Optimization of marketing budgets: Maximize your results and avoid regrets.

Optimization of marketing budgets: Maximize your results and avoid regrets.

In life as in business, let’s avoid regrets.

Nowadays, businesses have more data at their disposal than ever before to help them make informed decisions. However, using this information to make wise decisions can be a challenge. Among the tools used, the mixed media model (MMM) helps identify communication channels that have the most impact on sales.

Let’s illustrate with an example!

Company X is trying to determine the optimal allocation of its $1M budget between television, radio, and online advertising. The analysts used predictive models to estimate the return on investment (ROI) for each channel and obtained interesting results. The chart below illustrates the relationship between investments and ROI: initially, the more significant the investment in a medium, the higher the ROI. However, beyond a certain saturation point, the ROI either plateaus or decreases. It’s also noteworthy to mention the superior performance of online media, followed by radio, and then television.


If we assume that predictive models provide a reliable representation of the future, it’s also possible to use these models to determine the optimal allocation of the budget among the three channels. This strategy treats the issue as an optimization problem subject to constraints. Thus, the goal is to maximize the return on investment while adhering to budgetary constraints.

Various techniques can be employed to solve this problem, such as gradient descent. Using this method, it’s possible to achieve an optimal solution. In the current scenario:

Optimal Allocation 1

  • Radio : 420 000 $
  • TV : 0 $ 
  • Web : 580 000 $  

This allocation of funds corroborates the previous observation, with a larger portion allocated to online media and radio, while the amount allocated to television is zero. With this optimal distribution, an ROI of $1.9M, or 190% of the invested amount, is achievable. However, it may seem counterintuitive and risky to completely cut the television budget…


In decision theory, regret is defined as the difference between the outcome achieved following a decision made and the best possible outcome that could have been achieved if another decision had been taken. In the context of optimizing marketing budgets, regret can arise when actual results diverge from the predictions of the model used.

Let’s assume that Company X decides to strictly follow the results of the previous optimization and completely eliminate the television budget. In the following quarter, after gathering results and updating the predictive models, it is discovered that television performed better than in the previous quarter and that the models were not entirely accurate. The estimated ROI with this budget allocation is 130%.

After conducting an optimization based on the new models, it’s also observed that the following budget allocation would have generated a better return on investment, yielding an ROI of 170%:

Optimal Allocation 2

  • Radio : 230 000 $ 
  • TV : 170 000 $ 
  • Web : 600 000 $ 

The regret in this case would therefore be: 170% – 130% = 40%.

Minimization of Regret: Make agile and informed decisions.

Rather than limiting oneself to optimization based on deterministic models, it’s crucial to minimize regret by considering various possible scenarios. This approach allows you to make agile and informed decisions by anticipating the performance variations of different media and market uncertainties.

For instance, let’s imagine a scenario where television’s performance surpasses projections next quarter, while radio experiences a slight decline and online media maintains its performance. By utilizing techniques like Monte Carlo simulations, you can model these variations and many others. This enables you to arrive at a solution that minimizes potential regret across these situations, typically resulting in a more balanced budget allocation, such as:

Optimal Allocation 3

  • Radio : 300 000 $
  • TV: 150 000 $
  • Online : 550 000 $

This allocation might not be as efficient if the deterministic model proves to be accurate. For instance, by choosing the optimal allocation3, one would achieve an ROI of 170% and a regret of 20% (190% – 170%) for the first scenario1 . However, such an approach would minimize risks and optimize the return on investment in the context of uncertainty. One would achieve an ROI of 155% and a regret of 15% (170% – 155%) for the second scenario2.

This approach allows you to leverage the insights from your predictive model while safeguarding against potential risks and market uncertainties.

If you, too, would like to live without regret, contact our experts; they will know how to help you create your own mixed media model! Click here to learn more about our marketing performance services: https://numea.ca/services/performance-marketing/