Moving from R&D to Sales and Marketing: the business analytics experience of a statistician at Lilly

Transcription

1 Moving from R&D to Sales and Marketing: the business analytics experience of a statistician at Lilly Todd Sanger, Ph.D. Research Fellow, Advanced Analytics Eli Lilly & Company

2 Questions in Sales and Marketing Should How do we we treat bring our How Where Do What we do to often is to predict have the we launch should sell best the the a way our right portfolio we most new to integrated customers unparalleled healthcare of effective sales distribute move diabetes force sales product? sales samples? alignments? products? reps? reps? systems experiences? differently?

3 Healthcare Ecosystem Television Demographics Sales force Details / Samples Demographics Social media Brand equity / lifecycle Speaker programs Brand loyalty / preferences Patient Healthcare professional Treatment decision (Rx/sales $) News Managed care Managed care Websites Market events Direct mail Pharmacist Market events PR / info COMPETITION Direct Samples Touchpoints Mobile Influences Display / web

4 Data Sources Patient Longitudinal Data Health Care Professional Data Wholesaler Distribution/Inventory Data Market Research Data Marketing Touchpoints Data

5 Revenue Modeling Process Input data (touchpoints) Statistical models Output data (impact per touchpoint) 350 Detail TV s Time y t p i 1 x i p it i t Direct Mail Sales Decomposition Time Radio Mobile Time TV Direct Mail Radio Print Base Time Print Time Brand sales = B x units of touchpoint

6 Touchpoint Contributions to NRx 34% carryover 1% Iconnect DM 30% details 1% 3% samples details 1% direct mail 1% sports TV 2% speaker programs 19% national TV 2% paid search 3% print (news) 3% digital display

7 Design/Build Phase Selection of Data Level Analysis on disaggregated data with random effects yields insight refinement Aggregation of Customers Provides directional insights by Touchpoint β fixed 4.3 Level of Insights TRx/Detail Estimation of touchpoint impact on outcomes in aggregate ( fixed effect ) Segmentation of Customers Provides granular insights by Touchpoint and Segment β fixed 4.3 β random Estimation of touchpoint impact on outcomes by Customer Segment ( fixed and random effects ) Customer Level Provides granular insights by Touchpoint and Customer β fixed 4.3 β random * Fixed effect example; 1 detail generates 4.3 Rx Estimation of touchpoint impact on outcomes by Customer ( fixed and random effects ) HYPOTHETICAL DATA

8 Aggregate Touchpoint Impact vs Customer Experience While the models will produce a result that can be summarized as a total Trx contribution of each touchpoint the reality is that total contribution is an aggregation of a unique set of touchpoint impact experiences that vary by customer

9 Some Pitfalls Source: CDC and USDA

10 Some Pitfalls Eliminating ice cream was recommended as part of an anti-polio diet!

11 Remedy: Prospective Test-Control Gold standard method of determining effectiveness Only way to scientifically prove cause-and-effect relationships The process of randomization into test groups and control groups guarantees that the observed difference between them must be due to the treatment, and no other extraneous factors Test Group Customer Population Randomize Control Group

12 Remedy: Retrospective Match-Control Next best option to prospective test-control Does NOT guarantee cause-and-effect relationships, but gets us closer The idea is that for each customer in the test group, we find a twin customer that did not receive the treatment ( control group) Twins are customers that are similar to the test customer based on the criteria that we specify (historical Rx behavior, Details, s, etc.) Customer Population Twin match Test Group Twin match Approximate Control Group

13 Conclusions Vast amount of data in healthcare space available for drawing insights We need to understand how to use it properly Right data, even if not big, can be more valuable than Big data that is not right While data is useful, smart analytics is needed to make the data come alive