The Advice Everyone Hears
If you spend any time in Facebook advertising communities, you've heard some version of this: upload more creative. Test more variations. The algorithm rewards creative diversity. Give it options and it will find the winners.
This advice comes from everywhere — Meta's own marketing materials, media buying agencies, ad coaches, conference speakers. It sounds intuitive. More options should mean better outcomes. More creative should give the system more room to optimize.
"Upload more creative. The algorithm needs options."
This is one of the most repeated pieces of advice in Facebook advertising. It is also one of the least examined.
Measuring What Meta Won't Define
Meta never precisely defines what "creative diversity" means. They use the phrase constantly — in blog posts, in Andromeda marketing, in advertiser guidance — but they never provide a formula, a threshold, or a measurement. You're told to maximize something that has no clear definition and no reported metric.
That's a problem if you want to test the claim. So we needed a proxy — something measurable that captures the spirit of what Meta is describing. The best data available is how Meta itself allocates your spend across your ads. If the system spreads budget across many ads, that's the algorithm treating your creative as diverse. If it concentrates budget into a few ads, the system is effectively saying your "diversity" isn't useful enough to distribute across.
We built the Spend Diversification Score, or SDS, to measure exactly this. If one ad absorbs all the budget, your SDS is 1. If spend distributes evenly across ten ads, your SDS is 10. It's not a perfect measure of creative diversity, but it's the best proxy available using data advertisers actually have. And it gives us a way to test Meta's creative diversity claim with data rather than anecdotes.
The math behind SDS.
If you have 50 ads but 90% of your budget goes to 3 of them, your SDS is closer to 3 than 50. The metric captures how diversified your spend actually is — not how many ads you uploaded, but how many are meaningfully participating in delivery.
SDS is mathematically identical to the inverse Herfindahl–Hirschman Index (HHI) — a standard concentration metric used in economics to measure market dominance and in ecology to measure species diversity. It's the same formula applied to ad spend: how concentrated or distributed is the allocation across your creative portfolio? It is well-understood, widely published, and not something we invented. We just applied it to a context where Meta chose not to.
For those who want the exact calculation:
SDS = (Σ spendᵢ)² / Σ(spendᵢ²)Where spendᵢ is the dollar amount spent on each individual ad during the measurement period. The numerator is total spend squared; the denominator is the sum of each ad's spend squared.
What the Spend Diversification Score Actually Measures
Applied to ad spend, SDS answers a single question: if spend were evenly distributed, how many ads would it take to produce this concentration pattern?
This is a useful diagnostic. But it measures how Meta allocated your spend — not whether that allocation was good for your business.
That distinction matters. A high SDS means Meta is spreading your budget across many ads. It does not mean those ads are performing well. It does not mean the diversity is profitable. It does not mean more creative would improve results.
SDS is an output of the system's allocation decisions. But Meta's marketing narrative treats creative diversity as an input you should maximize — upload more creative, give the algorithm options, and performance will follow. SDS lets us test whether that's actually true.
What the Data Actually Shows
🖼 Asset needed — chart: "Same-Week Revenue Correlation: New Ads vs Active Winners." Two rows of scatterplots across four e-commerce business categories (Pet Products, Bath Products, Cleaning Products, Optical), each plotting weekly revenue. Top row — New Ads Launched — shows r-values scattered near zero (r = -0.00, 0.25, 0.54, 0.11). Bottom row — Active Winners — shows strong, statistically significant correlations (r = 0.87, 0.64, 0.87, 0.60). Re-export from the original analysis rather than screenshotting the PDF.
These charts compare two different ways to predict weekly advertising revenue across four e-commerce businesses. The "r" value shown on each chart measures how strongly two things move together — the closer to 1.0, the tighter the relationship. Green highlighting indicates a statistically reliable correlation, yellow indicates moderate reliability, and white indicates little to no meaningful relationship.
The top row shows that simply launching more new ads has almost no connection to revenue — the dots are scattered randomly, and the r values hover near zero. The bottom row tells a very different story: the number of Active Winners (ads that proved profitable early and remain profitable) strongly predicts revenue, with r values ranging from 0.60 to 0.87. The takeaway is clear: scaling ad revenue isn't about producing more creative — it's about producing more quality creative. This aligns with Meta's emphasis on "Creative Diversity," but with an important nuance: diversity only matters if you're diversifying with winners.
The Distinction That Changes Everything
The most important finding in our analysis was not about SDS itself. It was about which ads you include when you calculate it.
Raw SDS does not predict performance. Filtered SDS does.
When we calculated SDS across all ads with any spend — the raw, unfiltered version — the relationship to future performance was unreliable. Sometimes positive, sometimes negative, often statistically insignificant.
When we filtered to only include ads that were meeting profitability benchmarks before calculating SDS, the picture changed. A diverse set of profitable ads predicted better performance over the following days. Consistently. Across business types.
This makes intuitive sense once you see it. If you have ten ads and eight of them are burning money, having high SDS means the system is enthusiastically spreading your budget across eight losers and two winners. That's not diversity working for you. That's diversity working against you.
But if you have ten ads and seven of them are profitable, high SDS means the system has found multiple distinct audience pockets it can serve profitably. That is the diversity that helps.
The metric is identical. The ads going into it are completely different. And that difference is the entire ballgame.
Why Meta Promotes the Simpler Story
"Upload more creative" is commercially useful advice for Meta. Every new ad you upload is a new candidate for the auction system. More candidates mean more competition, which drives better ad pricing — for Meta. More creative also means more impressions Meta can serve, more total spend it can absorb, and more data it can collect about which formats and messages drive engagement.
None of that is sinister. Meta's system genuinely does benefit from having more options to evaluate. But "the system benefits" and "your business benefits" are not the same claim, and Meta's marketing materials don't draw that distinction.
Meta's incentive is volume. Your incentive is profit.
When Meta promotes creative diversity, it is describing what helps its auction system run efficiently. That may or may not align with what helps your business run profitably. The burden of figuring out which is on you.
Why "Spray and Pray" Fails
If raw creative volume drove performance, the strategy would be simple: produce as many ads as possible, upload everything, and let the algorithm sort it out. This is, in practice, what a lot of advertisers do. The industry has a name for it: spray and pray.
It doesn't work, and the mechanics we described in earlier posts in this series explain why.
Three forces work against creative volume without quality.
The pacing system is indifferent to quality. As we described in Meta Is Not Sequencing Your Ads, the pacing system's job is to spend your daily budget. If you give it 50 ads and only 5 are strong, it will allocate spend to the 5 winners first — but then it will continue deploying budget into the weaker options to maintain delivery. More bad ads means more places for the system to waste money while faithfully spending what you told it to spend.
The "best worst ad" problem scales with volume. For every audience segment where none of your ads are strong, the system picks the least bad option. If you have 10 ads, there are limited "least bad" options. If you have 50 ads, you've created 50 candidates to be the least bad option for niche audience pockets you don't even know exist — and the system will spend money on every one of them.
Learning phase costs multiply. Every new ad enters a calibration period where the system gathers enough data to estimate its performance. During this period, delivery is less efficient. If you upload 20 new ads at once, you're paying the learning-phase tax 20 times simultaneously. The cumulative cost of all those calibration periods can be substantial — and most of those ads will turn out to be losers that absorbed real budget while the system figured that out.
The experienced practitioners who say "quality over quantity" are responding to this reality, even if they can't always articulate the mechanism. They've seen what happens when you flood an account with mediocre creative. The numbers get worse, not better.
What This Means for Your Campaigns
Creative diversity is not a strategy. Profitable creative diversity is. The goal is not to maximize SDS. The goal is to maximize the number of ads that individually meet your profitability benchmarks — and then let the system distribute spend across those winners. That is a fundamentally different objective than "upload more creative."
Monitor SDS as a diagnostic, not a target. SDS tells you how concentrated or distributed your spend is. That's useful information. But whether high or low SDS is better for your business depends entirely on what's underneath the number. A high SDS where every contributing ad is profitable is a strong signal. A high SDS where the system is spreading budget across a mix of winners and losers is a red flag.
Audit what's getting spend, not just how much. Instead of asking "how many ads should I run?" ask "how many of the ads currently getting spend are meeting my benchmarks?" If the answer is three out of fifty, you don't need more creative. You need fewer ads and better ones.
Test deliberately, not indiscriminately. There is real value in creative testing. New ads can find new audience segments, refresh fatigued creative, and reveal messaging angles you hadn't considered. But testing should be a disciplined process: launch a limited number of variants, give them enough budget to exit the learning phase, evaluate against your benchmarks, and kill the losers quickly. This is the opposite of uploading 30 new ads every week and hoping something sticks.
Be skeptical when a platform tells you to produce more of what it sells. Meta benefits from creative volume. Agencies benefit from producing creative volume. The entire supply chain around Facebook advertising is incentivized to tell you that more creative is the answer. Your economics are the only counterweight to that incentive, and as we described in Optimizing for Purchases Is Not Optimizing for Profit, only you can see your economics.
A Simple Experiment
Try this with your own account.
Pull a list of every ad that received spend in the last 30 days. For each one, calculate whether it met your break-even CPA (or target CPA, or whatever profitability benchmark you established after reading paper #2 in this series).
Now calculate SDS twice: once for all ads with spend, and once for only the ads that met your benchmark.
If the two numbers are similar, your creative portfolio is healthy — Meta is mostly spending on profitable ads, and the diversity in your account is the kind that helps.
If all-ads SDS is much higher than profitable-ads SDS, you have a problem. The system is distributing budget across ads that aren't working. You're subsidizing losers. The "diversity" in your account is costing you money.
The gap between those two numbers is a better measure of your creative health than SDS alone will ever be.
The Bottom Line
Meta wants you to upload more creative. The coaching industry is split between "more is better" and "quality over quantity." Both sides are arguing about the wrong variable.
The question is not how many ads you have. It's how many profitable ads you have. Creative diversity measured across your entire portfolio — the raw SDS — is noise. Creative diversity measured across your profitable ads is signal.
That distinction won't appear in Meta's investor blog posts. It won't appear in Ads Manager. It requires knowing your own economics and applying them to the data the platform gives you.
If that sounds like a recurring theme in this series, it is. The system is sophisticated. Your data is incomplete. And the only person who can close the gap between what Meta optimizes and what your business needs is you.