Should you cut the creatives that show zero ROAS on Meta?

It is the first reflex of many media buyers. It is also the one that can break what drives sales. To understand why, you have to look at two things at once: how Meta delivers your ads, and how Ads Manager reports on them. The two follow different logics. This article explains that gap, then what it changes in your day-to-day decisions.

TL;DR

A customer moves through several levels of awareness and sees many ads, on average, before buying. As I read it, Meta builds its delivery around that journey: it chains several creatives, several placements and several devices until it gets the conversion. Ads Manager, for its part, credits each conversion to the last ad clicked or viewed. A creative with zero ROAS may therefore have set up the sale that the “winning” creative closed. Cutting it breaks the sequence and dries up what feeds your conversions. As long as your campaign optimizes for sales, you steer on a single metric, spend: where Meta chooses to spend, it is right.

A customer’s journey: five levels, many touchpoints

Let’s start with buyer psychology. It is often summed up by the acquisition funnel (the funnel that runs from discovery to purchase), split into five levels of awareness. Each level matches a part of your target market.

Take the example of an agency media buyer who presents the week’s numbers to their client every Monday morning.

1. 1Unaware. The person has no awareness of the problem. For this media buyer, presenting the numbers to the client on Monday morning is no trouble at all.
2. 2Problem aware. They are aware of the problem. Every Monday, their Looker Studio has broken, half the images fail to load, the numbers are wrong, and they have to call the developer who maintains the dashboard.
3. 3Solution aware. They know several solutions exist: get better at Looker Studio, learn the presentation by heart, or replace the tool with other BI software.
4. 4Product aware. They know a specific tool is one of those solutions.
5. 5Most aware. They know its features, its benefits and its price. All that is left is to convert.

Moving someone from that first level all the way to a purchase takes many touchpoints. You find plenty of figures on the subject; the order of magnitude I keep in mind is 17 touchpoints on average. In practice, a prospect sees one of your ads on Meta, searches your name on Google, later types your URL into their browser, sees another ad on Meta, and finally buys.

What Meta does: it optimizes a sequence

Meta knows this journey. It built its delivery algorithm, Andromeda, with this funnel in mind: the algorithm decides which creative to show, to whom, and when.

It all comes down to two sentences. Meta optimizes by sequence. Ads Manager reports results on a last-click basis.

Let’s start with the first: Meta optimizes by sequence. To bring a person to conversion, it shows them a first creative, then another, then a third, and finishes with the one that triggers the purchase. In the opening example, it may have chained C1, then C14, then C3, and ended with C2.

Each creative then plays a different role. Some go after people who are still unaware of their problem. Others answer people who are comparing solutions. The last one speaks to someone who already knows your product and only needs a reason to act.

One nuance matters here. Meta does not publish the details of how it chains ads for a given person. What I call optimizing by sequence is a reading of how it works, consistent with the buying journey described above. What follows, however, is documented.

What Ads Manager shows: the last touchpoint

Ads Manager reports your results on a last-click basis: it credits each conversion to a single ad, the last one. As Jon Loomer, a well-known Meta advertising expert, explains, the conversion is credited to the ad clicked most recently within the attribution window, and only when there is no click, to the last ad viewed. The default window is 7 days after a click and 1 day after a view.

Back to the sequence. Fourteen creatives spent $10 each. The final conversion is worth $100. Ads Manager tells you C2 spent $10 and brought in $100, a ROAS of 10 (ROAS is attributed revenue divided by spend). The other thirteen spent $10 each and brought in zero.

That table is accurate by its own calculation rule. It simply says nothing about the role C1, C14 or C3 played before C2 closed. Ads Manager presents performance this way for practical reasons.

What you break when you cut the zero creatives

If you cut the thirteen creatives to bet everything on C2, you deny acquisition theory, the funnel described above, and the way Meta optimizes the delivery of your ads. You break the sequence.

C2 is excellent in one precise role: closing with a person who is already most aware and knows your product. For those people, it is a conversion machine. It is far less effective at reaching people at the top of the funnel, the ones who do not know you yet.

A soccer team works the same way. For the striker to score, the goalkeeper has to keep the ball out, the defender has to play it up to midfield, and the midfielder has to make the assist. Keeping only the striker because they are the only one who scores empties the field of everything that let them score.

In other words, the creatives that show zero in Ads Manager may be filling the top of your funnel. Cutting them shrinks the number of people C2 can then convert.

The same trap with placements and devices

This reasoning goes beyond creatives. Meta also optimizes the platform, the placement and the device by sequence. It can deliver a first creative on Instagram, in Stories, on mobile, then the next one on Facebook, in Feed, on desktop. Ads Manager still credits the conversion to the last touchpoint.

Take an example. You break your results down by device. Mobile spent $1,000 for a ROAS of 2.1. Desktop spent $1,000 for a ROAS of 4.3. The temptation is to cut mobile to feed desktop.

Odds are that mobile does the discovery work, at the top of the funnel, and desktop closes. We see this pattern a lot on e-commerce sites with a high average order value: people discover the offer on their phone, in bed at night, and switch to their computer when it is time to pay for a big purchase. To be precise: this behavior is less and less pronounced, but it still exists.

The rule is the same as for creatives: don’t cut an element at the top of the funnel, or you risk drying up what feeds the bottom.

Meta itself describes a related phenomenon in its documentation, which it calls the “breakdown effect”: when you break a campaign down by placement, ad set or ad, one element can show a higher cost per result even though the system uses it to get the best overall result. Its explanation is about how the budget is distributed. Its practical conclusion matches ours: a campaign is judged on its overall result.

Two conditions for letting Meta do its job

This approach assumes you trust the algorithm. It rests on two conditions.

First condition: you have given Meta a sales objective. The reasoning holds for a campaign optimized for conversion (a purchase, a lead). That objective is what gives meaning to the algorithm’s spending choices.

Second condition: you are convinced Meta is better than you on this ground. Two reasons make me confident. The first is the information it has: 52,000 data points on each user. The second is its economic interest. Meta charges by the impression and can only raise its CPM (the price of a thousand impressions) by making your ads more profitable, while keeping you around an ROI of 3, the global average. Your profitability and its own move in the same direction.

Steering by spend

When both conditions are met, the conclusion is counterintuitive: you steer from a single metric, spend.

When you have given Meta a sales objective and it decides to spend on a creative, a platform, a placement or a device, Meta is right: you let it, even if Ads Manager credits no conversion to that element. Spend is the visible trace of what the algorithm considers useful in the sequence. The ROAS of each element, read on its own, only tells you who scored the goal.

In practice, this changes how you read your reports:

• a creative that spends and shows zero is first of all a creative Meta considers useful;
• a placement or device with a lower ROAS is often the one that opens the journey;
• the result that counts is read at the campaign and ad set level, adding up every step of the sequence.

What this method does not settle: it tells you not to destroy a sequence that works. It does not tell you what to do when the overall result deteriorates, or how to spend more. Those are steering decisions in their own right, and they call for their own method.

So, cut or let it run?

Before your next cut, ask yourself three questions. Is my campaign really optimizing for sales? Am I judging this element on what it closes, or on what it brings to the whole? What would become of my best creative without the ones that come before it?

Ads Manager shows you who scored. Meta plays with the whole team.