1
`2
`2
`c ails in 3 ~..-_::.5 ___ ___:0;._ ___ o:::..... ___ ...:7 ____ ::.3 ___ ___:5;._ ___ 7:...._ ___ ...;:0;.___.
`
`POE
`TR x
`
`Q t r 1
`6.5
`57
`
`Q t r 2
`6 . 2
`68
`
`Q t r 3
`4 .2
`5 0
`
`Q t r 4
`4 .7
`4 3
`
`Q t r 5
`6.5
`55
`
`Q t r 6
`6 .7
`44
`
`Q t r 7
`6 .9
`40
`
`Q t r 8
`6 .8
`60
`
`8 ~----------------------------~ 80
`ro
`7
`6
`
`60 I 50
`
`5
`
`~ 4
`3
`2
`
`1
`0
`
`Qtr 1
`
`Qtr 2
`
`Qtr 3
`
`Qtr 4
`
`Qtr 5
`
`Qtr 6
`
`Qtr 7
`
`Qtr 8
`
`1-
`
`PD E -o - TRx I
`
`10
`0
`
`40
`30
`
`20 i
`
`Wl. 1
`WI. 2
`Wt.3
`
`0 82
`0 .68
`00 1
`
`Calls in 1 I 5
`5
`3
`2
`5
`4
`3
`5
`C alls in 2
`2
`2
`1
`1
`2
`3
`3
`0
`c ails in 3 ~..-_.::.5 ___ ___::0 ___ __;0::...._ ___ .:...7 ___ ___::3;__ __ __;5::...._ ___ .:...7 ___ ___:0;____.
`
`P OE
`TRx
`
`Qtr 1
`4 .15
`57
`
`Qtr 2
`5.46
`68
`
`Qtr 3
`3 .82
`50
`
`Q tr 4
`2.39
`43
`
`Qtr 5
`4 .8 1
`55
`
`Qtr 6
`4.69
`44
`
`Qtr 7
`4.57
`40
`
`Q trs
`6. 14
`60
`
`7 . -- - - - - - - - - - - - - - - - - - - - - - - - - -- - . 80
`6
`70
`60 ~
`50 l!.
`40 Q
`30 ~
`
`5
`
`~ 4
`3
`
`2
`
`20 ~
`10
`0
`
`0
`
`Qtr 1
`
`Qtr 2
`
`Qtr 3
`
`Q t r 4
`
`Qtr 5
`
`Qtr6
`
`Qtr 7
`
`Qtr 8
`
`1-
`
`PDE -o -TRx I
`
`FIGURE 4: AN EXAMPLE OF DERIVING A SET OF NEW PDE WEIGHTS FOR AN
`ACTUAL PHYSICIAN WHO APPEARED VISUALLY NONRESPONSIVE
`WITH THE TRADITIONAL SET OF PDE WEIGHTS.
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`52
`
`Page 10 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`each
`information on
`In Step 8,
`physician’s
`responsiveness and
`the set of
`respective PDE weights for each of them is
`collected and stored. For the nonresponsive
`physicians, PDE is determined by taking the
`average PDE weights from
`the responsive
`physicians for each detailing sequence.
`
`In Step 9, physician responsiveness and PDE
`weights are merged with physician data and the
`company’s resource data to formulate a nonlinear
`programming model. The objective of this model
`is to determine the optimal plan for detailing the
`firm’s target physicians to maximize quarterly
`profit. This formulation is shown here:
`
`Maximize
`
`¦
`
`i
`
`[PRF
`i
`
`(
`
`PDE
`
`i
`
`u
`
`)
`
`Price
`
`-
`
`Cost
`
`(
`
`PDE
`
`i
`
`i
`
`/
`
`E)]
`
`[TP
`d
`
`u
`
`*
`
`PRF
`d
`
`(
`
`PDE
`
`d
`
`u
`
`)
`
`Price
`
`-
`
`Cost
`
`(
`
`PDE
`
`d
`
`d
`
`/E)],
`
`i
`
`
`
` I
`
`
`
`10
`
`
`1
`
`¦
`d
`s.t.
`
`PDE
`
`i
`
`3
`
` ¦
`
`
`1
`
`j
`
`u
`DW
`ij
`ij
`
`,
`
`for
`
`all
`
`i
`
`n
`
`j
`
`W
`
`j
`
`
`
`(9)
`
`(10)
`
`(11)
`
`(12)
`
`(13)
`
`(14)
`
` ¦
`
`
`1
`
`i
`
`j
`
`/
`nW
`ij
`
`j
`
`,
`
`for
`
`32,1,
`
`W
`
`j
`
`u
`
`TD
`id
`
`,
`
`for
`
`all
`
`d
`
`3
`
`1¦
`
`j
`
`PDE
`
`
`
`d
`
`,R
`j
`
`
`
`j
`
`32,1,
`
`¼º
`
`
`d»
`
`
`
`W
`
`j
`
`u
`
`TD
`id
`
`10
`
`¦¦
`
`j
`
`
`1
`
`d
`
`
`u
`DW
`ij
`ij
`
`
`
`
`
`n
`
`j
`
`¦¦
`
`j
`
`
`1
`
`i
`
`«¬ª
`
`
`
` variablesall
`
`t
`
`0
`
`where
`
`PDEi
`Rj
`Wij
`Dij
`TDid
`
`PRFi (x) promotional response function for physician or decile i, returning expected
`prescription volume for x detail in a quarter
`physician detail equivalent for responsive physician or decile i
`total quarterly resource for the jth position details
`detailing weight for responsive physician i for detailing position j
`total details that need to be made in the ith position to responsive physician j
`total details that need to be made in the ith position to nonresponsive physician in
`decile d
`total physicians in decile d
`price of a single prescription of the drug
`cost to detailing physician i
`efficiency factor to account for empty efforts directed to the physicians’ offices
`number of responsive physicians for detailing sequence j
`
`TPd
`Price
`Costi
`E
`nj
`
`The objective function, given by (9),
`maximizes the quarterly profit from the sales
`force efforts. The first summation in the function
`
`calculates the optimal detailing plan to generate
`maximum profit from the physicians who are
`responsive to the sales force’s detailing efforts:
`
`California Journal of Operations Management, Volume 9, Number 1, February 2011
`
`53
`
`Page 11 of 15
`
`
`
`John C. Yi, Ming Zhou, and Taeho Park
`An exploratory study to improve sales operations when selling multiple prescription drugs
`
`the promotional response function of PDEi
`details to physician i, given by PRFi(PDEi),
`produces the number of prescriptions written by
`physician i; this is then multiplied by the price
`per prescription, price, to arrive at revenue; the
`cost
`to detailing physician
`is given by
`i
`cost(PDEi/E), where E, which denotes efficiency
`factor and is less than one, accounts for the
`empty efforts made by the sales reps; and taking
`the difference between the revenue and cost per
`physician i and summing up the profit for all the
`responsive physicians gives
`the
`total profit
`generated by this group.
`The second summation in the function
`calculates the profit from the nonresponsive
`physicians: since there is no visually discernable
`response pattern,
`the promotional
`response
`function to detailing effort PDEd, given by
`PRFd(PDEd), is derived at decile level d based on
`average PDE weights from
`the responsive
`physicians; this function produces the average
`number of prescriptions written by an average
`physician from decile d; multiplying the number
`of prescriptions by price and subtracting the cost
`associated with
`the detailing effort, again
`including E, gives the profit per physician from
`decile d; and summing for all the deciles gives
`the total profit generated from this group.
`Constraint (10) defines PDE for physician
`i, based on the PDE weights found specifically
`for responsive physician i in detailing position j
`derived earlier, in Step 7. The set of weights
`provides the visually recognizable pattern of
`responsiveness for physician i, which enables the
`program to locate the optimal set of details for
`that physician.
`The average PDE weights for responsive
`physicians defined for each detailing sequence is
`in constraint (11). Constraint (12) defines PDE
`for the nonresponsive physicians per decile d.
`Constraint (13) sets the upper limit Rj on the total
`quarterly detailing resources for the drug for each
`
`sequence j, while the non-negativity condition is
`set by constraint (14).
`
`V. APPLICATION AND RESULTS
`
`To measure the effectiveness of the
`knowledge-based approach, the sponsoring firm
`allowed
`its
`implementation
`in a randomly
`selected territory where the research data were
`collected; this territory, located in the Northeast
`region with 72 target physicians, are called the
`test group. The implementation period is from
`March 2005 through June 2005, with the March
`used as a grace period to correctly implement the
`plan. The result from the latter th