Challenges With Marketing Mix Modeling in Pharma

Marketing mix analysis is an ideal way to measure the impact of different marketing channels for various industries, including pharma. The algorithm is quite simple, which further adds to its quick adoption and popularity. It provides us with intuitive, easy-to-explain, and useful results while offering many downstream applications of the channel’s incremental impact. For instance, we can use it to plan a budget, redistribute spend, and maximize ROI. Campaign analysis, sequencing the journey, and next best engagement (NBE) are some of the few projects where the betas can be useful.

Despite being standardized, the marketing mix (MMix) model has a lot of room to grow in the pharma space, such as: 

1. Large Datasets

The sales and marketing data in the pharma industry can be gigantic. And the longitudinal nature of the marketing mix data can lead to large size of final datasets having millions of rows. To validate the datasets, we need statistical numbers, such as mean, median, percentiles, skew, kurtosis, or graphs. 

• We can also use histograms to see the distribution of touchpoints along with scattering plots that can help us select the right transformation while modeling 
• Boxplots can help us understand the data distribution and outlier points
• Monthly trends can help us identify the distribution of sales and related activities 

But plotting these graphs is computationally expensive. Tools that can quickly process data, such as SAS and SQL, have lower data visualization capability. While the tools that have good visualization features available, like Tableau, Power BI, etc., have a lower computational efficiency.

2. Bringing Consistency

Another big challenge in marketing mix modeling in pharma is ensuring consistency while using it. Pharma analytics teams are cross-functional, and the results are shared and validated by differet of teams before you can present them to the brand teams for a final review. 

The cross-functional nature of the analytics team in pharma prevents easy sharing of the marketing mix results. The context, period, business rules, and other details that were specified during an analysis get lost in translation, and the new team often needs spend additional time in understanding these details. As such there are inherent inefficiencies built into the various steps within the marketing mix.

3. Data Gathering

The data for a traditional MMix is drawn from multiple sources, and these sources have different data formats, refresh frequency, and storage. For example, in pharma, the marketing data may have to be sourced from CRM giants like Veeva or Salesforce, or individual vendors like Doximity and Medscape. 

Data gathering requirements differ for cross-functional teams which can result in multiple iterations before the ready-to-use dataset is created. This is caused by the absence of dedicated teams for carrying out specific operations. 

4. Data Preparation

After gathering the data, the next step is to check the dataset for errors, missing values, right time frame, right columns, etc., to ensure that the dataset is ready for further use. 

In order to verify the business rules for processing the data, the MMix team needs to connect with the brand teams or an onshore counterpart who can be busy at times. Also, identifying the right person to validate the statistics can be time-consuming, which acts as a major roadblock. 

5. Occurrence of QC Errors 

Data processing is mostly done manually, leaving a lot of room for errors. The analysts spend a lot of time massaging the data and writing codes but fail to make it error-free. Even a small anomaly that doesn’t look harmful can add investigation time to QC the data, and worse, the error may go undiscovered. For example, if you must sum two columns in SAS and if there are missing values/”.” in a few rows of the data, then the result would be: “.”