Join our Analytics for Humans Facebook community

In the first part of this article, we explored the true definition of CAC as a metric and concluded that:

CAC is a simplified mathematical model that serves as a proxy for your advertising delivery efficiency.

Being a model, which is a simplified representation of a complex system, we have realized that just optimizing our ads on CAC alone blindly is destined to deliver us suboptimal results, and we need to understand the behavior of the “CAC Model” a lot more intricately to be able to truly optimize our promotional efficiency, and that’s the topic of today.

Today, we are going to focus on understanding the change in the behavior of your CAC as critical variables, such as time and spend, changes on your specific ads or platforms.

If you are taking away anything from this article, here is the central point:

“ Traditionally, we believe that the lower our CAC is, the better the ad is - and this belief is absolutely wrong because CAC is a measurement of efficiency, not the outcome (like Revenue or Profit) ”.

Instead, our CAC is the best when the ads are delivering the highest level of outcome (Net Revenue or Profit) for our company, and it is almost never at its peak efficiency.

To understand how we get to this conclusion, and how to find your own optimal CAC, let’s start with a discussion of marginal efficiency.

Marginal Efficiency’s Behavior in a Single Ad Situation

Let’s begin our discussion by defining a few key terms we will be using throughout this article:

Accumulated Revenue: The ultimate lifeblood of your business, the total amount of money you earn from running your ads.
Marginal Revenue: The amount of revenue you gain from running your ads as you spend one unit more money on it.
Marginal Efficiency: The efficiency of your ads as you are spending more money. Or, the projected spend it will take for an ad to generate per unit of revenue, usually measured by CAC.
Break-Even Point: Price of your product minus the variable cost (cogs, shipping, taxes) of the product, notice that advertising cost is not included here. This means that your CAC has to be below this point for your ads to be “really” profitable for your company.
Accumulated Profit: Accumulated revenue - total variable cost - advertising spend. If you are a bootstrapped business, you will be focused on maximizing this metric.
Marginal Profit: Gain/loss in profit as you spend more money on your ads.

Just from the onset, you probably have realized that CAC, or marginal efficiency, is not the end goal of your business.

Instead, it is a secondary measure that helps you understand how close you are to your ultimate business goal - whether that’s higher revenue or higher profit.

Before going too far ahead of ourselves, let’s first start by examining the behavior of our CAC, or marginal efficiency, as we spend more money on it.

In the theoretical world, your marginal efficiency of an individual ad/adset should roughly follow the following pattern (assuming no learning or other time effects on ads).

The movement of your CAC illustrated above can be divided very roughly into 4 key steps:

Learning: As your ads are created and being trained, its CAC will slowly decrease as the ad becomes more optimized for delivery on advertising platforms.
Acceleration: This decrease of CAC will accelerate rapidly as training happens, as your content resonate with your audiences. This is the moment where your marketing team is incredibly excited that something actually worked.
Performance Peak: The decrease of CAC will gradually topple off, reaching somewhat a balance at what many would call the “optimal CAC” (we will challenge this premise later). This is what most people are aiming for in terms of the performance of their ads.
Inefficient Scaling: As you further increase spending on an ad, your CAC will start going back up due to audience saturation, this is where most marketers will become scared and start scaling back spending.

After step 4, your ads CAC may continuously increase until oblivion and the ads becoming the lost driver of your company.

Optimizing Your Ads when you only have one ad running

Most people believe that the ad is serving you the most when it is reaching its performance peak, or when its CAC is the lowest - but this notion is simply incorrect.

To illustrate, let’s introduce another graph that represents the profit you are getting from your ads as your spending increases.

In the graph, the red area represents a negative profit (since higher CAC = higher cost), while the green area represents positive profit.

One immediate thing we notice on the graph above is that if we optimize our ads for the lowest CAC, we will only capture a minor portion of the potential profit of the ad.

In fact, as long as our CAC is below our break-even point, we will always maintain a positive marginal profit from our ad, so the optimal stopping point for most of us advertisers out there is when our CAC increases back up to our break-even point.

However, the story does not stop here. One key metric that was not mapped in this graph is accumulated revenue, which will continue to go up (albeit slower and slower) as you spend more money on the ad.

For companies with a traction/revenue optimization strategy, maintaining zero profit from your advertising campaigns might be an even more optimal strategy as it allows for more revenue and customers to come through your doors - assuming that you have a healthy amount of capital to keep cash flow positive.

So here is a simple three-fold rule for optimizing your advertisement under a “single ad” situation:

Your ad is NEVER the most efficient when your CAC is the lowest.
If you want to optimize profit, you should stop spending when your CAC equals to your break-even point on its way back up.
If you want to optimize for growth, you should stop spending when the total profit contribution of this ad is zero when CAC is on its way back up.

Please note that you should almost never stop your ads when its CAC is decreasing, and all of the rules above apply after the learning phase has finished and your CAC has dropped to its lower point.

The Problem of Hindsight

One of the biggest problems with the single ad analysis above is that we are viewing everything in hindsight after the ads have already finished delivering.

In reality, we never really know the precise trajectory of our ads when we are delivering it, and our CAC graph looks like something illustrated below.

In a single-ad situation, we can easily overcome this uncertainty by adopting a “wait and see” strategy and simply try it out by spending more on this ad.

Ultimately, we will be able to identify all of our most optimized spends if we follow the rule of thumb described in the previous section.

However, this “wait and see” strategy is only possible if we don’t have to allocate budgets across multiple ads - something we struggle constantly in practice as advertisers.

Therefore, we have to somehow assign a present value for the efficiency of the ads that happen in the future, and this will require the deployment of probability theory and stochastic processes.

As you would have imagined, modeling the future trajectory of your ads is not a simple mathematical feat, and requires additional information such as audience size, audience quality, past performance of similar ads, etc.

But in short, with the support of those maths, we should be able to assign an expected profit contribution for the ad above at different stages of spend, with us being less and less certain of the outcome further away we are from the current spend.

Keep this fact in mind, as this is essential for the multi-ad case that we are going to cover in the next section.

Looking for more content or discussions just like this? Join our Analytics for Humans Facebook community

Optimizing Ads In A Multi-ad Case

Everything we have discussed so far describes the behavior of one ad, channel, or your CAC performance in general.

However, in reality, we are not dealing with one of those curves, but rather multiple up to even hundreds of those curves operating simultaneously, competing on the same budget.

Your challenge now has become optimizing your advertising spend across different channels, adgroups, activities, with each at different stages of their lifecycle, and producing different CAC behaviors - where the fun actually begins.

Before talking about multi-ad optimization, let’s set a few assumptions of our analyses to simplify the world we are about to analyze:

We are going to assume that all ads and channels function independently of each other. That is, platform cannibalization and synergy does not happen.
We are going to assume all ads are placed for the exact same purpose and are comparable in terms of CAC performance (which is not true, as explained in a future article).

To conduct optimization mathematically, we will need to produce a trajectory projection for each of those adgroups or channels individually and compute the expected value for each trend, and then plug those trajectories into a profit optimization algorithm to produce the final result.

To do this without advanced math, you need to gather the spend and CAC of your adsets or campaigns at different “accumulated level” of spending and plot out their CAC as their accumulated spend increases.

Then, you can compare and contrast to obtain your top performers and continue your bets on them.

The problem of long-term, short-term tradeoffs

However, both the math and none-math approach above does not solve the following problem.

An ad will have inevitably high CAC at the beginning of its lifecycle, regardless of its future potential.

Therefore, if we focus on maximizing our profit or revenue for the present moment;

“ The most optimal strategy is to simply stop all ads at their early lifecycle, and go only for the ads that are in their later stages and can generate revenue/profit immediately. “

While this is a great short-term profit-boosting solution, it exhausts us of our resources in the long-term, and will ultimately produce a worse result for us.

In terms of data science or probability theory, this is called an exploration vs exploitation trade off - one of the most well-known challenges that have applications ranging from machine learning to life choices.

In our case, explore means running ads that might have high potentials in the future, but are at learning stages right now and does not produce high profit immediately.

On the other hand, exploiting means increasing the budget on ads that are already at their profit-generating stage, so we can reap their harvest.

There are people dedicating their entire lives analyzing creating solutions for this particular problem and by no means, I am one of the experts in the field, nor I want to dedicate much room in this article talking about this problem.

So to simplify, here are a couple of considerations to help you make decisions intuitively before making a choice on how much to explore vs exploit.

If you have ample cash flow and want to achieve long-term growth, always explore more than you exploit. Otherwise, exploit more than you explore.
If your organization is risk-averse, exploit will generally give you a much more stable outcome than exploring. On the other hand, if you are looking for higher rewards, exploring more is better.
If you have a very limited advertising budget and can only run an ad or two, exploiting only might be the best idea as you might simply not have enough budget for the exploration phase to complete for your learning ads.

Overall, the best choice for you is to set-aside an exploration budget for your company so when your executive team sees a jump in CAC, they do not panic and ask you to stop running those ads at their early stages.

Based on my experience seeing lots of ad account and talking with company owners over the years - you are exploring too little, go out and experiment on more early-stage ads.

A couple of practical concerns our analyses are missing

One keyword that I placed with all of the analyses so far is that it should be the case under a “theoretical” case, whereas the reality of ad optimization is even more complicated than what I have described above.

There are several big considerations that were not included above that you should be aware of here:

We are not directly considering the impact of audience saturation on your ads. If you are suddenly spending too much on one ad in a single day, your ad might not reach its supposed CAC potential as your deliverable audience of the day might be saturated.
We are also not considering the time of your delivery - spending $10,000 in one day will obviously be different than spending $1,000 for 10 days in a roll, but will remain similar to our model.
We are also not considering external factors such as competitor activities and seasonality, which also might impact our CAC significantly.

Wait, does this mean that everything I just told you above is crap? No, not really.

What I have shown you above is a very good “model” of how your ad functions as you spend more money on it, and as it changes over time.

Like any other model, it is a simplified description of a complex system that is almost impossible to explain clearly.

Furthermore, the elegance of the model above enables us to mathematically describe and model the performance of our ads with relative accuracy and efficiency, much beyond what we are able to do without it, and with CAC alone.

The non-practical elements of the model are strong considerations you should be aware of when using those models to plan and execute your ads, but modeling phenomenons such as audience saturation is a much more complex issue that is beyond even the scope of this article series.

Plus, there are two more topics that we haven’t covered yet in this article series that will also impact our CAC performance and optimization decisions - namely the different purposes of your ads, and the synergies among different platforms.

So let’s continue our discussion next time by exploring the impact of ad purpose and engagement depth on your ads.

In that article, I will provide you with a good framework to bucket and analyze your ads ad different stages and purposes, so you understand, with clearly, which channel is the best for which specific purpose.

Looking for more content or discussions just like this? Join our Analytics for Humans Facebook community to engage with like-minded people to discuss and share ideas on topics like this!

See you next time!