With advertising becoming more complex, there is a strong demand for new advanced measurement capable of keeping up. Measurement needs to be flexible enough to model a wide variety of marketing situations that include different mixes of advertising spend, levels of ad effectiveness, types of ad targeting, sales seasonality, competitor activity, and much more.
The desire to create more effective marketing strategies drives demand for measuring marketing performance. Marketers need to understand the long and short term impact of media advertising, trade promotion, and other marketing tools to create effective tactical and strategic marketing plans. Quantitative analysis is expected to provide accurate, reliable measurement of key performance metrics and the effects of various marketing strategies.
Facebook and Google provide integrated experimental capability; they report incremental advertising effects measured through randomized experiments in their Conversion Lift and Brand Lift products. Experimental capabilities in other media, such as television, are rare. While randomized experiments are well understood, they can be expensive and impractical. Measuring cross-media effect requires intricate experimental designs and advanced planning, and a lot of time to test. Measuring small effects requires a large sample size. Experimental studies generally provide only a snapshot views of the marketing environment.
Marketing and Media Mix Modeling generates conclusions by observing and analyzing historical data, rather than data aggregated on the user level. Drawing causal conclusions from MMM requires modeling assumptions concerning the nature of the marketing environment (e.g., how advertising changes user behavior, how ad channels interact, how pricing impacts sales, etc.).
While marrying the experimental and observational methods could become expensive and impractical, it is possible to represent their complexities in a simulated ad system. With simulation, it is feasible and inexpensive to consider various scenarios for methodology evaluation and run virtual experiments to measure the complex interplay between consumers, marketing tools, and environmental phenomena. By simulating multiple reasonable market conditions, we can capture consumer behavior’s critical aspects, including complicated purchase and response to marketing techniques. These insights help modelers develop measurement methods that are robust to different market conditions.
We start by segmenting the consumer population into distinct groups based on several key features that characterize the consumer’s relationship with the category and the brand. The meaning and usage of each category and brand state follow below:
- Market State (Category): This represents the target population and those interested in the category. Those who are not interested in the category will not be affected by any marketing or media activity. For example, females who are not mothers will never buy infant milk no matter how many times they’ve been exposed to advertising.
- Satiation State (Category): This is related to the category’s purchase cycle and whether the category’s demand has been satisfied by past purchases. Satiated consumers may become unsatiated with time. Price promotions can create dips in sales and a depressed market in succeeding weeks.
- Activity Sate (Category): This tracks consumers’ location in the path to purchase funnel. Consumers could be inactive, researching, or in the final stage of deciding what brand to buy. In each step, consumers have different media consumption and different response to marketing activities. For example, consumers who are researching may make generic or branded search queries as part of the decision-making process. TV and radio will reach consumers in a wide range of activity states, while other media like paid search will tend to target a smaller number of consumers further along the path to purchase.
- Brand Favourability State: This measures consumer’s perception towards a certain brand. Usually, a high level of favourability yield a high probability to purchase. Brand favourability can be high or low for multiple brands simultaneously.
- Brand Loyalty State: Consumers can be loyal to the advertiser brand, loyal to a competition brand or switchers. Unlike favourability, loyalty is exclusive to a brand. Consumers may have high favourability to multiple brands but remain loyal to one brand.
- Brand Availability State: This refers to how physically or mentally easy the brand is to buy. For example, brands with low distribution amongst population grocery stores have low physical availability. In online space, search ads increase the brand’s mental availability.
Consumers’ mindset towards brands changes with time, due to the impact of various market forces. Inactive consumers may become exposed to a display or magazine ad, then move to explore and discuss with a friend, or view a product review on YouTube, then purchase the product at the brand’s website. There are different kinds of events that drive consumer migration from one step to another. Some actions evolve naturally; for instance, even in the absence of marketing intervention, consumers’ disposition toward a category and brand may develop. Marketing interventions, such as media advertising, trigger some actions, and influence consumer mindset and drive sales. Marketing can cause changes in activity state, brand favourability, brand loyalty, and brand availability. Usually, marketing is expected to move individuals from less favourable to more favourable states, thus increasing the total number of purchases. The purchase action can also change the consumer mindset; for example, consumers may become unsatisfied with the brand, or satiated or even loyal. For instance, following a successful purchase, consumers may become habituated or develop a strong preference for a particular brand.
How to model marketing interventions in detail is beyond the scope of this article. The approach mainly uses separate events to simulate each ad channel’s behavior and then sequence them into custom simulation scenarios. A particular media channel is specified with specific parameters. For example, in paid search, we specify which query volume, paid impression, paid clicks, and search spending is generated. Different parameters determine the query rates and click-through rates for each population segment to create separate branded and generic search events. For traditional media, we specify the values of parameters controlling the media channel’s audience size and composition, the media volume (GRPs) and spends, and media effectiveness according to their simulation needs. The modeling takes into account the effects of pricing and competition on advertiser and competitor sales. It also takes into account differences in purchase behavior between consumers belonging to different population segments. We also model post-purchase changes in consumer mindset in addition to calculating sales, revenue, and profit.