Does this diagram look familiar?  If your response curves resemble this, you are most likely wondering, “Why do lower target deciles appear more responsive than higher ones?” Or, “Why is the promotion showing a negative impact on sales?”

Your promotion response model may not be capturing essential time trends and seasonality that impact your sales. Effective modeling with panel analysis can help.

First, it’s essential to understand what panel analysis is.

Panel analysis is a statistical modeling technique designed to measure individual entities’ responses over time. It is thus equipped to capture the uncontaminated, pure impact of promotion, by teasing out the baseline response of each individual physician and each time period. This ability to provide more precision when using predictive analytics, particularly promotion response, makes for more reassuring resource allocation decisions and has a significant downstream impact on areas such as pharmaceutical sales force sizing, marketing mix, customer segmentation and targeting, and several other applications in the realm of life sciences.

In fact, with enough good quality data, panel analysis may even be capable of providing individualized promotion response results beyond the baseline. Leveraging these individualized results can bring the industry closer to personalized promotion in the commercial space — a goal that has thus far seemed too distant to reach.

To give you a better sense of what panel analysis is, here we have compared it to two common types of analyses: Time-series Analysis and Cross-sectional Analysis.

  1. Time-series – data for a single entity (HCP, hospital, etc.) that varies over time
  2. Cross-sectional – data for a single time period that varies across entities
  3. Panel – data that varies across both time period and entities

The results from each type of analysis provide us with varying levels of insight. The time-series analysis provides insight into the impact of time on the response variable overall but is unable to capture variances across entities. The cross-sectional analysis on the other hand, provides insight into the unique response of each individual entity but is unable to capture variances over time.

The panel analysis, a sort of joyous union between the time-series and cross-sectional analyses, is able to capture the impact of time and the unique response of each individual entity. It thus provides for a richer analysis that yields more insight into the changing behavior of the response across individual entities and over time.

What is it that really makes panel analysis useful?

HCP prescribing often exhibits time trends and seasonal patterns independent of promotion activities. Traditional modeling techniques are not well equipped to capture such trends or seasonality.

Panel analysis works by extracting the effect of time while looking for true responses to promotion. As a result, it is able to more accurately attribute HCP prescribing to its individual contributors, be it trend, seasonality, or promotion activity.

Look back at the chart with the mismatched deciles and negative impact at the beginning of the article. Now compare that to what it looks like after modeling using panel analysis.

Another drawback of traditional approaches is that they assume all HCPs with similar historical Rx volume will have the same baseline response.

Panel analysis rejects this premise. Rather, it takes into account that every HCP is an individual and estimates each individual’s unique baseline response, beyond which response to promotion is captured. This is a more data-driven approach, using the data to point to the HCPs who best respond to promotion, rather than using inherent assumptions about physician volume to drive the analysis.

Finally, here’s our analysis of Panel Analysis.

Predictive modeling and analytics in the life sciences industry have long since struggled to cope with the challenges of identifying and decoding the complexity of human behavior and its changes over time. However, if used correctly today, it can be harnessed to optimize commercialization with planning, production, and distribution. The basis of a valuable predictive analysis lies in its handling of seemingly ‘unexplainable’ events – events whose impact on the desired response is difficult to estimate and extract, and the primary measurement of its success lies in how well it reflects reality.

Proper modeling with panel analysis can largely alleviate any undesired impact and produce a more accurate promotion response, thereby painting a truer picture of reality and having far reaching implications for sales strategy, marketing strategy, and overall decision science.

If you’d like to learn more about panel analysis and how it can be used, be sure to read our panel analysis white paper.