Marketing Analytics

Essential Tools for Data-Driven Decisions

Rajkumar Venkatesan, Paul W. Farris, Ronald T. Wilcox

• Publisher: Darden Business Books
• Year: 2021
• eBook ISBN: 9780813945163
• Topic: Business
• Subtopic: Business Education
• Index: Business

The authors of the pioneering Cutting-Edge Marketing Analytics return to the vital conversation of leveraging big data with Marketing Analytics: Essential Tools for Data-Driven Decisions, which updates and expands on the earlier book as we enter the 2020s. As they illustrate, big data analytics is the engine that drives marketing, providing a forward-looking, predictive perspective for marketing decision-making.

The book presents actual cases and data, giving readers invaluable real-world instruction. The cases show how to identify relevant data, choose the best analytics technique, and investigate the link between marketing plans and customer behavior. These actual scenarios shed light on the most pressing marketing questions, such as setting the optimal price for one's product or designing effective digital marketing campaigns.

Big data is currently the most powerful resource to the marketing professional, and this book illustrates how to fully harness that power to effectively maximize marketing efforts.

1

Resource Allocation

Imagine you need to decide how much of your budget to spend on online versus television advertising in a certain market, or which product to make the focus of your new campaign. How much will you invest in which marketing activities to optimize your investment? In other words, how will you allocate your marketing resources? Ideal resource allocation is the optimal level a company spends on each of its marketing levers (campaigns, new products, and so on) to maximize success. Figuring out how to do this depends on a range of variables, and which of those variables you choose to focus on is the crux of marketing analytics and the essence of managerial intuition.

Resource allocation is the endgame of marketing analytics, a framework that ties together the various tools and techniques to serve a firm’s strategic decisions. As such, it is important to keep in mind throughout this chapter and the rest of this book. The chapters that follow this one will delve much more deeply into tools and scenarios, but for now we focus on the four main steps in resource allocation. As we do so, we present ways to make the resource-allocation process more data driven.

Please note that in this chapter, we’ve included hypothetical cases and examples to familiarize you with the case method and our approach. In the first hypothetical case, about a pharmaceutical company, you are guided through the problem-solving process. In later cases in this book, both hypothetical and real, the steps may not be as explicit, giving you the opportunity to choose how to solve the problem, including which metrics and which analytics technique to use, enabling you to build your business intuition as you hone your analytical skills.

RESOURCE ALLOCATION: FOUR STEPS

1. Determine the objective metric that captures the business outcome of interest.
2. Develop a function that connects marketing inputs to the objective metric.
3. Using historical data on marketing inputs and the objective metric, employ marketing analytics to estimate the unknown values or parameters of the function identified in step 2.
4. Now reverse the process: keep the parameters identified in step 3 fixed, and find the optimal marketing input levels that would maximize the objective metric to determine the best resource allocation.

Step 1: Determine the Objective Metric

The first step of resource allocation is to determine the objective metric. The company needs to set a goal—such as maximizing profits or improving brand awareness—and then choose a metric to measure its progress toward that goal. Metrics often used for this include: paid search click-through rates, conversion rates from website visits, net profits, customer lifetime value, near-term sales lift, new buyers, repeat sales, market share, or customer retention rates.

Step 2: Develop a Function that Connects Marketing Inputs to the Objective Metric

The second step is to connect the firm’s marketing inputs to its resource-allocation objective metric. Managerial intuition is of paramount importance in this stage, as it allows the marketer to correctly decompose a metric. If a company is examining gross profits, for example, what components of the business contribute to those profits? Unit sales is a function of price, advertising, salesforce, and trade promotions. Advertising, salesforce, and trade promotions also incur marketing costs. Gross profit is sales minus unit costs, and net profit is gross profits minus marketing costs. Thus manipulating marketing inputs can improve sales, but the different inputs are also cost centers.

Once the marketing inputs are mapped to the objective metric, the marketing manager must determine which relationships between components that contribute to the metric are computational and which are empirical.

Computational relationships, also called accounting identities, can be calculated directly using known quantities. An example of an accounting identity is net profit: if both gross profit and marketing costs are known, net profit is simply gross profit minus marketing costs.

Empirical relationships are more complex, and driven by unknowns. They are functions, or transformations, that reflect analysis of historical data. An example of an empirical relationship is the relationship between marketing costs and unit sales. Unit sales cannot be obtained by a direct sum of the investments in marketing (as mentioned, these include price, advertising, salesforce, and trade promotions). The relationship is termed empirical because it is based in observation: the manager must analyze historical data to develop a function that transforms the marketing inputs into unit sales. For example, the manager could analyze the historical data on price and unit sales and develop a function that describes their relationship. The transformation function ideally yields a weight that translates a product’s price into unit sales; in other words, the weight is the amount that unit sales would be expected to change with every unit change in price. These weights do not provide perfect transformations: they are best guesses based on historical data, wherein several factors—not just price—affect unit sales.

The main difference between an empirical relationship and an identity relationship is that an empirical relationship implies a best guess or prediction, while an identity is certain.

Step 3: Estimate the Parameters of the Function Identified in Step 2

The third step in the resource-allocation process is to estimate the best values for the unknown weights of the empirical relationships identified in the second step. A common method for identifying these weights is to build an econometric, or regression, model. For example, consider a problem where unit sales is the objective metric and price is the marketing input. Let’s say in step 2, we specified the function connecting price and unit sales as: unit sales = a + b × price. We don’t know a and b because unit sales and price have an empirical relationship. In step 3, we collect historical data and use marketing analytics to find the values of a and b in the context we are analyzing. This process of finding the values of a and b is called parameter estimation.

In this model, the independent variables are what we have been calling the marketing inputs: in the current example, this is price. The dependent variables are the potential metrics that a business can use to assess its resource-allocation efforts: in the example, this is unit sales. The aim of the model is to identify which marketing inputs of interest (price, advertising, sales calls, and so on) should be considered as having an effect on the dependent variable.

Once the unknown values in the empirical relationship between marketing inputs and objective metrics are estimated, the marketing manager can begin to predict the objective metric at different levels of the marketing input. This is the mathematical model that describes the relationship between the independent variables (price, advertising, sales calls, and so on) and the dependent variable (market share, profits, customer lifetime value, and s...

Table of contents

• Cover
• Title Page
• Copyright
• Contents
• Foreword
• Introduction
• 1. Resource Allocation
• 2. Cluster Analysis
• 3. Conjoint Analysis
• 4. Linear Regression
• 5. Customer Lifetime Value
• 6. Marketing Experiments
• 7. Paid Search Advertising
• 8. Text Analytics
• 9. Logistic Regression
• 10. Recommendation Systems
• 11. Automation of Marketing Models
• 12. Implementing Marketing Analytics
• Acknowledgments
• Notes
• Further Resources
• Index