A founder messaged me in March: “We’ve doubled our Meta spend and revenue has only grown 30%. What’s broken?” Nothing was broken. They were running a 2023 operating model on a 2026 system. Their creative pool was too thin, their structure was too complex, and their testing budget was non-existent. The platform was working exactly as designed. They just weren’t.

Scaling paid media on Meta in 2026 is not the job it was in 2022. The auction is different, the levers are different, and the operating model has to be different. The brands who keep growing are the ones who’ve internalised that. The brands who plateau are the ones still optimising the wrong layer.

The practical operating model: what the week actually looks like

Forget the workflow your agency built in 2021. Here is what a well-run Meta operating model looks like across a 28-day cycle in 2026.

Week one: ship between eight and twenty new creative concepts depending on spend level, with a structured brief tied to two or three strategic angles. Set them live in a testing campaign at 10–15% of total spend. Don’t touch the production campaign except for budget pacing.

Week two: review testing results against a clear threshold, usually a CPA within 25% of the scaling campaign’s CPA at minimum spend volume. Promote winners into the scaling campaign. Cut losers. Brief variants of the winners within 72 hours.

Week three: ramp budget on the scaling campaign in 15–20% increments, watching MER not just platform ROAS. If MER holds, ramp again. If it softens, hold and refresh creative.

Week four: full creative pool audit. Count distinct concepts delivering 5%+ of spend. If the number is under five, the bottleneck is creative volume and that’s where the next month’s energy goes.

This rhythm is more disciplined than what most accounts run, and less complex than what most agencies sell. The discipline is the point.

KPI expectations in an AI-first system

The hardest conversation I have with founders in 2026 is recalibrating which KPIs matter. The metrics agencies still report (CTR, ad-set–level ROAS, CPM by audience) are mostly noise now. The system is making decisions on data and dimensions the dashboard can’t see, so optimising against dashboard metrics is optimising against a reflection.

Direct answer to the KPI question, in 50 words: in an Andromeda-driven account, the metrics that actually matter are MER (blended revenue ÷ total ad spend), first-order new customer CPA, contribution margin after media, and creative concept refresh rate. Platform ROAS, ad-set–level metrics, and CTR are diagnostics, not targets. Don’t run the account against them.

Two implications most brands don’t act on. First, weekly granular reporting becomes counterproductive. It tempts intervention on noise. The cleanest accounts I run report MER weekly, but the campaign-level review happens monthly, and structural changes happen quarterly. Second, “good” CPA is now an account-specific number derived from contribution margin, not a category benchmark. The brand next door selling at a higher AOV can tolerate a higher CPA than you can. Stop comparing.

The mistakes brands are making right now

I’ll give the four I see most.

Mistake one: optimising the structure instead of the inputs. Adding and removing ad sets, splitting and merging audiences, pausing and unpausing campaigns. Most of this work is theatre. The model is overriding it anyway. Time spent restructuring is time not spent on creative volume, signal quality, and offer testing. The first thing I do on a stalling account is reduce the structural fiddling, not increase it.

Mistake two: under-funding creative as a line item. Brands spending £80k a month on media will commission £6k a month of new creative and wonder why scale has stalled. The honest ratio for a scaling brand is closer to 10–15% of media spend reinvested into creative production. Below that you’re starving the only lever you have meaningful control over.

Mistake three: confusing platform ROAS with business reality. I’ve watched brands cut Meta budget because platform ROAS dropped from 3.8x to 3.2x, only to see total revenue drop the next month, because the “low ROAS” Meta spend had been driving demand that converted on other channels. Look at MER. If MER is holding, the platform is doing its job, even if the numbers underneath it shift.

Mistake four: treating Advantage+ as one of several options rather than the default. Most accounts I audit in 2026 should have 70%+ of their spend running through Advantage+ Shopping. Most have 30–40%. The gap is the operating cost of clinging to an older mental model.

A framework for growth-stage versus scale-stage brands

The right operating model depends on where you are. The framework I use:

For growth-stage brands (under roughly £40k a month in Meta spend): focus is on finding the creative-offer fit that breaks scale. Run one production campaign and one testing campaign. Push 70% of budget into Advantage+. Invest disproportionately in creative volume. Aim for 10–12 concept-level new ads per month at minimum. Accept platform ROAS that looks worse than your peers’ early on, because you’re paying to find the winners that compound later. Don’t measure success in weekly MER yet; measure it in how many concepts cross your “scalable” threshold per quarter.

For scale-stage brands (£40k–£200k a month): focus shifts from finding winners to maintaining throughput. The mathematics of fatigue matter more than novelty. Operating model is two production campaigns (split by major market or product line), one testing campaign, and a relentless creative pipeline shipping 15–25 concepts a month. Budget moves on MER and contribution margin, not on platform ROAS. Reporting cadence becomes weekly MER, monthly account review, quarterly structural decisions. The work is less interesting and more repeatable. That’s the point.

The brands I see fail tend to be growth-stage brands trying to act scale-stage (over-structured, under-experimented) or scale-stage brands still acting growth-stage (chasing every novel format, never settling into rhythm). Knowing which mode you’re in matters more than the tactics inside either.

Where this leaves you

The thing nobody is saying loudly enough about scaling on Meta in 2026 is that the work has become more about operations than about cleverness. Anyone can copy a tactic. Almost nobody can run the operating model with discipline for twelve months. The compounding edge is in the rhythm, not the trick.

If your Meta account has plateaued and you can’t decide whether the problem is the platform, the creative, the structure, or the operating model, it’s almost always the last one. The platform is the easy thing to blame. The operating model is the hard thing to fix.

If you’d like a second view on which mode your account is actually in and what the next 90 days should look like, I’m easy to reach.