Note: ROAS = return on ad spend. To calculate it, divide the dollars of revenue attributed to an ad campaign by that campaign’s cost.
We have officially entered Q4, the period where ROAS is highest for many brands. But the long term benefits of holiday sales are questionable at best. So I thought I would use this edition of the newsletter to explore how ROAS can be used and abused, and what metrics we should be using in its place.
ROAS is not completely useless. It’s good for one thing: getting a read on the performance of a specific advertising campaign. Unfortunately, ROAS is also frequently used as the north star for budget planning and eCommerce strategy. Those are not jobs ROAS is equipped to do well.
Let’s use an analogy. Your business is an ecosystem, kind of like your own body. And like any ecosystem, your physical health is affected by multiple factors that occur both inside and outside your body.
ROAS is the equivalent of a simple data reading, like taking your temperature or checking your blood pressure. A reading like this can provide valuable information about a specific body system, within the context of a specific time period.
For example, you could use temperature readings over the course a week to determine your progress in fighting the flu. Or you could use monthly blood pressure readings to evaluate the effectiveness of attempts to improve cardiovascular health through diet and exercise.
But if you used a week’s worth of temperature or blood pressure readings to make major decisions about your health, you would probably wind up with some pretty bad outcomes. Like, you wouldn’t amputate a leg or decide to go raw vegan based on a few readings. But in the eCommerce world, we often do just that.
By the way…if you appreciate a more strategic lens on digital marketing, you’ll love the No Best Practices newsletter.
There are two reasons why ROAS is not up to the task of guiding your digital marketing efforts:
1. Your daily sales come from one of two places: new or returning customers. Sales volume from returning customers is fairly predictable and has a defined ceiling. You need to make up the distance between that ceiling and your sales target with acquisition.
ROAS does nothing to help you understand your progress with either acquisition or retention. All else equal, your most loyal customers have the highest likelihood of converting again. They also make up the smallest percentage of your customer base. ROAS-first thinking will bias your efforts towards these customers–a small pool of rapidly diminishing returns.
2. Paying money to run an advertisement is only worthwhile if it changes the viewer’s behavior. Paying to “convince” someone to do something they were already planning on doing is a waste of money. This is why the concept of incrementality is so critical to marketing. Using ROAS as your primary KPI actually discourages you from winning incremental sales.
Let’s say I run a shop that fixes flat tires. If I pay someone to twirl a sign outside as motorists with busted tires roll up, is that activity “convincing” anyone to do anything? No! But the view-to-conversion rate will be close to 100%, and the ROAS numbers would be amazing. Hold your team or agency accountable to a ROAS target alone, and you’ll get this kind of “effective” marketing strategy.
Most eCommerce businesses have seasonal peaks and valleys in their revenue numbers. The peaks are often driven by macro consumer behavior–back to school season and the holiday shopping season are two examples. But just as often, the peaks are driven by promotional events and sales.
Your ROAS will almost always be higher when you’re on sale. And the deeper the sale, the higher the ROAS. Why? When your brand is on sale:
Conventional budgeting strategies allocate the annual digital marketing budget in proportion to annual sales. So if 20% of annual sales happen in March, you should also spend 20% of the budget during March.
But what if it turns out that March is your brand’s annual clearance sale? Your budgeting situation is going to create a flywheel that makes it harder to sell at full price as time goes on:
Some marketers claim that clearance or markdown customers can be turned into full price customers over time. I outline why that isn’t true here. If you run your business like this for long enough, you will destroy it.
During the annual financial planning process, the performance marketing budget for the year will be set at some percentage of the year’s forecasted total sales. The specific percentage depends on a lot of factors specific to your stage of growth and what you’re selling.
Let’s say you are trying to achieve $10M in sales next year.
In the middle example there is an underlying assumption that every dollar you invest in digital marketing will result in five dollars of sales. Of course, that’s almost impossible. Nothing exists that drives that type of return consistently and at scale.
But you don’t need digital marketing to cover the entire $10M goal. some of your returning customers will come back without paid advertising. And you’ll acquire some new customers through word of mouth or other organic means.
But leadership will still insist on a requirement like “all digital marketing must have a blended ROAS of 5”. Where are you likely to get a ROAS of 5? Running ads when you’re on sale and/or targeting audiences who already have high purchase intent. Again–do this for long enough and you’ll wind up with a business that doesn’t grow, and that can’t sell at full price.
As marketers, the entire purpose of our jobs is to influence human behavior. And generally, if we want to change consumer behavior–tear people from the grasp of inertia and get them to pay attention to us–we need to change our own behavior.
ROAS does not help us evaluate if behavior has been changed. In fact, it often rewards activities that produce no net change. Here are some alternative metrics we can start using instead.
A/B holdout testing is the gold standard for determining if marketing activities are actually influencing consumer behavior.
In a holdout test you define a target audience, expose half of them to an advertising channel or campaign, and prevent the other half from seeing the same campaign.
If the test audience converts at a higher rate, and that lift is statistically significant, you can say with some certainty that the test treatment is causing the gains.
In a perfect world, you would be holdout testing all of your digital marketing activities in this way. eBay did just this for their entire branded search spend, and found it was mostly useless.
Think about how much budget you’re investing in branded search right now. I’m not saying “pull the plug”, but wouldn’t it be worth a second look?
I also acknowledge that the world is not perfect. But at the very least, you should be holdout testing any new vendor or tactic when it’s technically possible.
A rule of thumb: the closer a channel or tactic is to the bottom of the funnel, the more vulnerable it is to ROAS abuse.
So the more aggressive you should be about holdout testing it.
CAC = cost to acquire a customer. This is all the money spent on acquisition marketing during a period divided by the number of customers acquired in the same period.
You can compare this to your average order value and average contribution margin per order over the same period to gauge profitability in real-ish time.
MER = media efficiency ratio. This is all your sales in a period divided by your total marketing spend for the same period. You can use this metric to understand if, in general, your total marketing budget is becoming more or less effective over time.
These two metrics will help you manage the high-level profitability and effectiveness of your marketing spend.
You do need a metric or two that will provide some guidance on same-day campaign performance. You want to understand if new campaigns or creatives are meeting expectations. You also want to be aware of wild swings in performance so you can investigate them.
Sometimes, ROAS is the right tool for this job. Other times, traffic quality metrics or in-platform attribution from the advertiser will do the trick.
But you need to treat these metrics as what they are: campaign level temperature taking.
You should not be running your business with them.
Supercharge your growth with a purpose-built ecomOS for brands and agencies.