Is ROAS a standardized metric? Think again.

During my presentation “Measuring What Matters” at last year’s ANA Smarter Measurement, Smarter Marketing conference, I discussed why one of our industry’s most popular metrics—Return on Advertising Spend (ROAS)—often misleads more than it informs. My stance remains the same today: ROAS has fundamental issues in how it’s calculated, defined, and applied.

The Simple ROAS Calculation

To most people, ROAS is a standard metric with a fairly simple calculation:

Sales from the exposed audience during a specific window divided by the ad spending.

For example, if the sales from the exposed audience during the look-back window of 14 days is $100, and the ad spending is $40, then the ROAS is $100 ÷ $40 = 2.5-- for every dollar spent in the ad campaign, it "generated" $2.5 in sales.

That's not necessarily true. Here's why…

The Profit Margin Problem

First, ROAS often uses gross sales as the numerator, which means it is always positive (unless there's a massive recall/ return that exceeds sales). However, if we look into profit margin instead of gross sales, we will see a vastly different picture.

Using the previous example, if the profit margin is 30%, then the so-called ROAS is $30 ÷ $40 = 0.75-- for every dollar spent in the ad campaign, the "profit" is only 75 cents. If we deduct the ad spending $30 - $40 = -$10 , then the ROAS will be negative: -$10 ÷ $40 = -0.25. Basically, the profit from the sales is less than the ad spending. To break even, we need to increase sales by at least one-third-- not an easy task.

Enough math. If years of teaching taught me anything about the "audience", it would be that most people don't like to look at numbers and equations.

Does the Ad Really Generate Those Sales?

Going back to the definition: does the ad campaign really "generate" those sales?

If a campaign is targeting people who are most likely to buy, e.g., geo-targeting those customers who are in the proximity of the store, or even at the parking lot of a retailer, chances are the sales had little to do with the ad. If the data indicated that a customer is getting a coffee at Dunkin every morning, or buying pet food from Walmart every four weeks, surely you can serve an ad to "remind" them, but the purchase is likely to happen regardless. After all, most people are habitual by nature.

Simply targeting current customers won't grow your business significantly-- how many cups of coffee can one person drink on a daily basis?! To expand, you need to acquire new customers. This is why #INCREMENTALITY matters.

The Grading-Your-Own-Homework Problem

And then there's my favorite topic: the grading-your-own-homework reporting. When a customer is reached by messages on multiple social platforms, emails, different retail networks as well as linear TV, out-of-home display… Everyone can claim 100% of the credit. If we add all the amazing ROAS together, the sales can be so overly inflated that it may never be reconciled with your CFO's balance sheets.

Not to mention the inconsistency of definition and calculation of ROAS makes it impossible to compare different reporting, and can be very challenging when it comes to benchmarking, which requires some level of standardization.

Inconsistent Definitions and Benchmarks

I was once asked to look into a report with double-digit ROAS-- as much as the client preferred a stellar performance, the number was "unrealistic" therefore the client (rightfully so) rejected the report.

The client's reaction brought up an interesting question: In what range would the client just accept the report because "it looks ok"?

To understand, and hopefully manage, clients' expectations, I tried to find industry benchmarks for ROAS. What I saw was a wide range of numbers, varied by verticals, products, brands, and of course, platforms. Last time I checked, the Google AI Overview stated:

The Time Frame Problem

I did get into the rabbit hole of all the variations. The ultimate conclusion was: No matter what the number shows, we should never take it as is, simply because "it looks ok".

Another element of the ROAS measurement is the time frame: How do we define the ad's lasting impact on purchase. Other than using last-touch-last-click (LTLC) for digital conversion which is another topic all by itself, the inclusion of day-zero purchase can be iffy. Additionally, the arbitrary look-back window used for attributed sales could miss a great deal of long-term impacts while falsely claiming credits for short-term impacts-- No. It's not a wash.

According to a Nielsen study (commissioned by Google), the returns on long-term media investments (defined as 5-24 months) are the same as the returns of the "short-term" media investments (defined as 0-4 months). Yet most of the ROAS calculations I have seen used look-back windows anywhere between 7, 14, or 30 days, not 4 months, and definitely not 24 months. With such a narrow window, the ROAS would only capture the immediate returns that are most likely generated by those ready-to-buy shoppers instead of the net-new customers.

The Long-Term Brand Impact

While ROAS overstates the impact of performance-focused media which target the finite "low-hanging fruit", the limited look-back window ignores any impacts from long-term, cumulative branding messages which is crucial for business growth.

Measurement Without Action

Measurement without learning and actions is just another set of numbers; however, actions against the wrong data and metrics can cause real harm to your business. The inflated ROAS gives marketers a false sense of security and encourages increase in performance-based media for the immediate returns. The short-sighted focus is likely to overlook the impacts of long-term investment and under-sources the efforts that build brands. The relatively "instant" gratification often over-shadows the slow efforts in building brand equity since the latter is harder to measure. Besides, the increase in ROAS tells a pretty darn good story at QBR meetings.

Even with all these issues that prove ROAS is not a standardized metric, this industry won't suddenly eliminate it from all the reporting. Nevertheless, it will be wise to treat ROAS as another reference point, maybe pair with other measurement tools like Marketing Mix Modeling (MMM) to assess the incremental revenue by channel over time. Most importantly, avoid using it as the primary KPI for media allocation and channel optimization at the cost of sustainable business growth!