Skip to content
Skayle Marketing

Meta Ads

Nobody was looking for you, so the creative has to do the work

Meta is demand creation. Audience settings have largely consolidated into automation, which leaves the creative as the main variable you still control — and that turns advertising performance into a production problem as much as a media one.

The premise

On Meta, the creative is the targeting. The system will go and find whoever responds to what you made, which means the thing you made is the instruction.

That is not a slogan about the power of storytelling. It is a description of how the auction now behaves. As audience controls consolidated into automation, the number of decisions an advertiser still makes shrank, and almost all of the remaining ones are creative decisions.

Which means the question stopped being who do we target and became what do we say, in what format, often enough to learn anything. That is a production question, and most accounts fail on it before they ever reach a media buying problem.

What changed underneath

Signal loss, modelled conversions and what to believe

Since the iOS tracking prompt, a meaningful share of conversions cannot be observed directly. Meta closes the gap by modelling: estimating outcomes it cannot see from the ones it can. That is a reasonable engineering response, and it also means the number in Ads Manager is partly an estimate rather than a count.

Two things follow. First, restoring legitimate signal is worth real effort — server-side event sending through the Conversions API, deduplicated against the browser pixel, with events configured and prioritised properly. Better input produces better delivery, not just better reporting.

Second, the channel needs a number it cannot mark its own homework against. We agree one before launch: total revenue or total qualified enquiries measured against total spend across the period. In-platform figures then do what they are good at, which is telling us which creative to make more of.

How we test

Three variables, tested in this order

Testing in the wrong order is how accounts spend a quarter learning that a green button performs much like a blue one.

  • The concept

    The argument the ad makes. A demonstration, an objection answered, a comparison, a founder explaining a decision, a customer describing the problem in their own words. Concepts are where the large differences live, and where fatigue actually happens.

  • The format

    How the concept is carried: static, short video, carousel, creator-shot footage, text-led. The same argument performs differently as a fifteen-second video and a single frame, and the placements it will run across should decide the shape before production starts.

  • The execution

    Hook, pacing, captions, thumbnail, the first three seconds, the offer wording. Real gains sit here, but only once a concept has proven it deserves the attention. This is the last variable we touch, not the first.

Scope

What running Meta with us involves

Roughly half of this is media buying. The other half is a production process, because without one there is nothing meaningful to buy media against.

  • Customer language research from reviews, support tickets, sales calls and comments
  • A concept backlog with a stated hypothesis behind each one
  • Scripting and direction for video, including creator and user-generated formats
  • Static and motion asset production, or direction of your in-house team
  • Campaign structure kept simple enough that results are readable
  • Pixel and Conversions API implementation, deduplicated and event-prioritised
  • Catalogue and feed setup where products are being advertised
  • Offer and landing page testing run alongside the creative tests
  • Weekly creative review: what is scaling, what has fatigued, what gets made next
  • Monthly reporting that shows platform-reported and platform-independent numbers side by side

The rhythm

How a creative cycle runs

  1. Mine the source material

    Reviews, objections, support tickets, sales call recordings and the exact phrases customers use. Ads that work are usually built from something a real person said, not from a positioning document.

    You get: Message bank and objection list

  2. Write concepts, not adverts

    Each concept is a distinct argument with a reason to believe it will work. We commit to a batch, not to a favourite, because the point is to learn which argument lands rather than to defend the one we liked in the meeting.

    You get: Concept backlog with hypotheses

  3. Produce in formats that fit the placements

    Vertical video for the placements that demand it, statics that survive being seen for one second, captions for sound-off viewing. Production is scheduled, so volume is planned rather than hoped for.

    You get: Batch of finished assets on an agreed cadence

  4. Launch clean and let it settle

    Simple structure, adequate budget per test, and enough patience to get past the learning period. Most accounts are damaged more by constant editing than by bad creative.

    You get: Live tests with defined read dates

  5. Read, retire, reinvest

    Scale what works, retire what has fatigued at the concept level, and feed what was learned back into the next batch. Then agree what changes: the creative, the offer, or the page.

    You get: Weekly decisions and an updated backlog

Questions

What advertisers ask about Meta

Did Facebook ads stop working after the iOS privacy changes?

They became harder to measure rather than less effective. When people opt out of tracking, some conversions cannot be observed directly, so Meta reports a mixture of observed and modelled results and the numbers in Ads Manager stop matching the numbers in your shop.

The practical response is to restore as much signal as you legitimately can through server-side event sending, configure your events properly, and hold the channel accountable to a platform-independent number as well as the reported one.

How much creative do we actually need?

More than most teams expect, and the useful unit is concepts rather than files. Five genuinely different arguments for why someone should care are worth far more than fifty crops of the same photograph.

We usually plan around a small number of new concepts per month, each produced in the formats the placements require. That is a production schedule, so we agree it against what your team or ours can realistically deliver before anything is promised.

Should we still build detailed interest audiences?

Far less than you used to. Meta now expands beyond your selections by default in most campaign types, so tightly built interest stacks often end up describing a starting point rather than a boundary.

What still earns its place is exclusion and constraint: customer lists, existing buyers, age and location limits, and anything defined by your actual business rules. The rest of the steering now happens through what you put in the ad.

Why do Ads Manager and our ecommerce platform disagree?

Because they count different things over different windows. Meta attributes a conversion to an ad within its own attribution setting and includes modelled results where direct observation is not possible. Your shop counts orders when they happen, regardless of what someone saw.

Neither is lying. We reconcile them by agreeing one platform-independent measure up front, usually total revenue or total qualified enquiries against total spend, and treating the in-platform numbers as directional signals for deciding what to test next.

Can you guarantee a return on ad spend?

No. Return depends on your margins, your offer, your pricing against competitors, your site and the auction, and only some of those are ours to influence.

We will tell you before we start what we think is plausible given your economics, and we will tell you quickly if the channel is not going to reach it. That is more useful to you than a figure nobody can stand behind.

Find out whether your problem is the media or the creative

Send us the account and the last six months of creative. We will tell you whether performance is limited by how the budget is being spent, by how few genuinely different ideas have been tested, or by the offer itself.

Last updated · Reviewed by Zubair Afzal

We use analytics to understand which pages are useful. Nothing runs until you choose, and we do not sell or share what we collect. What we would set.