Same budget, which creator portfolio delivers more?

How much one creator swings

Start with the picture. Each row is one of the more active creators, drawn from their worst post to their best:

The middle creator's best post is 14× their worst, and half the creators swing wider than that. A third of what one creator delivers in a quarter comes from their single biggest post.

The chart is not saying some creators are steady and some are not. Every one of them is unsteady — almost every creator's results spread across a very wide range.

So don't plan around which post will take off. Creators can't predict that about themselves either.

But picking creators is not a raffle

The other half of the same data: group the posts by author, and half of the variance in outcomes is decided by who posted. Picking controls a roster's average level; what it can't control is any single post.

One number surprised us: follower count explains only 7% of the variance in post outcomes. Picking from public profiles is close to not picking at all. The signal that works is a creator's record across real campaigns — how much effective exposure each post actually earned, and where it ranked in its own campaign. That record doesn't live on any public profile; it lives in our settlement data. It's what we build brand rosters from.

How far can you trust the record? We split the campaigns into an earlier and a later half: of the strongest earlier cohort, half stayed in the top tier, and a third dropped below the middle. Far better than a coin flip — nowhere near "buy last season's number one and relax."

Same hits, different people every time

The top 10% of posts carry half of all effective exposure. Splitting the later half's hit posts — the top decile within their own campaign — by the author's track record:

Veterans are genuinely strong: roughly one hit in every four to five posts, and 45% of all hits are theirs.

But 42% of the hits came from creators with no record at all — people no data could have picked in advance.

And only a third of these hits came from someone who had ever produced a hit before. The other two thirds hadn't hit in the previous round.

Build the list from records alone, and that 42% is never yours.

How we build the portfolio

We backtested it on real history: same budget, same number of posts, a roster picked on records earns about seventy percent more than one thrown together — sometimes double.

Now give a few of those slots to newcomers: overall exposure barely moves, but the newcomer hits — and the fresh distribution they open up — can be worth far more than recycling the same accounts.

As for pushing the whole budget onto two or three stars: the paper math looks best, but you can't buy it — one creator posts one or two originals per campaign, nowhere near a full campaign's volume. Even if you could, when one person has an off run, all their posts sink together; top-tier pricing runs higher too. It's also hard to scale — too much of it is non-standard and uncertain.

So this is how we build brand rosters now: most slots go to creators with proven records, so the campaign's volume has a floor you can rely on; a few slots go to newcomers, a bet on where the hits come from. Like a stock portfolio — every single position is uncertain, but a well-built portfolio balances the whole.