Your next quarter,
mapped before it happens.

Every tool you have looks backward.

Your bookkeeper tells you what happened. Your accountant tells you what happened last quarter. Your bank statement tells you what happened yesterday.

None of them tell you what’s about to happen.

For a small business, the thing you don’t see coming is the thing that kills you.

The more you share, the sharper the read.

Two numbers — revenue and costs — are enough. The engine reads your trajectory from those alone.

Share your balance sheet and it sharpens further. Cash runway, leverage pressure, margin compression — things revenue alone can’t show you.

The 9-field engine fires on 98.9% of out-of-sample predictions where balance-sheet data is available. The balance-sheet signals adjust the engine's read toward your specific financial profile and overlay real cash-out detection. Modal hit rises to 89.4% (vs 86.6%), top-2 to 91.1% (vs 89.1%), and outside-any-band drops to 2.1% (vs 2.6%) — better across every accuracy metric. The cost is modestly wider bands (1.35× current revenue vs 1.17×): the engine knows more about what could go wrong and admits more uncertainty in the tails. Catches cash-burning rising-revenue cases the basic engine misses.

Honest numbers at every scale.

The pooled 86.6% number is pulled up by large, predictable companies. Here’s the breakdown:

Larger businesses are inherently more predictable — diversified revenue, stable customer bases, less exposure to single-client swings. Smaller businesses swing harder, and the engine reflects that honestly.

The sub-$1M tier is dominated by pre-revenue junior explorers, speculative shells, and biotech capital-raising vehicles — not operating businesses. The 50% reflects how unpredictable those are and is included in the pooled number for transparency, not because we think the engine is broken.

Your data goes in. Scenarios come out.

Upload a spreadsheet. The engine reads your revenue trajectory, calibrates against thousands of similar businesses, and simulates hundreds of possible futures for next quarter.

Those futures cluster into 3–7 natural groups. Each group becomes a scenario card — a revenue range, a probability, a direction.

Some markets work. Some don’t.

The pooled 86.6% hides a wide regional spread. The engine hits its highest accuracy exactly where Simpler is launching — and its lowest where public markets are dominated by speculative shells with corrupted reporting.

Why Southeast Asia leads. SEA exchanges are overwhelmingly real operators — manufacturers, retailers, industrials, logistics — on mandatory quarterly reporting. The engine was designed for businesses like these, and the 88.9% accuracy reflects it.

Why East Asia is highest. Mainland Chinese, Japanese, Korean, and Taiwanese main boards mandate quarterly reporting and are dominated by established operators — a calibrated read on a well-disclosed market.

Why Europe and Oceania run lower. Both contain markets that file financials semi-annually rather than quarterly. The data feed reconstructs quarterly figures from those half-year filings, and the resulting series carry artifacts that any prediction engine will read as extreme volatility. On real-quarterly-reporting markets the numbers sit in the high 80s.

The Oceania subset remains in the pooled 86.6% headline. We don't exclude regions that look bad.

Every region number above is raw — no reweighting, no region-specific tuning, no cherry-picking.

Full data: 86.6%. Clean data: 87.3%. Almost no gap.

We ran the whole validation twice. First on the full dataset — every quarterly-reporting company, shells and commodity cyclicals and semi-annual-synthesis exchanges included. Second on a cleaned subset that excludes companies with (a) three consecutive identical revenue quarters (the semi-annual-as-quarterly synthesis pattern), or (b) raw sector classification in Basic Materials, Energy, Utilities, or biotech — the "real operating businesses on clean quarterly data" cohort a rigorous critic would demand.

The clean-vs-full gap is under 1pp. Shells, cyclicals, and semi-annual-synthesis exchanges drag the pooled number by less than a percentage point. The engine handles messy inputs without hiding behind them.

Both published. 86.6% is the honest "include every messy input" number a critic would audit; 87.3% is what the math does on clean operator data. Neither is more correct; they answer different questions.

More than 9 in 10 land in the predicted band.

When we miss, it’s almost always a business that just doubled or halved.

The engine misses about 7% of predictions. We characterized every miss:

The engine correctly captures normal business variation. It correctly misses the extreme events — a sudden major contract, a key client walking, a market shock. No spreadsheet predicts those.

359,896 predictions. One method.

Use the history available at that point in time. Predict next quarter. Compare against what actually happened. No peeking. No retraining on the answer.

These numbers validate the probabilistic math — the part that turns your spreadsheet into scenario cards with revenue ranges and probabilities.

Out-of-sample protocol: sector priors are fit on pre-2020 data only; predictions are tested on 2020-and-after quarters with no leakage from the test period. Every number above comes from a single unified run, 5,000 Monte Carlo trajectories per prediction. Engine = source of truth: probabilities, ranges, and labels emitted by the engine flow through unchanged to these numbers — no post-hoc massaging.

Stated 72.5%. Actual 86.6%.

“Bands too wide” is a fair criticism of any probabilistic forecast with a high hit rate. If the engine’s confidence in a card doesn’t match the empirical hit rate, the probability is lying. So we measured every one of 359,896 out-of-sample predictions — the modal card’s assigned probability vs whether the actual revenue fell in that card. A well-calibrated engine has them matching.

The engine substantially under-claims confidence — stated 72.5%, actual 86.6%, a 14pp gap in the safe direction. Truth lands inside some emitted scenario 97.4% of the time. Bands are wide by design — for an SMB operator, “revenue could be $X to $Y” is more useful than “revenue will be $Z” because the second is a lie that lets you plan wrong. When the engine says a range, the actual outcome lands inside it at the stated rate or better.

What we can’t do.