What it takes to run a creative system

The question a brand team asks once it understands the loop is whether it can run the loop itself, and the answer is yes, under conditions. A creative system is Strategy, Production and Optimisation run as one weekly cycle on a connected ad account. What makes it a system rather than a project is that every part happens every week, at a volume the arithmetic sets, to a standard everyone in the room shares, by people whose job it is.

The arithmetic of variety

Start with the number of ads. The delivery system matches each ad to the people who respond to it, and Meta's own guidance describes creative diversification as "the best lever to find the most relevant audiences", so a portfolio needs enough distinct ads for the matching to work, in our experience twenty to thirty live at any time, each built for a group of category buyers and a category entry point in Jenni Romaniuk's sense, the situations, occasions and feelings that connect a buyer to the category.

Distinct is the word doing the work. Cutdowns and aspect ratios do not count, because ads that open alike and earn the same response from the same people are consolidated in delivery and the rest sit live and starved. So twenty new ads a month means twenty situations, twenty casts, twenty locations and twenty openings, real people in real environments, each carrying the distinctive assets from the first second, so that consistency lives in the brand and variation in the execution. A studio with three presenters and one kitchen can make twenty variants of one ad in a month. It cannot make twenty ads. The production that meets the arithmetic is distributed, creators and talent across many lives and locations working to one asset standard, and it is standing capacity, because the loop asks for new work every week.

The weekly rhythm

Each week the account's data is pulled per ad, per platform, stop rate, hold rate and watched rate, cost per watched impression, served frequency against watched frequency. The read produces three lists. Ads to pause, where attention never arrived or has decayed, decided on the watched rate against the brand's own baseline and never on the calendar, an ad is retired because it was not watched, not because it is old. Ads to remix, where the hypothesis is alive but the execution is leaving value unclaimed, which turns evidence into improvement without a new shoot. Concepts to replace, the net-new hypotheses the portfolio needs, briefed from who stopped and who stayed.

The briefs go out the same week, production turns them inside the cycle, every new ad passes the gate and is live before the next read, so the loop closes fifty-two times a year rather than four. Each decision is recorded against the hypothesis it tested, which is how an account builds a memory instead of a history, and the media team receives the creative signals, which ad is tiring, which buyers are responding, in the same meeting it reviews pacing and frequency. A quarterly review cannot be made weekly by working harder inside it, the cadence is set by the structure.

The pretest standard

Every ad passes a gate before media spend attaches, fast enough to run on every ad every week or it becomes a bottleneck and gets waved through. It checks the checkable. Brand impact, the distinctive assets present early, the codes used correctly, the product as the answer to the situation. Creative quality, the opening's impact, native feel, legibility with the sound off, safe zones, copy accuracy. Safety and compliance, platform policy, brand mandatories, claims and legal. Each ad is scored with weights the brand sets once and comes back with action items rather than a verdict. What the gate cannot do is predict attention, a panel paid to watch tells you what an ad does with attention it has been given, not whether it earns any in a feed, so that question goes to the feed, which answers it within days. What pretesting can tell you, and what only the feed can draws the line.

The measurement standard

A loop runs on numbers, so everyone in the room has to mean the same thing by them, and the standard is published in How to read paid social results. Results are read in one order, effectiveness first, did the brand move, from lift studies on their own clock, then performance, did the ads earn attention, weekly, then efficiency, what it cost. Stop rate is three-second plays over impressions on Meta, hold rate is ThruPlays over three-second plays, watched rate is ThruPlays over impressions on Meta and focused views over impressions on TikTok, never blended, one platform and one threshold. CPM is media efficiency and carries its objective beside it, and all-in cost per watched impression, media plus fee, against the prior creative on media alone, is marketing efficiency. The objective stays set for the brand's goal, reach for brand building, while the creative is optimised for attention. We read attention, we do not buy it.

The people and the hours

Five functions have to exist, however the headcount is arranged. A portfolio owner, who holds the map of buyers and entry points, decides which hypotheses the account backs and writes the briefs from the read. A producer, who runs the distributed talent, casting, locations and turnaround. An analyst, who pulls the data per ad, runs the read in order and brings the three lists to the meeting. A brand approver, who sets the gate's standards once and owns the exceptions. And a media counterpart, who receives the creative signals and owns the settings, because no media buying happens inside the creative function.

The hours are the honest part. The read and the decision are a standing weekly commitment, because a skipped week is a week spent on stale information. Production is continuous, not a burst before a launch. The gate runs on every ad, which at twenty to thirty a month is a daily task. The record of what was tested and learned has to be kept by someone and outlive them. None of this is a side-of-desk job, and a committee or a quarterly meeting fails the cadence test before it starts.

Building it in-house

The counter-position is a reasonable plan and deserves stating as one. A brand team with a content studio, a creator programme and an analyst in the media team proposes to run the loop itself, keep the knowledge in the building and avoid a fee. Nothing here argues against that. The conditions are the ones above, standing production across enough different people and places to make twenty distinct ads a month rather than twenty variants, a named owner with the authority to pause work that is spending, a gate fast enough for every ad, a measurement standard the media and brand teams both use, and the patience to treat the first ninety days as calibration. A team that holds all five every week has built a creative system, whatever it calls it. A team that holds three has found where the other two sit, whether it hires for them or brings in a system to hold them. The alternative to building it is running the creative half of the account as a deliverable, which is where most of the unlock in paid social still lies.

What to do with this

Write the five functions down the side of a page and put a name against each for your own account, then write the four standing commitments beside them, the weekly read, continuous production, the gate on every ad, the record of hypotheses, and mark which happen today. Then run the arithmetic on last month, how many distinct ads went live, distinct by person, place, situation and opening, and how many were retired and why. The gaps are the specification for what to build or buy, and they are the unlock, not a failing. Media creates the opportunity to see your ads. Creative determines whether they work. Whoever manages the second half of that sentence, How to audit your own paid social account turns this page into the numbers to pull.

Questions

Can we build a creative system in-house?

Yes, under conditions. Standing production capacity across enough people and places to make twenty to thirty distinct ads a month rather than variants of a few, a named owner with the authority to pause an ad that is spending, a pretest gate fast enough to run on every ad, a measurement standard the brand and media teams both use, and ninety days of calibration before anyone reads a verdict.

Teams that hold all five every week are running a creative system whatever they call it. Teams that hold some of them have located the rest, which is the useful outcome either way.

How many people does it take?

Five functions, and the headcount depends on the size of the media plan and the portfolio it needs. The functions are a portfolio owner, a producer running distributed talent, an analyst who does the weekly read, a brand approver who sets the gate's standards once, and a media counterpart who keeps the settings.

What does not vary with scale is that the read, the decision and the gate are weekly standing commitments rather than something done when there is time, so the roles need hours protected for them, and the production has to be capacity that exists every week rather than a shoot booked for a launch.

How long before we know whether it is working?

Performance signals arrive within days per ad, stop rate, hold rate and watched rate against the brand's own prior creative, and they are read weekly from the first week. The portfolio itself takes about ninety days to calibrate, the first month casts the net, the second starts retiring and extending, and by the third the surviving ads are concentrated where the market responded and the briefs are written from evidence.

The brand outcome, lift in mental availability, association or intent, is read on its own clock from lift studies against a holdout, over quarters, and a watched rate is never a proxy for it. Effectiveness is reported when it was measured, and when it was not, we say what was measured and stop.

Sources

• Meta for Business, "Demystifying Creative Diversification", April 2025, as quoted by Jon Loomer, 2025, and Engineering at Meta, "Meta Andromeda", 2 December 2024. Creative diversification as the best lever to find relevant audiences, ad diversity and advertiser outcomes. Platform wording, Meta sells the media.
• Meta Business Help Centre, video metrics. Definitions of 3-second video plays and ThruPlay (completion or 15 seconds, whichever comes first). Platform documentation, Meta sells the media.
• TikTok Ads Manager Help Centre, video views objective. Definition of a focused view (at least six seconds, completion, or a positive interaction within the first six seconds). Platform documentation, TikTok sells the media.
• Jenni Romaniuk, Building Distinctive Brand Assets, 2018, and Better Brand Health, 2023. Category entry points, consistent assets with varied executions. Ehrenberg-Bass Institute, sponsored by large advertisers.
• Our own pre-flight gate, described generically. Brand impact, creative quality, and safety and compliance checks with weighted scores and action items, applied before media spend. Our own platform, our interest declared.
• How to read paid social results, the measurement standard, with its own sources footer. Our own standard, platform definitions checked September 2026.