How to set your 2024 budget with marketing mix modeling
Jan 8, 2024
Introduction
For marketing leaders, January means annual planning season and justifying their budgets to executive leaders. While everyone aims to get this wrapped up before the holidays, it almost always bleeds into Q1.
This process is challenging since there’s no way to guarantee the revenue generated by the marketing budget. Many factors impact sales—like your product, the quality of reps’ pitches, or your company’s pricing—so it’s difficult to isolate the impact of marketing alone on revenue.
The next best option, however, is using Marketing Mix Modeling (MMM). With this tool, marketing leaders can use years of performance data to estimate (not guarantee) how channel spend will affect their team’s target metrics—such as sales, leads, and qualified traffic.
What is Marketing Mix Modeling?
Marketing Mix Modeling (MMM) is a form of statistical analysis that estimates the impact of each marketing channel on target metrics, such as qualified traffic and leads. Using one or more years of historical data, this model estimates how channel spending will impact target metrics. We recommend reading this comprehensive resource to learn more about MMM.
How to use MMM to evaluate and determine spend for each channel
Budgeting is the heart of the MMM process. Armed with data about how every dollar spent on a channel affects target metrics, marketing leaders can plan spending for each channel and create tests to confirm their model’s estimations. We’ll walk through the three key steps of using MMM to form budgets.
Build the model
Depending on your budget and internal team, there are two options for Marketing Mix Modeling: you can build the model yourself with a spreadsheet and R or Python code, or you can use a platform designed explicitly for MMM.
The former DIY option is possible, but it requires significant resources. You’ll need to hire a data scientist to lead your marketing analytics team throughout the building process. This group must know how to build the model on top of open-source libraries, like Google’s LightwightMMM and Meta’s Robyn. Beyond requiring expertise, the DIY approach takes anywhere from two to four months.
We recommend using specialized MMM software. These tools are often more cost-effective than building MMM on your own and are easy to implement.
Paramark puts MMM in the hands of every marketing leader. Rather than hiring a team, our users rely on a platform built by data science and marketing analytics experts. We created our platform following data science best practices, like checking model “fit” with a mix of techniques and testing model predictions against observed data.
Compare the relative costs of each channel
Once your model is set up, you should see the following data for multiple points in time across several years.
Target metrics (like pipeline generated or leads generated)
Marketing inputs for each channel (such as impressions, event attendance, or emails sent)
The costs of each channel
Use this data to compare each marketing channel’s relative cost for your target metric. Say your MMM indicates a dollar spent on Facebook ads generates more leads than a dollar spent on Google search ads. Given this target metric, Facebook ads are more cost-effective.
Create scenario planning around goals
Along with indicating the most cost-effective channels, your MMM lets you estimate how much you would need to spend on a marketing channel to generate your desired target metric output.
Say your MMM shows that a certain level of channel spend generated a 30% increase in your target metric, qualified leads. You share this data with your CFO, and they recommend increasing qualified leads by 40% to 50%. Using your MMM data, you can estimate the additional budget allocations you’ll need to hit that goal.
The word “estimate” is critical in this step. Marketers often assume they can scale outcomes linearly by increasing their spend on a channel. If I increase my budget for this channel by X, I’ll raise my target metric output by Y%. However, every channel has diminishing returns, so you can’t assume a certain budget will generate a specific outcome, and MMM helps account for those diminishing returns.
Tips for confirming and communicating the budget
It’s tempting to review your MMM and assume you know the best way to allocate your budget based on the data. But to reach the best outcomes for your marketing department, you don’t want to go into annual planning with just one suggestion. Be ready to negotiate and discuss multiple options with your CFO by preparing a variety of data and scenarios.
Start by presenting historical data
Before you share a budget with your CFO, explain how your marketing spend has impacted your target metrics over the last year using your MMM. This historical data will build your CFO’s trust in your forecasting and help them understand the reasoning behind your budgeting recommendations.
Consider covering these historical data points for the past year:
What marketing channels your team invested in and what target metrics you were measuring
The top 3 channels you spent the most on (and how much you spent)
The top 3 channels in terms of performance for each target metric
Total marketing contribution towards the output metric
How that output metric converts to new sales and bookings
Keep in mind, a large amount of historical data may overwhelm your CFO and cause them to lose interest. Instead, focus on what they need to know to understand the scenarios and budget recommendations you’ll present (as we’ll cover in the following steps).
Forecast best, worst, and base case scenarios
It’s tempting to present just one budget to your CFO, but this can lead to unreliable planning. You don’t want to waste your budget by failing to account for all possible scenarios.
Plus, your CFO may disagree with the one scenario you present. They’ll push for the budget they want, and you won’t have any leverage to negotiate.
Increase your odds of aligning with your CFO by forecasting multiple scenarios before meeting with them.
A best-case scenario: An unlikely but possible situation where you exceed your target metric goal.
A worst-case scenario: An unlikely but possible situation where your channels perform worse than expected.
A base-case scenario: The most likely situation where your marketing channels perform as expected.
Say you’re forecasting your budget for search ads. Your best-case scenario is that increasing your search ad budget by 10% will increase your target metric qualified leads by 10%, and your worst-case situation is qualified leads will remain the same or decrease. The base-case scenario factors in varying competition and keyword volumes with search, so it predicts qualified leads will increase by just 5%.
Back up recommendations with testing data, if available
Marketing Mix Modeling is ideal for estimating the incremental impact of marketing. But since it only approximates the effect, it’s best to combine MMM with past experimental results to support your suggested scenarios and build your CFO’s trust when budgeting.
Let’s return to the earlier example and imagine your marketing department ran experiments the last two summers about the impact of search ad budget on leads. All factors were identical, except the search ad budget was increased during the second summer. Leads increased during that time, so you could use this testing data to support your recommendation of expanding the search ad budget.
How to use MMM to optimize budget regularly
Along with helping teams create budgets, MMM is useful for regularly assessing your spend and reallocating as needed. Analyze MMM output data and in-channel reporting every quarter to determine what’s working and worth doubling down on (and what isn’t) with your strategic finance partners.
Monitor in-platform data for each channel weekly
Before making reallocation decisions with MMM data, first make sure you’re using each channel optimally. Make small monthly adjustments to each channel’s campaigns and content based on platform-level data. For example, you might optimize your YouTube videos’ thumbnails if the videos aren’t garnering enough views.
With these improvements, you can rule out poor channel practices as a reason for the outputs you’re seeing in your modeling. Here are several in-platform metrics you might check for each channel on a weekly or monthly basis:
Likes
Comments
Impressions
Reach
Plays
Compare MMM data to platform-level data every quarter
Unlike each channel’s reporting, MMM data isn’t worth analyzing weekly. You need time to make channel comparisons based on your target metric, so we recommend reviewing MMM data monthly and quarterly.
Check MMM data every month to determine how channels have impacted your target metrics since the last month. Then, use each channel’s reporting to stay on top of weekly performance.
Say your model indicates that Google non branded search ads are a cost-effective source of incremental leads. You check Google’s reporting, which also indicates that your cost-per-click is low, so you feel confident in increasing your search ad budget. The reverse would also be true—poor performance in both your MMM and channel reporting would suggest that you should decrease your search ad budget.
Sometimes, your MMM and a channel’s reporting will show opposite trends. For example, imagine Google’s reporting shows that the cost-per-click for search ads is low, but your MMM indicates these ads are an expensive source of leads. In this case, use experimentation to confirm how a channel impacts your target metric.
Request reallocations based on your data
Use your MMM and platform-level data about each channel to make informed budget recommendations for your CFO. Request a budget increase if your MMM and platform-level data both show a channel is positively impacting your target metric (and vice versa).
Consider the earlier example where both Google and your MMM show search ads are a cost-effective source of incremental leads. In this case, you would ask your CFO whether they would like to increase your search ad budget to increase leads.
Likewise, if CPC is high and MMM shows a decrease in leads, you could suggest decreasing the Google budget to your CFO. You might also recommend reallocating that spend to a channel that’s performing well based on your MMM data and platform-level reporting.
Optimize your budget with a specialized MMM approach
MMM is a powerful tool for understanding the incremental impact of your marketing channels. However, building a model and interpreting its results requires significant expertise. You’ll need data scientists and marketing analysts to:
Check your source data before building the model
Creating a model on top of open-source libraries
Interpreting the model’s “fit”
Finding ways to improve the model
Instead of building your own model, we recommend using an end-to-end solution built for MMM. Marketing Mix Modeling is a core pillar of our platform. If you’d like to learn more about how you can benefit from it, please reach out for a free consultation.
