Creative is the constraint in Q4, not budget

Every year the same thing happens in the first half of November. Spend goes up, results go sideways, and somebody asks what changed. Almost nothing changed. That is exactly the problem.

Let us concede the well-covered part first, because the agency blogs have been saying it for two years and they are right. Creative is now the main performance lever on Meta. DataSauce calls it the biggest lever there is. AdVisible says budget and targeting tweaks cannot fix tired creative. Nest Commerce puts it most usefully: you cannot out-target the algorithm, but you can out-supply it. MTA Digital argues creative strategy is no longer an add-on to performance but its foundation. We have made the same argument ourselves for ecommerce accounts.

So the what is settled. The question worth asking is when.

Because the advice that follows the consensus is almost always "produce more creative", and the businesses that lose Q4 usually did produce more creative. They produced it in November. By then the expensive part of the quarter is already running, and a new asset is the worst-positioned thing in the account.

Why more budget makes a tired account worse

The mechanics are not subtle. Frequency climbs, click-through rate slides, CPMs rise with no improvement in conversions, and cost per acquisition drifts up on identical spend (DataSauce). At that point the instinct is to move money, and every direction is wrong. Lowering spend buys fewer impressions rather than better ones, raising it accelerates the fatigue, and pausing and restarting the ad resets nothing because the platform remembers (AdVisible).

Privacy changes made this sharper. Smaller retargeting pools and thinner behavioural data mean the creative carries more of the load, and the same people see the same ad more often (AdVisible). Q4 then adds competitors bidding harder into the same feed.

None of that is a budget problem. Budget in a fatigued account just buys more impressions of an asset the audience has already decided to ignore. It is worth remembering how much of the outcome sits in the asset itself: creative accounts for 49% of the total sales impact of advertising in research Google cites from NCS (Google).

One of our own Q4 results says the same thing from the other direction. A US retailer lifted ROAS 14% on essentially flat spend across a Q4 on Meta (Dadek Digital). The money did not move. What the account was working with did.

What actually happens before the auction

Most advertisers picture one contest: your ad against everyone else's, decided by bid and predicted performance. There are two. Before the auction there is retrieval, and retrieval is the stage that decides who gets to compete at all.

Meta describes it plainly. Retrieval is the first step in its ads recommendation system, tasked with selecting from tens of millions of ad candidates down to a few thousand, and it processes three orders of magnitude more ads than the stages after it (Meta). Andromeda, the system Meta built for that job with NVIDIA hardware, delivered a 6% recall improvement to retrieval and an 8% ads quality improvement on selected segments (Meta). Where the old system chose from thousands of ads per auction, this one chooses from tens of millions (Nest Commerce).

That is the shift everyone summarises as "creative is the new targeting". Fine. But it has a consequence for timing that the summaries skip.

The part almost nobody mentions: new creative has no history

A brand new asset arrives with nothing. No accumulated performance, no learned expectation, no reason for the system to trust it.

That matters more than it sounds. Older creatives tend to dominate delivery precisely because they have already accumulated historical data, and genuinely better new creatives can barely receive impressions because the system has not learned what to expect from them yet (pValues). Meanwhile the platform needs stability to learn, and changing several things at once resets that learning (Nest Commerce).

Put those two facts next to a November upload date and the picture is unpleasant. You are asking an asset with no track record to win a place in the candidate set against your own proven ads, in the most expensive auction environment of the year, while the account is least able to absorb a learning reset. The creative might be your best work. It will not get the impressions to prove it.

This is the real deadline, and it is not "have the files by November". It is have the assets live early enough that they have already earned delivery before the peak. In practice that means the shoot happens in September, the assets go live across late September and early October, and by the time the expensive weeks arrive you are choosing between proven options instead of introducing unproven ones.

It also means last year's Q4 winners are not the answer. Recycled creative declines faster than fresh creative, with click-through rates dropping quicker and CPMs rising faster, partly because audiences carry the earlier fatigue and the platform still holds the old performance record (DataSauce).

What businesses overestimate: volume

Here is where we part company with some of the consensus.

The published benchmarks have become enormous. Competitive brands are described as running 30 to 50 assets a month as a minimum, leaders in the hundreds, with the trajectory pointed at 500 (Nest Commerce). That advice is not wrong, but read who it is written for. Nest states its own client floor as £75,000 a month in paid media (Nest Commerce). Most Australian businesses planning Q4 are nowhere near that, and chasing the number half-heartedly is worse than ignoring it.

Worse, because volume without difference does nothing. Same image with a different headline, same concept with a different background, same hook with a different thumbnail: none of that is meaningful variation, and the test is simply whether these ads would appeal to different people (Nest Commerce). Two images can be different files and still look similar enough that the variation makes no difference to performance (Jon Loomer Digital). A single creative is not a test at all (MTA Digital).

Volume is also the cheapest thing in the system now. More than a million advertisers used Meta's generative tools to make over 15 million ads in a single month, and Meta estimates businesses using image generation see a 7% lift in conversions (Meta). If anyone can generate variations, variations are not an advantage. Distinct ideas are.

So the Q4 question is not how many files you have. It is how many genuinely different ideas you have: different hooks, different formats, different angles, different people in mind. Count those. Most accounts hold more files than ideas, and quite often the honest answer is one.

And you do not always need more ads to get there. Flexible Format and Dynamic Creative take up to ten images or videos inside one ad, alongside five primary text and five headline variations (Jon Loomer Digital).

What businesses underestimate: whether the loop can actually close

The standard recommendation is a closed loop. Performance data flows into the creative brief, production is informed by what is working now, results feed the next cycle (Nest Commerce). MTA frames the same thing as a repeatable cycle of research, brief, test, insight and iteration (MTA Digital).

We agree with the model. We just think almost nobody can run it, because the loop is open at the join.

In most Google Ads setups you cannot reliably connect the specific creative a person saw to what that person did afterwards, which makes post-click engagement quality, retention by creative cohort and lifetime value differences close to unmeasurable (pValues). It gets less comfortable. In a Performance Max or App campaign, 90% of the performance you are attributing to an asset may be coming from Search placements entirely (pValues). And creative lifecycles are usually too short and too sparse to model against, so the patterns teams believe they have found are often auction dynamics and delivery bias rather than creative resonance (pValues).

That is the gap worth closing before the quarter starts. If you cannot join the asset to the revenue, your November decisions get made on platform-reported numbers, during the one month of the year when being wrong is most expensive. Which is the same argument we made about tracking readiness earlier this month, for the same reason.

On Google, the constraint is a different shape

Creative means something narrower on Google, and it is worth separating the two.

The surface is genuinely restricted: 30 characters per headline and 90 per description on responsive search ads, no control over fonts or colours, and text overlay on images can be limited (Nimbl). Nobody wins Google on art direction. You win it on coverage.

Google's own numbers are specific about what coverage is worth. Lifting Ad Strength from Poor to Excellent is associated with 12% more clicks and conversions, image assets on search ads with a 6% average click-through lift, showing a business logo and name with 8% more conversions at similar cost, and including all three video orientations in Performance Max with 20% more YouTube conversions than horizontal alone (Google). Armani Beauty added responsive search ads at Good or Excellent strength across its unbranded ad groups and saw a 61% increase in click-through rate and 11% more on-site conversions (Google).

Google will also tell you what is missing. Creative guidance flags video ads without a prominent logo in the first five seconds, without a human voice-over, outside the recommended length for the objective, or missing any of the three aspect ratios (Google). Running that check is twenty minutes of work and it is the highest-return creative task in most Google accounts before Q4.

What to do in the next fortnight

Open the ad account and count ideas, not assets. If everything traces back to one hook, you have one creative, however many files sit in the library. Then work out what is missing: which formats you have never run, which objections you have never answered on camera, which customer you have never spoken to directly.

Book the production for September, not November, and get the assets live in early October so they have accumulated delivery history before the auction gets expensive. While that runs, fix the join between creative and revenue, so the decisions you make in the peak weeks are made on your numbers rather than the platform's. Then leave the working things alone, which is the discipline we wrote about at the start of the month.

Dadek Digital builds the creative testing structure and the measurement layer underneath it, so you can see which asset earned the money rather than which asset the platform credited. If you would like that in place before the quarter turns, a Q4 readiness review is the fastest way to find out what is missing.