Marketing Budget Optimization: Lightweight MMM vs SaaS MMM

Have you ever wondered how to get the most out of your marketing budget? Like a savvy traveler planning a trip with limited funds, marketers must figure out how to best distribute or allocate their resources for maximum impact. This is where budget allocation optimization comes into play, ensuring every dollar or euro works as hard as possible towards achieving your business goals and increasing your ROI.

But how exactly do you navigate this process? Which tools and methods should you adopt to make sure your marketing budget is not just spent, but invested wisely? In today's dynamic and often overwhelming digital landscape, making these decisions can be daunting. Maybe you have heard the famous quote of John Wanamaker from the late 1800s claiming that: “Half the money I spend on advertising is wasted; the trouble is I don't know which half”. Since he was one of the global pioneers in marketing, his statement has been defining the career paths of many marketers world wide. However, in the current day and age it doesn’t have to be like that anymore.

Two notable options that have been gaining traction are Lightweight MMM (Marketing Mix Modeling) developed mainly by Google for budget allocation and SaaS-based MMM ((software as a service marketing mix modeling) Budget Optimizer or Budget Planner solutions. In the case of the latter we will focus mainly on the Sellforte Media Optimizer which is our tool for optimal marketing budget allocation. Each of these approaches offers a unique set of features, capabilities, and benefits that can help streamline your budget allocation process.

In this blog post, we'll dive deep into the world of budget allocation optimization. We'll explore Google's Lightweight MMM, discussing its strengths and areas where it might fall short. We'll also delve into SaaS MMM, dissecting its advantages and potential limitations. By comparing these two methods, we aim to provide you with a clearer understanding of which tool might best align with your budget optimization needs. So, get ready, it’s time to ensure that both halves of your marketing budget are actually a measurable investment.

If you want to know what to know more about Lightweight and SaaS MMM please read this article.

Media Optimization and MMM: Why is it important?

In today's fast-paced and interconnected digital landscape, Media Optimization as part of MMM has emerged as an essential tool for any marketing strategy. Media optimization is a significant aspect of MMM. It pertains to the process of effectively allocating marketing resources across different media channels to optimize the return on investment (ROI). Optimization, in simple terms, refers to a process of finding the best solution to a problem or making the most efficient use of available resources. In the context of budget allocation, optimization helps determine how to allocate funds effectively in order to obtain the best ROI for each media channel or campaign. So, in a nutshell, media optimization in marketing is all about smartly spreading your resources across various media channels like TV, online ads, social media, etc., to ensure you get the best bang for your buck.

Why is it important?

• Better Resource Allocation: It ensures the most efficient use of your marketing budget by helping you understand where to invest for maximum impact.
• Informed Decision Making: It offers insights into the effectiveness of various marketing channels, enabling data-driven decisions.
• Improved ROI: By optimizing media spend, businesses can maximize their return on investment.
• Future Planning: Through the predictive power of MMM, businesses can foresee potential outcomes of different marketing strategies and plan accordingly.
• Competitive Advantage: Understanding your marketing performance at a granular level can give you an edge over competitors who are not leveraging these insights.

We have established that in order to optimize budget allocation based on data driven insights, one should use an MMM solution. The true question is: Which one? There are several available on the market, however, they all have their key differences even if at a glance, they seem to be similar. Let’s dive into the Lightweight and SaaS MMM comparison.

What are Lightweight and SaaS MMMs?

Lightweight MMM, an open-source Python library, serves as a robust instrument for marketers implementing Bayesian Marketing Mix Modeling. It leverages the Bayesian algorithm and Numpyro backend, enabling users to incorporate their current data into the model for enhanced accuracy and precision in their analysis. Bayesian marketing mix models utilize probability distributions to represent uncertainty in data. Bayesian approaches allow the model to update its predictions as new data becomes available, making it particularly effective for navigating uncertain or rapidly changing marketing environments.

Bayesian modeling is particularly beneficial in marketing because of its flexible structure for handling data uncertainty and variability. By utilizing probability distributions to signify uncertainty, Bayesian models also offer the possibility to incorporate business intuition by specifying the prior distributions for the parameters.

Another significant benefit of Lightweight MMM being open-source is that it's accessible for anyone to use, adapt, and enhance. However, the library is mainly developed by Google’s employees and they do not engage frequently with the public nor do they bring forward continuous updates.

To put it simply:

Lightweight MMM is like a free tool kit for marketers. It's an open-source library built in Python, meaning anyone can use or tweak it as needed.

SaaS MMM represents a sophisticated, cloud-based solution for marketing analysis and strategy optimization. This model employs advanced data analytics techniques to evaluate the effectiveness and contribution of various marketing channels and strategies on key business metrics like sales or market share. However, Sellforte in particular uses Bayesian models. For an in-depth understanding of the precision of Bayesian models, check out this article.

In a SaaS setup, the MMM solution is hosted on the cloud and accessed via the internet, eliminating the need for local installations and maintenance. This delivery model provides enhanced scalability, flexibility, and cost efficiency. Users can access the platform from any location with an internet connection, facilitating real-time insights and data-driven decision-making. Additionally, SaaS set up is handled by the provider, meaning that you will not have to build up the model on your own with your resources.

Furthermore, the SaaS delivery model facilitates rapid deployment of updates and new features, ensuring that users have access to the latest analytic techniques and functionalities.

To put it simply:

Think of SaaS as a tool that you can use over the internet, without having to download or maintain it. It's like watching a movie on a streaming service, instead of shooting your entire feature film from scratch. You don't have to worry about storage space or upkeep; all you need is an internet connection.

Lightweight vs SaaS MMM: Budget Allocation Optimization

Set-up SaaS

Choosing a SaaS-based MMM solution means opting for simplicity and convenience. The initial setup, handled entirely by the provider, removes the complexity often associated with such sophisticated analytics tools. This ease of deployment is one of the key advantages of a SaaS approach. From the installation to the ongoing maintenance and updates, everything is taken care of by the provider and is included in the subscription, freeing you from the technicalities and allowing you to focus on your core marketing tasks.

Your provider will carefully assess the quality of your existing data which is a critical input for your MMM. Data forms the backbone of any marketing mix model, and its quality directly influences the accuracy of the insights the model can deliver. Your provider will guide you on how to refine your data collection and management practices, if necessary, to ensure that your model is fed with high-quality, relevant data.

Once your MMM is established and functional, it becomes a powerful tool in your marketing toolkit. It will allow you to optimize your budget allocation by revealing which marketing channels and strategies contribute most to your return on investment (ROI). Whether it's online advertising, social media campaigns, television spots, or in-store promotions, MMM will provide a quantitative assessment of their impact on your key business metrics.

Set-up Lightweight

Conversely, the Lightweight approach to MMM demands a hands-on, self-managed approach. Unlike the fully serviced SaaS solution, the setup, operation, and maintenance of Lightweight MMM rest entirely on your shoulders, or more accurately, on your data science team.

To deploy Lightweight MMM, your data scientists must set up and run the necessary Python libraries, write the required code, and manage the data inputs manually. This includes ensuring the proper functionality of the code, fixing potential errors, and continually maintaining and updating the system as required.

Data quality, a crucial factor in any MMM, also falls entirely under your review. The quality of the data being fed into the model directly impacts the quality of the insights derived. Therefore, your team must continuously assess and improve the quality of the data to ensure the effectiveness of your MMM.

Curves SaaS (Sellforte MMM)

Sellforte’s MMM leverages a negative exponential function to model the relationship between marketing spend and uplift. This is often referred to as a diminishing returns (DR) curve. You can read more about DR curves here.

Imagine you're on a roller coaster ride. The moment it starts, there's a significant initial incline (like the substantial impact of the first units of marketing spend), but then the ride slowly levels off at the top (just as the incremental impact of further marketing spend reduces). This progression outlines a concave shape, much like the curve in the diagram below, which signifies the fundamental assumption of the DR curve.

The assumption with this approach is that every additional euro invested in a particular marketing channel will yield a lower return compared to the previous one. In other words, the first euro you spend in a marketing channel has the highest ROI, and each subsequent euro provides a smaller ROI. This is why it's referred to as diminishing returns - the benefits (or returns) diminish with each additional unit of investment.

Each DR curve is characterized by two key parameters: shape and saturation point. The 'shape' refers to the steepness of the curve or how quickly the returns diminish. The 'saturation point' is the level of spend beyond which any additional investment will not yield any increase in return.

The DR curve helps identify this saturation point for each marketing channel. It's a data-driven way to say, "When approaching this point, your channels are becoming completely ineffective" This insight is critical for marketers when planning and optimizing their budget allocation to ensure maximum ROI. Essentially, this mathematical function ensures that you have a correct estimation of the saturation point every time you run the optimization function.

Curves for Lightweight MMM

The Lightweight MMM employs the Hill function, resulting in a response curve that exhibits varied behavior. Sometimes, it might resemble the Sellforte diminishing returns (DR) curve, where the ROI decreases with each incremental euro spent beyond a certain point (as depicted by the green curve in the example below where the incremental uplift decreases right from the start. The first euro is the best and then it’s becoming progressively worse). At other times, it can take on an S-shape, indicating different stages of ROI progression with increasing marketing spend.

However, one inherent complexity with the Hill-shaped response curves is the potential for misunderstanding the curve's inflection points, which might lead to suboptimal marketing decisions. For instance, the S-shaped curve might suggest a saturation point, indicating a plateau in ROI with increasing spend. But, in reality, this might just be a temporary dip, with the curve escalating and yielding an incremental ROI beyond this point (i.e. the blue curve in the example below).

This ambiguity with the S-shaped curves can be a disadvantage in certain scenarios because, unlike the consistent diminishing return principles of the DR curves, these Hill curves could lead to less accurate interpretation of marketing spend efficiency.

Output of SaaS (Sellforte MMM)

The output from the Sellforte’s MMM solution is an interactive dashboard. This dashboard doesn't just display your optimization results but also allows you to explore multiple potential scenarios based on different budgets, thereby enabling you to find the best strategy to meet your marketing objectives.

The optimizer within the dashboard is flexible and allows for a range of inputs. You can optimize based on historical budget spend, indicating how your previous budget allocations have performed. Alternatively, you can input a new budget or set specific sales targets to see how to best allocate resources to reach those goals.

A noteworthy feature of the optimizer is its allocation benchmark, which is usually set at approximately ±30% of your previous channel spend. In essence, it suggests that a given channel's budget allocation would not typically deviate by more than 30% from its previous allocation. This benchmark serves as a helpful guideline for budget distribution while preserving some continuity with past strategies.

However, the optimizer's flexibility doesn't stop there. You can override this default range and give the optimizer free rein, allowing it to explore a broader spectrum of allocation possibilities for potentially higher returns.

Below, you can see some images from the interactive dashboard. These provide a visual depiction of your budget allocation, potential returns, and different optimization scenarios. The user-friendly interface of the dashboard allows you to easily interpret and utilize the insights generated by the Sellforte’s MMM for effective decision-making. In case you want to try the dashboard for yourself, you can check out our live demo here.

Output of Lightweight MMM

Lightweight MMM, on the other hand, has a more simplified and hands-on approach compared to the SaaS MMM solutions like Sellforte.

By default, Lightweight MMM assumes that you wouldn't want to drastically change your historical spending habits, so it has a default deviation of 20%. Essentially, this means the media optimizer will aim to improve your budget allocation within the realm of a 20% change from your previous spend. However, it's not a fixed constraint; it can be adjusted to fit your specific needs if you change the original code.

The optimization process in Lightweight MMM is primarily based on your historical spend, taking into account how your past budget allocations have affected your outcomes. While this is a logical and efficient starting point, it also implies that changes in marketing strategies or market conditions would require manual adjustment in the model.

In terms of the output presentation, Lightweight MMM offers a more modest approach, providing images and simple plots instead of an interactive dashboard. This minimalist approach can be easier to interpret and work with, but it lacks the dynamic and interactive capabilities of a dashboard.

Moreover, Lightweight MMM does not inherently support scenario planning. Unlike SaaS solutions, where you can experiment with multiple potential scenarios within the dashboard, Lightweight MMM requires you to rerun the optimization with different inputs each time you wish to explore a new scenario or make any adjustments. This could mean more time and resources are spent on re-optimization and analysis. You can see an example of how the output would look like below.

Speed SaaS (Sellforte MMM)

The concave shape of the return curves in Sellforte's MMM has a significant mathematical implication: it results in a convex optimization problem.This is due to the relationship between a function and its curve: if all the curves are concave, then the optimization problem is convex, otherwise it isn’t.

Convex optimization problems are a special subset of mathematical optimization problems that have some beautiful and powerful properties. They are so named because the feasible region - the set of solutions that satisfy all constraints - is a convex set. In other words, if you were to draw a line between any two points in the feasible region, that line would entirely lie within the feasible region.

Why is this important? Well, in a convex optimization problem, any local minimum is also a global minimum. This is a fancy way of saying that if you find the lowest point in any one area (local minimum), it is also the lowest point overall (global minimum). This is a powerful guarantee because it ensures that you will always find the optimal solution, no matter where you start your search.

Moreover, convex optimization problems have efficient and robust algorithms for finding this global minimum. In the case of Sellforte's MMM, the algorithm is so efficient that it can compute the optimal budget allocation in less than a second! This offers significant advantages in terms of speed and reliability in arriving at the optimal marketing spend allocation. Therefore, it is actually instantaneous and always correct.

Speed Lightweight MMM

On the other hand, the Lightweight MMM takes a different approach with its use of the Hill function. This method has its own unique strengths and challenges.

One consequence of the Hill function assumption is the longer computation time. The math behind this S-shaped curve is more complex, and as a result, running the optimization algorithm can take considerably more time. Depending on the amount and complexity of the data, computation times can extend up to 10 minutes or more.

Furthermore, the Lightweight MMM lacks an inherent capability for scenario planning. Unlike some SaaS MMM solutions, you can't simply input different parameters into an interface and instantaneously see the results of various scenarios. Instead, each potential scenario or adjustment requires you to rerun the entire model, which can be a time-consuming process.

This requirement for manual adjustment and re-optimization is a significant consideration for businesses with dynamic marketing strategies or fast-changing market conditions. They will need to allocate the necessary time and resources for these repeated reruns of the model.

Before & after optimization comparison SaaS (Sellforte MMM)

One of the standout features of Sellforte's MMM is the budget allocation comparison function. It enables you to evaluate, side by side, the potential outcomes of different budget allocations.

This feature helps to shine light the complex world of marketing investment decisions. By viewing different allocation strategies in a comparative format, you can understand the potential impact of each decision more clearly. What if you allocate more budget to digital ads and less to TV commercials? How does increasing your social media spend affect your overall return on investment? These are the types of questions that the comparison function can help answer.

This level of insight is invaluable when it comes to marketing strategy planning. It allows you to anticipate the potential consequences of your decisions, giving you the confidence to choose the allocation strategy that best aligns with your business goals.

Below you have a few examples of budget allocation scenario comparisons. Alternatively, you can also try out to plan your own scenarios in our live demo.

Before & after optimization comparison Lightweight MMM

If you want to evaluate and contrast different budget allocation scenarios in Lightweight MMM, you will need to manually adjust the functions you are using and generate the comparative plots from scratch.

This task demands a strong proficiency in data manipulation and visualization, and an understanding of the Lightweight MMM's underlying mechanics. While this provides a high degree of control and customization, it also requires substantial time and effort. Moreover, each alteration or new scenario necessitates a rerun of the entire model, which further adds to the complexity and time commitment.

Below you have a few examples on how the budget allocation comparisons could look like in Lightweight MMM.

Summary Table

Let’s recap the core differences between Sellforte MMM (SaaS Solution) and Lightweight MMM (Self serve solution). While there are many different aspects between the two, we will focus on the crucial ones we touched upon previously.

Curious to learn more? Book a demo.