Marketing Mix Modeling - How Does Advertising Really Work?
Enormous amounts of money are being invested in advertising worldwide. The Dentsu Aegis Network expects $ 609.9 billion in advertising spending in 2019. The rising advertising investment is driven by digitalization and consumer behavior, and results in a vicious cycle: The flood of advertising messages that is washing over customers and users leads to increasing marketing tolerance, and even frustration. However, companies need to keep up in the fight for the favor of (potential) customers. The big challenge that is facing companies is to use a given advertising budget efficiently, and at the same time advertise purposefully such that advertising meets the customer when it has the most leverage - without being overwhelming, repetitive, or irrelevant.
With Marketing Mix Modeling, we can help to overcome this challenge.
Only companies that act on a data-driven basis can differentiate whether revenue increases are actually generated by advertising contacts or, for example, by an increase in demand due to seasonal trends. Statistical models can also identify which of the many advertising channels are efficient. Comprehensive knowledge of the efficiency of each channel is essential to optimally allocate the advertising budget, avoid poor investments, and create competitive advantage through a tailor-made advertising strategy.
With Marketing Mix Modeling, the advertising budget can be controlled and evaluated using data. An optimal distribution of the given advertising channel budgets (online, print, TV, etc.) is achieved, such that each euro of the advertising budget maximizes profit. This is where customers benefit, too, as they only receive advertising where it proves effective instead of being overrun and scared away by an avalanche of advertising messages.
Basically, Marketing Mix Modeling is an approach to comprehensively understand, quantify, and evaluate a company’s advertising impact.
In order to understand advertising impact and identify interactions, the various time series of ad spending and revenue per day are examined. This quantifies the effect size of each channel included in the marketing mix. For example: Is additional revenue more likely to result from TV advertising? Or by online marketing actions such as display advertising to potential new customers? Or is it actually the combination of the two channels that motivates customers to buy?
Statistical models are used to assign the advertising effect to the individual channels. These allocate the generated sales into the additional contributions induced by the various advertising channels.
After understanding the interactions between marketing channels and sales, this knowledge is used to optimally divide future spending between channels and to prevent a misallocation.
There are many subtleties to consider when modeling. The selection of a statistical model depends largely on the requirements of the challenge to be tackled. For example, an extremely simple Marketing Mix Model can be achieved with a multiple linear regression model. But if the full potential of the data structure underlying this question is to be exploited, more complex models are needed.
Marketing Mix Models should consider the following important aspects:
- Time Delay: Responses to advertising are often delayed. For example, customers receive a newsletter, but only open it later.
- Interaction: The individual advertising channels influence each other. For example, customers may be encouraged by a TV campaign or print ad to respond indirectly (e.g., through increased online research for the advertised product) to the ad campaign. This creates an interaction between the TV/Print channels and online marketing actions.
- Long-term Effects: Even a one-time marketing campaign can have long-term effects. The short-term effect is reflected both directly and promptly in the sales figures, but can also provide additional positive effects downstream. Although these decrease with time, they must be evaluated with regard to their cumulative effects over time.
Including these points in the model poses a challenge for the analyst. Standard models such as multiple regression can not take these considerations into account. Disregarding these points, however, would lead to erroneous results and inadequate - or, in the worst case, incorrect - conclusions.
Using multivariate econometric time series models, all aspects described above can be taken into account. This flexible model class makes it possible to investigate a large number of advertising channels at the same time, thus revealing the not-directly-observable interactions between different channels. By taking into account the interactive effects that exist over a longer period of time, completely new insights into the advertising effect on the customer base are created. This makes it possible to determine the advertising elasticities taking into account the interactions, and thus to determine the long-term optimal budget and the optimal allocation of the advertising material use.
- What are time series data? Data collected repeatedly over a long period of time.
- What is econometrics? Methodology in which data are analyzed with the help of statistical models to understand economic issues.
- What does multivariate mean? The simultaneous influence of several mutually-correlated quantities is investigated.
- What is elasticity? Measure for the relative change in the size of a variable (here, turnover) due to the relative change of a causal variable (here, channel-specific advertising pressure).
The simultaneous modeling of direct, time-delayed, and seasonal effects, taking into account the direct and time-delayed interactions between the advertising channels, requires a high degree of statistical know-how and experience in this area. As a data science company and statistical consultancy, we are your Marketing Mix Modeling partner, helping you to understand and optimize the advertising impact of your business.
The leverage of a custom marketing mix solution - compared to, for example, SaaS solutions - is enormous, particularly for larger companies. With an optimized advertising budget, more customers can be more efficiently (i.e. with less budget) motivated to buy, which can be a decisive competitive advantage in the market. It is essential to address today’s customers - overwhelmed by advertising - precisely and purposefully, and to deliver tailored advertising where it has the most impact.
Since every market and every business is different, a simple, unified software that determines the optimal advertising budget and distribution for all companies is illusory. SaaS solutions that offer marketing analytics products with a one-click interface may be worthwhile for small and medium-sized businesses in terms of cost-benefit. Tailor-made Marketing Mix Models, however, provide a much better cost-benefit ratio for larger companies.
Marketing Mix Modeling is the supreme discipline in the field of marketing analytics and includes control of advertising measures in their entirety. The data-driven evaluation of a marketing mix and its associated optimization gives companies a clear competitive advantage.