What $1M a month on Meta actually teaches you.

At $2K/day spend, Meta's auction feels like what it is: you bid, you win some impressions, you lose some, and performance is about your creative and your audience. At $30K/day spend across a single Meta account, the auction starts to feel like a thermostat. You're so much of the demand for your own inventory pool that changes you make move the auction dynamics in ways that feel wrong until you've seen them enough times.

Specific things we saw repeatedly at $15K+/day per account that we'd never seen at lower spend:

1. Self-competition is a real and expensive thing. Two ad sets with overlapping audiences in the same account at $1K/day each behave like two ad sets. Two ad sets at $15K/day each behave like one ad set bidding against itself. You'll see frequency spike and CPM climb without any apparent cause until you collapse the structure.
2. Learning phase becomes a seven-day tax, not a three-day one.At small budgets, Meta's 50-event threshold gets hit in a day and the account stabilizes. At 7-figure monthly spend on a specific campaign, you're often blowing through the threshold in hours — but the algorithm doesn't actually stabilize for five to seven days because the auction dynamics you're operating in are statistically noisier at high volume. Our rule of thumb: at $30K/day+, treat every new ad set as a seven-day investment, not a three-day one.
3. Advantage+ campaigns “go quiet” in a way that feels like a bug but isn't.We've seen Advantage+ Shopping campaigns at 7-figure monthly budgets suddenly concentrate 80% of spend on a single creative for 48 hours — a creative that wasn't the top performer a week before. This is Meta doing its thing. Whether you let it run or intervene is a judgment call. At $1M/month, “let it run” for 48 hours is a $60K decision.

The single largest structural difference between high-spend and low-spend accounts is how fast creative decays. At $2K/day, a winning creative can run for six to eight weeks. At $30K/day, that same creative in that same audience is cooked in ten to fourteen days. Sometimes faster.

The math falls out of saturation. Meta is a finite-eyeballs medium. You can't show the same ad to the same person thirty times without CTR collapsing. When you're spending enough to saturate the audience in a week, creative has to refresh in a week.

Those numbers are rough averages across different verticals. Home services concepts live longer than men's health concepts. Brand-driven creatives outlive offer-driven creatives. But the shape is the same: if you're at 7-figure monthly spend and your creative pipeline is producing three new concepts a week, you are losing money to saturation you can't see because the loss shows up as “auction inflation” in next month's CPM, not as a flagged event in your dashboard.

Why most agencies under-invest in creative at scale

The honest answer is pricing. Agency retainer math is usually set up so that creative production is a cost center and account strategy is the value center. When a client pays $15K/month for management of a $1M/month budget, the retainer is 1.5% of spend. Out of that the agency has to pay the buyer, the account manager, the reports, the tooling, and the creative. Creative is what gets squeezed.

The right answer is to set expectations up front that creative production scales with spend, not with headcount. A client spending $1M/month needs a creative engine that ships 30+ new ads a month. That's not a “creative person on the team” — that's a production system with briefs, specs, a vendor or studio pipeline, iteration cycles, and weekly performance feedback loops.

This is also why AI-assisted creative tooling is a bigger deal at high spend than at low spend. At $20K/month accounts, a human creative can keep up with the test cadence. At $1M/month accounts, the creative bottleneck is permanent and the only way through is to compress brief-to-shipped-variant time dramatically.

Every performance marketer knows Meta's reported conversions don't match the CRM. That gap is real at $10K/month. At $1M/month it's both larger in absolute terms and more structurally misleading, and both properties matter.

Three specific divergence patterns we saw repeatedly on 7-figure books:

1. Meta over-credits itself on view-through at scale. Click-attribution is roughly directionally right. View-through attribution gets dramatically more generous as spend concentration grows. At $1M/month spend, the click-only number vs. the default 7-day-click/1-day-view number can diverge by 40–60% on the same account. Neither is the CRM truth.
2. iOS signal loss isn't evenly distributed. Specific verticals and audience types take much harder hits than the platform average. Men's health, mental health, and medical lead-gen all saw material under-reporting vs the CRM of record through 2025 — more than the industry average numbers imply. The number you see isn't the number you get.
3. Attribution windows shift platform behavior, not just platform reporting.When you change your attribution window in Ads Manager, Meta's bidding algorithm optimizes differently. Moving from a 7-day-click/1-day-view window to a 7-day-click-only window on a $1M/month account isn't just a reporting change. It's a live change to how the algorithm decides who to show your ads to. This surprised us the first time and cost a real amount of money before we understood it.

The implication is that at high spend, you cannot treat Ads Manager as the system of record. You have to wire a pipeline from your CRM or appointment system back to the ad platform via CAPI and treat that data as authoritative for decision-making, even when it conflicts with the numbers in the UI. Most agencies know this in principle. Fewer have actually built the pipeline. At $20K/month it's optional. At $1M/month it's table stakes.

Most agency playbooks assume a morning-check-in rhythm: buyer logs in at 8am, scans Ads Manager, makes decisions, moves on. At $1M/month per account, that rhythm is a specific liability.

The math: spread evenly, $1M/month is $33,333 per day, or roughly $1,400 per hour. A signal that fires at 2pm and isn't acted on until 8am the next morning is an 18-hour delay. If the signal is meaningful — a creative collapse, an audience shift, a tracking break — that 18-hour delay is $25,000. Over a year's worth of delayed decisions, that's a number you can't ignore.

What actually works at high spend:

1. Alert-driven, not time-driven.A system that surfaces specific events (CPL up >20% day-over-day on a $5K/day campaign; frequency over 3.5 on a winning ad; a new creative crossing $2K spend with sub-threshold CTR) is worth far more than “checked it this morning.” The alerts need tight thresholds or they become noise. At scale you can afford to be picky.
2. Decision-filtered, not data-surfaced.At $1M/month spend, the dashboard has hundreds of data points. Looking at all of them is the wrong job. The right job is answering: “what's the smallest number of things I have to do in the next four hours to avoid a bad outcome, and what's the smallest number of things I should do in the next week to create a good one?”
3. Weekly compounding, not daily firefighting. Daily rhythm is for risk mitigation. Weekly rhythm is where the real performance work happens: creative reviews, audience rotation plans, LP iteration briefs, offer tests. Agencies that don't protect a weekly creative-and-offer session end up reactive on creative, which produces the decay we described earlier, which produces more daily fires, which consumes the weekly time. The loop is vicious.

Meta's UI has been pushing Advantage+ campaigns and broad- audience CBO as the default for three years. For most accounts at most spend levels, that's reasonable advice. For $1M/month accounts, it's a simplification that costs money.

The specific failure modes we saw on high-spend accounts using default platform suggestions:

1. Advantage+ Shopping over-concentrates creative at scale.At $30K/day, Advantage+ will happily park 70% of your budget on one creative for a stretch. If that creative is a genuine winner, great. If that creative is a winner only because Advantage+ didn't give the challengers enough volume to prove themselves, you've just locked in a local max. At 7-figure spend, “local max” can be 30% below what a true winner would produce.
2. Single-CBO structures lose the ability to isolate learnings.When you're running one CBO campaign with twenty ad sets at $1M/month, you can't cleanly read what audience is working versus what creative is working versus what offer is working. The signals are entangled. At small spend this is fine; the cost of experimentation is low. At high spend, blind signal is expensive.
3. Broad-audience targeting at high concentration distorts your own audience profile.Meta's algorithm is downstream of your pixel. When you spend $1M/month broadly and convert heavily among a specific user profile, the algorithm concentrates on that profile more aggressively over time. At 5-figure spend this is a feature. At 7-figure spend it can narrow your addressable audience in ways that hurt you the moment you try to expand or pivot.

What actually works at high spend is a hybrid: a high-conviction, tightly-targeted set of campaigns running against known-good audiences, plus a separate creative-testing structure that feeds winners into the first set, plus a controlled broad-audience Advantage+ campaign capped at a percentage of total spend. The structure is more complex to run. The returns on the complexity scale with spend.

This post has mostly been about Meta because that's where the biggest spend concentrations we ran lived. A quick note on Google at 7-figure monthly spend because the failure modes are different and worth naming briefly.

1. Performance Max is a black box that gets blacker at scale.At small spend, PMax is a convenience. At 7-figure monthly spend, PMax concentrates budget across channels in ways you can't see into without running a paired search-only campaign in parallel to benchmark. Several books we've seen discovered that their PMax was 80% brand-search traffic they would have gotten for free; the real incremental lift was a fraction of what Google reported.
2. Brand search cannibalization is a $100K/year mistake hiding in plain sight.When you run a big non-brand Google budget alongside an unmodified brand search campaign, Google will often divert meaningful non-brand budget to bid on your own terms. You're paying Google for traffic that would have converted anyway. At $1M/month Google spend, getting this wrong is a material number.
3. Offline-conversion upload cadence matters more as spend grows.Google's smart bidding gets dramatically better when you feed it offline conversion values. At low spend the effort isn't worth it. At high spend, a weekly offline-conversion upload with revenue values can materially compound over a quarter.

If you're running 7-figure monthly spend, or preparing to, here's the compressed checklist:

1. Your creative engine is your real moat.Eight new concepts a week is the floor. If you can't produce that, you're decaying faster than you're scaling and the dashboard will tell you about it three months too late.
2. Build your own attribution source of truth. Server-side CAPI, CRM-back feed, offline conversion uploads. The UI number is a polite fiction at this spend level.
3. Move to alert-driven intraday rhythm.A 24-hour response time to a meaningful signal is a $20K+ decision cost. Build the alert stack; trust the thresholds; don't micro- check when nothing's firing.
4. Separate your creative-testing structure from your performance structure.Let Advantage+ do what it's good at; don't let it also decide what's tested. The hybrid structure produces more durable performance.
5. Protect the weekly compounding time.Daily firefighting is inevitable. Weekly strategy time is the first thing to fall out of the calendar and it's the thing that determines whether next quarter's performance compounds or erodes.

Zeke AI is the system we wished existed while we were spending at this level. It reads the CRM, watches the portfolio, alerts on what matters, cites why, and learns from what you approve and reject. It doesn't replace the buyer's judgment. It gives that judgment a dashboard worth trusting and a rhythm that doesn't collapse as spend grows.