One Channel Is 70% of Your Spend. That’s a Risk, Not a Strategy.

Four shocks that don’t care what your ROAS was

Concentration isn’t dangerous in the abstract. It’s dangerous because there are four documented ways a single channel stops working, and a concentrated account absorbs all four at full weight with nowhere to route around them.

1. Price shock — the slow one

Average Meta CPMs reached roughly $13.48 in 2026, up about 20% year over year, before seasonality. Q4 CPMs run 30–60% above the annual average, with Black Friday week regularly clearing two to three times normal levels.

A diversified account treats that as a routing problem: the auction reprices, budget moves to wherever the marginal dollar returns most. A concentrated account has no such option. It takes the full repricing at 71% weight and calls it a tough quarter.

2. Access shock — the fast one

In early-to-mid 2026 a wave of Meta ad account bans hit performance marketers, including long-running accounts spending six and seven figures a month. Many lost pixels, custom audiences and years of campaign history with the account. Google is no gentler — as we covered in Don’t Vibe-Code It, a suspension can land on detection without warning and survive most first-round appeals.

Here’s what makes this a concentration problem rather than a compliance one: the probability of a suspension is roughly independent of how much budget rides on that platform. The cost of one is entirely determined by it. You can do everything right on policy and still be running an account where one automated decision takes 71% of your acquisition offline.

3. Signal shock — the structural one

April 2021, Apple ships App Tracking Transparency. That February, Meta’s CFO told analysts the iOS headwind on 2022 was “on the order of $10 billion”. Conversion tracking broke, retargeting pools collapsed, lookalike models lost their signal.

Note who made that decision: not Meta, not you. A third company, for its own reasons, neither of you consulted. Advertisers concentrated on Meta absorbed the full degradation with no hedge. Rule changes arrive from outside the system — which is what makes them impossible to forecast and expensive to be undiversified against.

4. Saturation shock — the invisible one

Concentration and audience saturation compound, and this one costs you money while everything still looks fine. Any single channel has a finite addressable pool. Push more budget in and frequency climbs while incremental reach per dollar falls — you start buying repetition instead of people. The account reads as scaling right until it’s just paying more to reach the same audience twice.

Concentration is what forces you to keep buying into the flat part of that curve. There’s nowhere else to send the money.

Measure it: three numbers, ten minutes

You cannot manage an exposure you haven’t quantified. None of this requires anything you don’t already have in an export.

• Top-channel share. Spend on your largest channel divided by total spend. Most teams have never written this number down, which is precisely why it drifts.
• A concentration index. Borrow the Herfindahl-Hirschman Index from market-structure analysis: square each channel’s percentage share, then sum. A single-channel account scores 10,000; an even four-way split scores 2,500.

As a working heuristic — ours, not a regulatory standard — above 5,000 is single-channel dependency, 3,000 to 5,000 is a dominant channel with thin hedges, below 3,000 is a real mix.

One caveat matters more than the number itself: the index assumes your channels are independent, and most aren’t. Facebook and Instagram inside one Business Manager are one position, not two — a single access shock takes both. Score the failure modes, not the logos, or you’ll congratulate yourself on a diversification you don’t have.

Spend share versus contribution share. Put them side by side. If a channel is 71% of spend and 78% of contribution, concentration is earning its keep — on return. It’s still entirely unhedged on risk. Two separate questions that almost every account conflates into one.

Then run the test that turns the chart into a decision:

The 14-day dark test. Take your largest channel. Assume it goes to zero tomorrow, no notice, for fourteen days. Write down what happens to pipeline — and how much of the displaced budget the remaining channels could actually absorb inside their own learning constraints.

Most teams have never computed that second half. It’s the half that hurts.

Diversification isn’t free, and pretending it is would be its own kind of vibes

Spreading budget thin is a real failure mode, not a hypothetical one.

Every channel carries a minimum viable budget set by its learning requirements. Meta’s long-standing guideline for exiting the learning phase is roughly 50 optimization events per ad set per week — a signal-volume floor, not a performance guarantee. Below it, delivery stays unstable and the data you get back is noise you’ll then make decisions on. Split $30,000 across six platforms and you can comfortably fund six underpowered accounts, converting a concentration problem into a measurement problem and making both worse.

So the conclusion isn’t “diversify.” It’s:

• Know your concentration number — share and index, written down, updated monthly.
• Know whether the concentrated channel earns it on contribution share, not spend share.
• Know the minimum viable budget for any channel before you add it. If you can’t fund it past its learning threshold, you’re not diversifying, you’re donating.
• Decide deliberately, then re-decide as the numbers move.

Concentration can be the correct answer. An early-stage company with one channel that works and not enough budget to properly fund a second should be concentrated — and should know it’s carrying an unhedged position rather than executing a strategy, because those call for very different contingency planning.

What isn’t defensible is arriving at 71% by accident. Which is how essentially every concentrated account gets there.

Concentration is a drift problem, not a decision problem

Nobody chooses to put 71% of budget on one platform. You drift there, for three reasons that operate continuously:

• Platforms reward incumbency. The channel with the most conversion history optimizes best, so it earns more budget, so it accumulates more history. The loop runs whether you’re watching or not.
• Your rebalancing cadence is slower than the drift. Media mix gets revisited monthly or quarterly. Auctions reprice hourly.
• Manual corrections decay. By our own reckoning, an average of six days pass before a manual process surfaces a budget leak. Concentration re-forms in that gap, and every gap after it.

A quarterly diversification decision cannot hold a line against a system that re-concentrates daily. This is why teams that know they’re over-indexed stay over-indexed for years. The knowing was never the bottleneck.

The fix is a floor, not a rebalance

If the drift is continuous, the correction has to be too — and it has to be a constraint rather than an intention.

Two mechanisms, both enforced automatically: floor budgets per channel, so a secondary platform can’t be starved below its learning threshold by a winning primary, and a ceiling on top-channel share, so no position grows past the exposure you’ve agreed to carry. Not a reminder to diversify. A limit the system cannot cross.

That’s what W-04 monitoring looks like inside Sturnix: top-channel and contribution share tracked continuously rather than at quarter’s end, per-channel floors and caps enforced as hard guardrails you set once, and cross-platform reallocation that runs as marginal returns move — not when somebody finally notices the pie chart.

Your ROAS tells you how the position is performing. It has never once told you how large the position is.

Concentration isn’t the mistake. Unmeasured concentration is.