Walk into any modern creative agency and watch the standup. It's a stand-up about volume. Six variants for Meta. Three for Google. Two for the landing page. A handful for the partner pitch. Everyone nods. Nobody mentions that the conversation is about output units, not about ideas.
The industry named this content at some point in the last decade. It was always manufacturing. Manufacturing is a noble discipline — we just have to admit we're doing it before we can decide what we're doing it for.
The factory worked for a specific decade
From roughly 2013 to 2022, the agency content factory worked. Three inputs held steady and reinforced each other:
- Labor was abundant and junior rates covered craft work. A strong copywriter or designer in their mid-20s delivered senior output at a salary that kept the P&L margin intact. Talent supply outpaced the number of agencies bidding for it.
- Platform algorithms rewarded throughput. Meta, TikTok and YouTube optimized for creative refresh rate. More variants — even mediocre ones — meant more impressions at lower CPMs. The factory's output curve matched the platform's demand curve.
- Attribution made volume look causal. Multi-touch attribution assigned credit to whatever variant ran last, and the factory always ran something last. The math told a story that justified the process.
All three quietly broke between 2023 and 2025.
What actually broke
Labor re-priced. Senior creatives are paid senior rates now. Mid-career talent that used to staff junior roles moved in-house or left the category. Hiring timelines doubled. The margin that used to absorb 40% rework evaporated.
Algorithms stopped rewarding raw volume. Meta's systems now privilege signal quality over asset count — a point the platform never says out loud, but every performance marketer has felt. A hundred mediocre variants now underperform five good ones in ways that would have been impossible to measure in 2019.
Incrementality tools arrived. Conversion lift studies, holdouts, and MMM caught up. The story attribution told is now routinely tested against reality, and the factory's throughput looks less causal under the test than it did under the model.
The post-mortem nobody wants to run
Here is the honest conversation an agency leadership team would have, if they were willing to have it in the open:
“Our throughput is no longer what the client is paying us for. They're paying us for judgement — what to test, what to kill, what not to build — and our operating model is still optimized for asset count. The gap between what we sell and what we deliver is where the margin disappears.”
Most agencies don't have this conversation because the answer ends in painful scope re-negotiations with every retainer client. But the conversation is inevitable. The only variable is whether it happens in a boardroom or in a quiet resignation by the senior strategist who has been carrying it alone.
What “naming the factory” actually means
Naming the factory is not an admission of failure. It's a precondition for building something better. Once an agency accepts that the factory is the base layer, three things become possible that weren't before:
- You can measure the right unit. Cost per variant at shipping quality. Cycle time from brief to approved. Rework rate. These are factory metrics and they're the ones that actually drive agency P&L. The old unit — “hours billed” — is a proxy that stopped meaning anything.
- You can separate the two products. The factory makes variants. The strategist makes decisions. Those are different products, sold to different parts of the client's org, with different economics. Most agencies bill them as one line item and lose margin on both.
- You can automate what is actually automatable. Most of the work the factory does is genuinely repetitive — the platform-sizing, the format variants, the localization, the brand compliance checks. This is where AI sits well. Not as a shortcut to creative; as a shortcut to the parts of creative no one should ever have been doing by hand.
What this looks like in practice
A healthy modern agency runs two distinct tracks. The factory track — high throughput, high governance, mostly automated, measured on cycle time and unit cost. And the judgement track — senior operators deciding what to test, what to kill, which insight is worth a week of investigation. The factory track funds the judgement track. The judgement track is what the client actually hires you for.
This is a harder agency to run, because it requires leadership to be honest about which track each project belongs to. It's also a more defensible one, because the judgement track is where the category is still unautomated.
What happens if you don't name it
The agencies that refuse to name the factory are following a predictable trajectory: retainer margin compresses, headcount is cut from the judgement track (because factory headcount feels more load-bearing), the factory output stops being distinguishable from any other factory's, and the agency becomes a line item in a procurement review. At which point a client replaces it with three factories in different regions and keeps its own judgement track in-house.
The way out is not to run the factory faster. It's to be honest that the factory is what you're running — and to price the judgement track separately, openly, with different outcomes on the table.