Pharmaceutical Dynamic Paid Search Budget Allocation Model

Overview

This project was for a British-Swedish multinational pharmaceutical and biotechnology company headquartered at the Cambridge Biomedical Campus in Cambridge, England. 

With 2020 revenues of $26.6 billion dollars, the firm develops pharmaceuticals for oncology, cardiovascular, gastrointestinal, infection, neuroscience, respiratory, and inflammation. It is perhaps best known for its involvement in developing the leading COVID-19 vaccines.

In early 2020, the pharma client was gearing up for a Paid Search (Google Ads) and social media advertising campaign with the goal of increasing market share and awareness of seven specific drugs across three therapeutic categories:

1. Bydureon (Diabetes)
2. Qtern (Diabetes)
3. Farxiga (Diabetes)
4. Xigduo XR (Diabetes)
5. Brillinta (Cardiovascular)
6. Lokelma (Renal)
7. Roxadustat (Renal)

As part of a larger pitch, this company approached AdVenture Media to design a sophisticated budget allocation model for the campaign. We needed to develop a data-driven and intelligent approach to the paid budget, presenting clear guidelines for how a $30,000,000 paid search spend would be best allocated across all seven in-scope brands and therapeutic areas.

Challenge

The pharma prospect wanted to generate awareness amongst two distinct audience groups: Patients and physicians. Patients suffering from addressable diseases need to be made aware of the available drugs, and physicians diagnosing the patients need to know the drugs exist as a viable treatment.

We needed to first develop insight and strategy related to both target markets, and then build that insight into a dynamic, forward-looking allocation model. Our model had to allocate budget across each therapeutic area, but the budget allocation had to be further segmented, allowing an allotment for both target audiences.

The model also needed to account for our own insight and market research, predicting the therapeutic areas, specific drugs, and audiences that would generate the most significant business outcomes for the client.

Lastly, the entire model had to be dynamic. Meaning, the model had to work for additional drugs, fluctuating drug prices, and other variables impacting the company's marketing goals.

Building the Model

Therapeutic area classification
We first established three categories in which to classify each of the prospect's in-scope brands: Diabetes, Cardiovascular, and Renal. Included in the diabetes category are Bydureon, Qtern. Farxiga, and Xigduo CR; the cardiovascular category includes solely the Brilinta brand; and the renal category includes both Lokelma and Roxadustat.

Determining Total Addressable US Sales
In order to ultimately allocate budget across a range of drugs and audiences, we had to determine the total potential US sales by drug, represented by gross revenue.

Using public data available on GoodRX, we sourced the retail prices for a month’s supply of each in-scope brand. Using the average frequency of consumption rates, we were able to reach a reasonably accurate annual sales estimate, per person, in the United States for each brand.

When establishing the average retail price for the diabetes category, we first sourced the average retail price for each brand included in the category. These prices were then used to calculate the average retail price for the entire diabetes category. Finally, this price was multiplied by a factor of 1.5 to account for the possibility of overlap within the category as patients may be prescribed more than one of these medications conjointly (e.g. a patient taking both Qtern and Farxiga).

Addressable US Market Size
Now that we understood the total US revenue per person, we needed to model the total market size for each drug across the US.

In other words, with full market penetration, how much revenue could the company earn from each drug each year in the US?

To determine the total potential US customers for each drug:

1. Began with the total US population (~330,000,000)
2. Determined the specific ailments treated by each in-scope brand
3. Using market research and publicly available data, we found the percentage of US adults diagnosed with each ailment 
4. Multiplied the percentage of impacted population by the total US population
5. Subtracted the percentage of the total population by the percentage of uninsured population (sourced from the CDC).

Finally, we have the total number of individuals in each drug’s addressable market. We are also now able to calculate the total maximum US annual sales per in-scope brand by multiplying the total addressable market with the annual revenue per person.

The total US sales potential for the seven in-scope brands is $1.6 trillion dollars annually.

Brand Budget Allocation
Using the percentage of total revenue potential, we allocated a category spend, using the total $30,000,000 marketing spend as a baseline.

For example, Brlinta accounted for 13% of the total potential revenue, so the model gives Brilinta 13% of the available $30 million budget in the baseline scenario (​​$3,768,994).

To start, we allocated the budget equally across the Diabetes drugs (83% allocation).

X-Factor Adjustment
After establishing the spend by category, we included a variable referred to as “X-Factor” to allow for the flexibility to add or subtract from the allocated budget based on notable factors impacting the digital marketing strategy.

These factors may include additional marketing initiatives targeting a given demographic, patent expiration dates, emerging trends, market saturation and competition, consumer awareness, new market exploration—any dynamic changes that influence sales or digital marketing goals can be accounted for through the X-Factor.

In the case of Roxadustat, we applied an X-Factor deducting from the allocated spend to account for the fact that the medication is not currently available to the market.

Based on client feedback and background research, we applied X-Factor adjustments to each in-scope brand.

As the budget allocation among brands is adjusted to reflect "X-Factors", the total spend may exceed the available budget. The reallocated spend by brand ensures that the total spend allocation never exceeds the total available budget ($30 million annually).‍

Defining and allocating two target markets
Our two unique audience segments were labeled Health Care Providers (HCP) and Direct to Consumer (DTC).

To determine the breakdown of HCP versus DTC percentage allocation, we analyzed the likely impact of a successful search from both segments. An HCP click is more impactful as they have an established network and the potential to prescribe the medication to multiple patients, whereas a DTC click will generally only provide the company with one new potential consumer. 

For this reason, we allocated a larger percentage of the budget (65%-80%) to HCP for each brand — with the exception of Roxadustat where 100% of the budget was allocated to HCP to account for the primary influence HCPs have over the awareness of a medication at such an early stage, capping our recommendation at the maximum available budget for HCPs based on branded clicks.

Our HCP (healthcare provider) audience became our primary target audience.

Maximum Available Spend
Just because we’d like to allocate 65-85% of our budget to our primary audience (HCP) through paid search using Google Ads, it doesn’t mean that we’ll be able to.

To determine the maximum available budget for HCPs, we first established the average number of impressions for a paid search targeted to this segment for each brand.

We then determined a click-through rate (CTR) for each brand, using historical analyses and other industry-standard modeling tools.

We then multiplied that by the average number of impressions to calculate the total number of clicks from HCPs. Finally, we determined the average cost-per-click (CPC) and multiplied it by the total number of clicks to calculate the maximum available achievable spend on HCP.

The average number of impressions, click-through rate, and average cost-per-click have all been sourced from research tools and predictive modeling and are not wholly reliable—accurate data cannot be obtained until a campaign has launched.

Using our predictive modeling and advertising research, we’re able to determine the maximum available spend on Google Ads for our healthcare provider audience.

For some brands, our model predicts not being able to spend our ideal budget allocation on the primary audience (HCP) alone. For example, we wanted to spend 75% of our $7 million Bydureon allocation on healthcare providers, but we predict being only able to spend $3.8 million on Bydueron campaigns for HCP’s—only 54% of Bydureon’s total budget.

‍Reallocating leftover budget to DTC
The model dynamically and automatically calculates what remaining budget we have based on our CPC, Impressions and CTR research, funneling that budget to our DTC audience.

The total, adjusted spend by brand, broken out by HCP and DTC will always equal the total available budget, $30 million.

Dynamic and Complete
The foundational model is now complete, allowing us to easily make real-time adjustments as more information comes to light.

1. We can easily update the model to allocate more budget to one audience over another, and brand-level budgets will automatically be adjusted.
2. If a certain brand or drug becomes discontinued, the X-Factor adjustment can be used to quickly reallocate budget appropriately across the other six in-scope brands.
3. If retail prices change, the model will automatically update budget allocation based on total addressable revenue in the context of the new pricing.
4. If research reveals patients tend to refill a certain drug infrequently, the model will account for that data and update budget allocation accordingly.
5. If Google search volume for a certain drug increases or decreases, our model will adjust, allowing more or less budget for the primary target audience for that specific drug.
   

Visualizations

No model is worth its salt without colorful visualizations, and we made sure to deliver the prospect with charts aplenty.

Budget allocation by drug, segmented by target audience:

Total budget allocation by category and target audience:

Results

This big pharma organization needed a sophisticated budget allocation model that incorporated market research, business insight and predictive models.

Our allocation model took multiple factors into account, always with a focus on maximizing the client’s revenue.

The model is dynamic, allowing a full range of inputs to quickly update the model based on real-time data and updated goals.