Facebook Ads Agency or Platform · Honest Decision Framework

Every marketing leader running a Meta Ads budget over $30K/month eventually faces the same decision: hire a Facebook ads agency, or move the operation onto a platform. The pitch decks on both sides are polished. The honest tradeoffs are almost never on the table during the conversation. By the time the wrong choice surfaces — usually 6 to 9 months in — the switching cost is high.

This is the framework we use when a brand asks us, off the record, which path to take. It is not a sales document. The right answer depends on five questions about the operation. Answer them honestly and the choice becomes obvious.

The agency-vs-platform tension is not new · the decision math is

Five years ago, the choice was structural and stable. Agencies brought senior craft and relationships with Meta's rep team. Platforms brought workflow tooling and reporting. A brand with under $20K/month in Meta spend ran the operation in-house; a brand over that threshold hired an agency. The line was clear.

The line moved. Two shifts pushed it. First, Meta's delivery system became aggressively automated — the senior craft that agencies sold in 2019 (manual bid management, manual audience targeting, manual placement optimization) became less differentiating because Meta does it natively now. Second, AI-native platforms emerged that produce the creative and operational work agencies used to provide, with a structural cost advantage. The shape of that cost advantage is the math we laid out in AI Ads · The System Meta Actually Rewards.

The result is a market where agencies and platforms compete for the same operational scope, often without the buyer realizing they are alternatives rather than complements. The five questions below cut through that.

Question 1 · How many distinct creative concepts do you ship per month?

Not how many variants. How many distinct concepts — meaning, how many genuinely different creative ideas, each of which can spawn variants.

• 0-2 concepts/month. An agency is the better fit. The volume is too low to justify platform overhead. The work is craft-heavy and benefits from senior creative attention.
• 3-8 concepts/month. Either model works, but the cost crossover starts to favor a platform around the upper end. An agency at this volume will charge $25K-$60K/month depending on geography; a platform delivers comparable output at $5K-$15K/month plus media.
• 9+ concepts/month. A platform is the better fit. The volume exceeds what most agencies can absorb without scaling the team, and the variant production within concepts is exactly the layer AI-native platforms handle better than human teams.

The honest follow-up: most brands overestimate how many distinct concepts they actually run. Pull last quarter's campaign list and count the genuinely different ideas, not the rerun-with-new-colors variants. The honest number tends to surprise the marketing lead. The reason this happens at all is the variant-vs-concept confusion we unpack in Ad Creative AI · The Uncomfortable Truth About Variant Generation.

Question 2 · How much of the work needs senior creative judgment?

Senior creative judgment is the work that requires a strategist or creative director to decide what to do, not how to execute. Examples: positioning a new product, breaking into a new audience, recovering from a brand crisis, launching in a new geography. Cannot be automated. Should not be automated.

Junior creative judgment is the work that requires craft but operates within an established positioning. Examples: producing this quarter's seasonal creative, refreshing underperforming ads, variant generation against an approved angle. Can be automated. Often should be.

• If most of your monthly work is senior judgment, you need an agency relationship or a senior internal team. Platforms execute below the senior layer and will frustrate you on this work.
• If most of your monthly work is junior judgment in service of an established positioning, you are paying agency rates for work a platform does at platform rates. The math doesn't hold.

Most established brands underestimate how much of their work is junior judgment. The positioning was decided years ago. The audience is mature. The angles are well-tested. What the brand needs in this state is production capacity at sane unit cost, not senior craft.

Question 3 · How fast does your creative need to refresh?

Meta's delivery system increasingly rewards creative refresh cadence. Audiences fatigue on a creative within 14-30 days at moderate spend. A platform that can produce a new orthogonal variant set on a one- or two-week cycle outperforms an agency that produces a new set on a four- or six-week cycle, at the same media spend.

The honest test: pull your last 90 days of Meta ad creative. Sort by launch date. Look at the actual gap between refreshes. That is your refresh cadence, regardless of what the agency contract says about turnaround. The week-by-week view of an operation that hits the weekly-refresh cadence is documented in One week inside a Hi Luca agency.

Question 4 · Who else needs to operate on the same creative system?

Modern marketing operations rarely involve just one party. The realistic operating reality is:

• The brand-owner team (CMO, brand managers, marketing leads).
• The internal performance marketing team.
• One or two external agencies in different specialties.
• Freelancers brought in for specific projects.
• Sometimes a regional distributor or franchise group.

All of these parties need to operate against the same brand memory, the same approved positioning, the same creative library. An agency engagement gives one party (the agency) access to the work; everyone else interacts with the agency. A platform engagement potentially lets all parties operate on the same surface with role-appropriate permissions.

• If your operation is one-to-one (brand and one agency), the agency model works cleanly.
• If your operation involves three or more parties, an agency-only model becomes a coordination problem the agency cannot solve. A platform that supports multi-role workflow becomes the better structural fit.

We described why the system-vs-tool difference is sharper in multi-party operations in AI Marketing Tools vs AI Marketing Systems.

Question 5 · What is the realistic 24-month roadmap?

Marketing operations evolve. The right model for the next 90 days might not be the right model for the next 24 months. Three forward-looking signals matter:

• Are you scaling the operation 2-5x in the next 24 months? Agency model scales linearly with cost; platform model scales with marginal cost approaching zero. Above 2x scaling, the platform math compounds.
• Are you planning to bring more of the operation in-house? If yes, a platform creates the operating layer that the in-house team will inherit. An agency relationship doesn't transfer well to an internal team.
• Are you adding new channels (Google Ads, TikTok, programmatic) in the next 12 months? A platform that handles multiple channels gives you a single operational layer; an agency would need to be re-scoped or replaced as channels expand.

Red flags · when to walk from either model

Some agencies and some platforms should not get the engagement, regardless of the model math. Three red flags worth treating as deal-breakers:

• Opacity on creative production cost. An agency that won't tell you how many person-hours a variant takes, or a platform that won't tell you the marginal cost of an additional variant, is hiding something. Both should be answerable in honest numbers.
• Reporting that lives only in their environment. If you can't pull your own performance data from your Meta account without going through them, you are operating with a dependency that becomes a switching cost when you decide to leave.
• No clear methodology for creative diversity. Both agencies and platforms have to answer: how do you ensure variants are orthogonal in the way Meta's delivery system measures, not just visually different? If the answer is vague, the operation will burn budget in suppressed variants.

The honest framework · in one paragraph

Hire a Facebook ads agency if your monthly work is senior-judgment-heavy, your refresh cadence is quarterly or slower, your operation involves one or two parties, and your 24-month roadmap is roughly stable. Operate from a platform if your monthly work is production-heavy on established positioning, your refresh cadence is weekly to monthly, your operation involves three or more parties, and your 24-month roadmap involves scaling 2-5x or in-housing. Run both if you have the budget — with the platform handling production and the agency handling senior strategy — but be honest that the agency's scope has compressed.

Most brands that face this decision in 2026 land on the platform side of the framework. That's not a bias; it's what the market structure is selecting for. Senior creative judgment is still scarce and worth paying for. The volume work it sits on top of is not.

Hi Luca is built for the platform side of this decision, with native multi-role workflow for the cases where an agency partner remains in the picture. The model isn't agency or platform — it's platform-with-optional-agency-collaboration. That structure matches what most modern brand operations actually need. The agency-side operational view of this shift is narrated in Before / after running Hi Luca in an agency; the segment-specific framings live at /for-agencies, /for-global-brands, and /for-d2c.

Whichever you choose, decide on the math, not the relationship. The relationship can be rebuilt; the budget that gets burned by the wrong structural choice does not come back.