ROISROIS tells you whether an ad campaign is worth the spend — before a dollar goes out.
Bayesian statistics and Monte Carlo — transparent math, not a black box. And when the signal's too thin to call, it says so.
The problem
Teams greenlight campaigns on gut feel, last week's lucky ROAS, or a number a dashboard rounded up. The budget goes out, the result disappoints — and the honest question, was this spend ever worth it?, never actually got answered. Because nothing was answering it.
How it works
Category, margin, audience size, and whatever performance data you already have. None yet? That's fine — it works from benchmarks.
Bayesian priors from real benchmarks, updated by your data, run through thousands of Monte Carlo trials.
Run it, hold, or walk away — with the probability of profit and the math behind the call laid out in the open.
The output
78% probability of profit
10,000 Monte Carlo trials · priors updated with 1,240 of your conversions
Why ROIS
Thin data? ROIS says “insufficient signal” instead of inventing a number. Statistical honesty over false confidence.
Classic statistics you can audit — Bayesian updating and Monte Carlo. Nothing hidden; every verdict shows its work.
Starts from real 2025–26 Meta ad data, then adapts to your brand as the numbers come in.
Under the hood
Every number ROIS produces traces back to its inputs — nothing you can't audit. Two methods do the heavy lifting:
Turns benchmarks plus your own data into a probability distribution — not a single guess. Less data means a wider, more honest range.
Draws thousands of possible outcomes from those distributions to get your probability of profit, expected ROAS, and downside.
Backed by a toolkit of classical methods
New brands borrow strength from their category until their own data takes over — ideal when you're small.
Beta–Binomial and friends update beliefs cleanly as each conversion arrives.
Weighs expected value against downside to make the run / hold / stop call.
Wrings honest uncertainty out of the limited data you do have.
Decides when there's enough evidence to keep spending as results stream in.
Compares variants the Bayesian way — which is winning, and how sure we are.
The difference
Big teams running thousands of creatives have the data to feed ML. If you don't, statistics is the honest choice — at a price that fits.
Grounded in real data
ROIS starts from real 2025–26 Meta ad benchmarks — the priors below — then Bayesian-updates them with your brand's own numbers as they arrive.
FAQ
No. With zero data, ROIS runs on category benchmarks and tells you the prior expectation — and flags that the signal is weak. As your conversions come in, the verdict sharpens.
Today, no. ROIS is classic statistics — Bayesian updating and Monte Carlo, with every number traceable to its inputs and nothing you can't audit.
Dashboards tell you what already happened. ROIS answers whether the next dollar is worth spending — with a probability of profit, not a vanity metric.
Then ROIS leans on benchmarks and says so. It would rather tell you “not enough signal yet” than fake a confident number you'd bet a budget on.
Game theory. Ads are auctions, so competitor bidding pushes your CPM. We're modeling that as an auction game in a later phase — only once the competitor data is real enough to avoid false precision.
The name
ROIS is French for kings — the piece every game of chess is built around. And a chessboard is never luck. It's a closed system of odds, where strong players don't move on hope: they read the position, anticipate the opponent's reply, and play what wins more often than it loses.
The creative world is no different. From a first thought to a word to a decision, human behavior runs on patterns you can put numbers to — and an ad is just a move on a bigger board. Every dollar is a position; every audience, an opponent. ROIS reads that board before you commit — and because ads are auctions, your opponent's move is the next thing we'll model.
Get a statistically honest verdict on your next campaign — before you spend.
Join our waitlist→