Hidden Risk in Your Best Channel

Abstract

Your highest-ROI channel is likely your riskiest. We analyzed 1,400+ Marketing Mix Models across the Cassandra platform and found that 73% of brands concentrate more than 40% of their marketing spend in a single channel -- the equivalent of putting most of your portfolio in one stock. When that channel underperforms (algorithm change, CPM spike, auction dynamics shift), there is no hedge. This is concentration risk, and most marketing teams do not measure it. The Herfindahl-Hirschman Index (HHI), borrowed from antitrust economics, quantifies the problem -- and reveals which marketing portfolios are dangerously exposed.

The Concentration Trap: Why High-ROI Channels Attract Too Much Budget

There is a pattern we see in almost every marketing budget that comes through Cassandra. It starts rationally. A brand tests five channels. One channel -- usually Meta or Google Search -- delivers the highest ROI. The team shifts budget toward it. The next quarter, it still leads. More budget moves. Within 18 months, 45-60% of spend sits in a single channel, and nobody flags it because every reallocation was individually justified by the data.

This is the concentration trap. Each marginal decision to move budget toward the best performer is locally optimal and globally dangerous. It is the same mistake retail investors make when they load up on a single stock that has been winning: the more concentrated you become, the more your entire outcome depends on one bet continuing to perform.

The financial analogy is exact. A portfolio with 60% in one stock has a higher expected return than a diversified portfolio -- until the day it does not. And the day it does not is the day that matters, because the loss from concentration is always larger than the gain from diversification was small.

In marketing, that day comes as a Meta algorithm update, a Google auction shift, a TikTok regulatory threat, or an iOS privacy change. It is never a question of if. It is a question of when. And most marketing teams have no idea how exposed they are when it arrives.

We quantified this exposure across our client base. The results are worse than expected.

Measuring Marketing Concentration Risk: The HHI Framework

The Herfindahl-Hirschman Index (HHI) is the standard measure of market concentration used by the U.S. Department of Justice, the Federal Trade Commission, and antitrust regulators worldwide. When the DOJ evaluates whether a merger would create a monopoly, they compute HHI. When central banks assess systemic risk in banking, they compute HHI. It is the canonical metric for answering the question: "How concentrated is this portfolio?"

The formula:

HHI = sum of (share_i)^2 for all i

Where share_i is the percentage share of each component, expressed as a decimal.

Interpretation thresholds (DOJ standard):

These thresholds are not arbitrary. They were established through decades of empirical analysis of market outcomes. Markets with HHI above 0.25 consistently show higher price volatility, lower resilience to shocks, and greater systemic fragility. The same dynamics apply to marketing portfolios.

A worked example

Consider a brand spending across three channels:

HHI = (0.60)^2 + (0.30)^2 + (0.10)^2 = 0.36 + 0.09 + 0.01 = 0.46

An HHI of 0.46 is nearly double the DOJ's "highly concentrated" threshold. By antitrust standards, this portfolio would trigger a regulatory investigation. In marketing, it is considered normal.

Now compare a more diversified allocation of the same $300,000:

HHI = (0.30)^2 + (0.25)^2 + (0.20)^2 + (0.15)^2 + (0.10)^2 = 0.09 + 0.0625 + 0.04 + 0.0225 + 0.01 = 0.225

Still moderately concentrated, but the risk profile is fundamentally different. The first portfolio lives or dies by Meta. The second portfolio can absorb a 30% decline in any single channel and still deliver 85%+ of expected revenue.

The math is elementary. The insight is that almost nobody computes it.

What the Data Shows: Analysis of 1,400+ MMMs

We computed the marketing HHI for every brand in the Cassandra platform with a completed Marketing Mix Model -- 1,427 models spanning ecommerce, B2B SaaS, fintech, DTC, and retail, representing $2.1 billion in measured ad spend.

Concentration is the norm, not the exception

Nearly three-quarters of brands have marketing portfolios that would trigger an antitrust investigation if they were markets. Only 8% achieve genuine diversification. The median brand uses 3-4 channels but allocates more than half of spend to just one.

The ROI-Variance Paradox

Here is the finding that should change how marketing teams think about channel selection. We computed both the median ROI and the coefficient of variation (CV) -- standard deviation divided by mean -- for every major channel across our dataset.

The paradox: channels with the highest average ROI also tend to have the highest variance. Meta Ads averages 3.2x ROI but with a CV of 0.68 -- meaning the standard deviation is 68% of the mean. In any given quarter, Meta's actual ROI could be anywhere from 1.0x to 5.4x. Google Search averages a lower 2.1x but with a CV of 0.31 -- its quarterly range is roughly 1.5x to 2.8x. Email has the highest ROI and the lowest variance, but it cannot absorb large budgets due to audience saturation.

This is the risk-return tradeoff, playing out exactly as it does in financial markets. High-return assets carry high variance. The channel your dashboard celebrates as the "best performer" is the same channel most likely to produce a disastrous quarter.

The HHI Contribution column shows why concentration compounds the problem. Meta's average 38% budget share contributes 0.144 to the portfolio HHI -- nearly reaching the "moderately concentrated" threshold by itself, before any other channel is counted. When the highest-variance channel is also the largest position, you are maximizing your exposure to exactly the wrong risk.

Concentration by industry

DTC and gaming brands are the most concentrated, driven by Meta's dominance in direct-response advertising. Omnichannel retail is the least concentrated, largely because these brands have legacy TV spend that provides structural diversification -- even when that TV spend was allocated for brand reasons, not risk management reasons.

The Hidden Cost of Concentration

Concentration risk is not theoretical. It has a measurable cost that shows up in three ways.

1. Algorithm dependency

In Q3 2024, Meta rolled out Andromeda -- a major update to its ad delivery algorithm that shifted optimization toward broader targeting and reduced the effectiveness of interest-based and lookalike audiences. Brands that had built their entire targeting strategy on these audience types saw CPAs increase by 25-45% overnight.

We tracked 89 Cassandra clients through this transition. The results split cleanly by concentration:

The CPA increase was similar across groups -- Meta got more expensive for everyone. But the revenue impact was 3x worse for concentrated portfolios because there was nowhere else for the budget to go. The low-concentration group had established channels (Google, YouTube, email) that absorbed reallocated spend within 2-3 weeks. The high-concentration group had to build new channels from scratch while simultaneously absorbing the Meta cost increase.

2. CPM volatility amplification

We measured the correlation between a brand's budget share in a channel and the CPM volatility it experiences in that channel. The relationship is not linear -- it is convex.

Brands spending 60%+ of budget on Meta experienced CPM volatility nearly 3x higher than brands at 20-30%. Part of this is mechanical: larger budgets reach broader audiences, which means more auction competition. Part of it is structural: heavy concentration in a single auction system means every fluctuation in that system hits you at full scale.

The Q4 seasonal spike is the clearest illustration. Every brand's Meta CPMs increase in Q4. But brands with 60%+ concentration see a 44% increase versus 12% for diversified brands. The concentrated brand's Q4 effectively costs 32% more than the diversified brand's Q4 -- and that premium buys no additional reach or return. It is pure concentration tax.

3. The cascading failure scenario

We documented one case in detail. A DTC brand with $1.2M monthly ad spend allocated 62% ($744K) to Meta, 24% to Google Search, and 14% to Google Shopping. HHI: 0.44. When Meta's Andromeda update hit, the brand's Meta CPA jumped 41% in three weeks. The team tried to shift $200K to Google Search, but Google's auction responded with its own CPA increase of 18% as the brand bid for incremental volume beyond its usual range.

The result: total marketing-driven revenue dropped 35% in a single quarter. Not because any individual channel failed catastrophically, but because the portfolio had no slack. Every reallocation triggered a second-order cost increase. The brand was playing chess with one move available.

Had the same brand entered the quarter with an HHI of 0.20 -- spread across five channels with no position above 30% -- the same Meta shock would have produced a 12-15% revenue decline. Still painful, but survivable without cutting headcount or inventory orders.

The difference between a 35% revenue drop and a 15% revenue drop is the difference between a bad quarter and an existential crisis. That difference is concentration risk.

Book a demo

The Portfolio Approach: Optimal Channel Diversification

How to calculate your marketing HHI

The calculation takes five minutes and requires only your current budget allocation:

Step 1: List every channel and its share of total marketing spend.

Step 2: Square each share (as a decimal).

Step 3: Sum the squares. This is your HHI.

Step 4: Compare to thresholds:

Target HHI ranges by business stage

Not every business should target the same HHI. Early-stage companies may need to concentrate spend to achieve minimum viable scale in a primary channel. Mature companies have no excuse.

These targets are calibrated from the Cassandra dataset. Brands in each revenue range that fell within the target HHI range showed 18-24% lower quarterly revenue variance compared to peers at similar revenue with higher HHI.

Rebalancing frequency

How often should you revisit your channel mix? The answer depends on your HHI and the volatility of your primary channel.

The goal is not to react to noise. It is to have a decision framework in place before a shock occurs. Companies that define rebalancing triggers in advance respond 3-4x faster than companies that convene ad hoc meetings when performance drops.

How to Apply This to Your Budget

Step 1: Calculate your current HHI

Pull your last quarter's channel spend breakdown. Apply the formula. Most brands are surprised by the number.

Step 2: Identify your concentration driver

Which channel dominates? What percentage of spend does it hold? What is its coefficient of variation? If your largest position is both the biggest and the most volatile -- the typical case -- you are compounding concentration risk with volatility risk.

Step 3: Model reallocation scenarios

Do not reallocate blindly. Use your MMM (or Cassandra's platform) to model how shifting 10%, 20%, or 30% of spend from your largest channel to underweight channels would affect:

• Expected total revenue (using diminishing returns curves)

• Portfolio variance (using channel-level variance and correlations)

• HHI (using the new allocation shares)

The Marketing Efficient Frontier framework provides the visual tool for this analysis: plot expected return against portfolio risk for every feasible allocation, and find the combination that sits on the frontier.

Step 4: Set an HHI target and a timeline

Based on your business stage and risk tolerance, choose a target HHI. Set a realistic timeline -- typically 2-3 quarters to move from an HHI of 0.40 to 0.25 without disrupting revenue. The transition requires building new channels incrementally, not slashing the dominant channel overnight.

Step 5: Integrate HHI into your reporting

Add marketing HHI to your quarterly board deck alongside ROAS, CPA, and revenue. The number takes 30 seconds to calculate and communicates portfolio risk in a way that any finance-trained board member understands immediately. It also creates accountability: if HHI rises above your target for two consecutive quarters, someone needs to explain why.

This is where the portfolio construction framework becomes operational. HHI is the diagnostic. The efficient frontier is the prescription. Together, they turn budget allocation from a quarterly negotiation into a disciplined capital management process.

Conclusion

Marketing channel risk is not about individual channel performance. It is about portfolio structure. A channel with volatile returns is manageable when it is 15% of your budget. The same channel becomes an existential threat at 55%.

The HHI is not a new metric. It has been the standard measure of concentration risk in economics and antitrust law for decades. Its application to marketing is overdue. The calculation is trivial. The insight it provides -- that most marketing portfolios carry concentration risk equivalent to a monopolistic market -- is not.

We found that 73% of brands in our dataset exceed the DOJ's threshold for "highly concentrated." These brands are not reckless. They arrived at concentration through a series of individually rational decisions: allocate more to what works. The trap is that "what works" is measured by ROI, which captures return but not risk.

The fix is straightforward. Calculate your HHI. Compare it to the thresholds. If you are above 0.25, model reallocation scenarios using your MMM's diminishing returns curves and the risk-adjusted returns framework. Set an HHI target. Rebalance toward it over 2-3 quarters. Monitor the number alongside ROAS the way a fund manager monitors portfolio beta alongside alpha.

Your best channel is not your safest channel. It is almost certainly your riskiest. The question is whether you know that before the next algorithm change, CPM spike, or platform disruption tells you the hard way.

How to Run This Analysis

Your marketing HHI takes five minutes to calculate with a spreadsheet. Understanding what to do about it -- how to model reallocation scenarios against diminishing returns curves, how to find the optimal diversification point on the efficient frontier, how to quantify the downside exposure of your current allocation -- requires a Marketing Mix Model with full posterior distributions.

Cassandra computes your marketing HHI, channel-level risk profiles, and optimal reallocation recommendations automatically as part of every model run. If you want to see your concentration risk quantified and know exactly where to reallocate, book a call with our team.

Ready to measure the concentration risk hiding in your marketing portfolio?

We'll calculate your marketing HHI and show you exactly where diversification would reduce risk without sacrificing returns.

Book a demo

Methodology: Analysis of 1,427 Marketing Mix Models across ecommerce, B2B SaaS, fintech, DTC, and retail, representing $2.1 billion in measured ad spend on the Cassandra platform.

Author: Gabriele Franco, Founder & CEO of Cassandra