27 tensions. Six domains. One read.
It watches your whole business continuously and surfaces the one thing about to break, before it does.
Two healthy numbers. One unhealthy pair.
A tension isn't a bad metric. It's two normal numbers colliding: stretched, not yet snapped. The kind of thing no single chart can show.
+16% MoM
Healthy on its own. The chart is green and everyone's pleased.
3.1 months
Normal on its own. Thinner than last quarter, nothing alarming.
A countdown
Growth eats cash faster than it refills, and runway crosses one month in six weeks.
Each of the 27 tensions is a pair like this, watched continuously, on your model's thresholds.
Where this goes.
The pair above, left alone. Every tension card ships with this chain: phases, dates, and the lever that breaks it.
Caught at phase two, this is a billing-terms change. At phase four, it's a layoff plan.
Reading, not reporting.
The mechanics behind the one weekly read.
27 tensions, 6 domains
Liquidity · Demand · Capacity · Decisions · Structure · Retention. Watched continuously, on your model's thresholds.
Where the corners cross
The worst problems only show where domains meet. The engine reads them together, not in silos.
The bigger shape
When several tensions fire at once, it names the pattern, not the symptoms.
Says when it's guessing
Confidence scores, kept subtle. It asks one question instead of faking certainty.
How a number becomes a decision.
Nothing here is a dashboard query. Raw events become signals, signals become evidence, evidence is matched against the library, and only what survives confidence becomes a read.
deal won · invoice sent · ticket closed · check-in filed
cash buffer · backlog age · founder share · churn rate
which signals, from which hub, how fresh, how complete
27 patterns, weighted, on your company type and stage
one dominant tension · confidence · one lever
The output is a Tension Case: the object you actually work with, the way a CRM works with a deal.
Every decision teaches the next one.
The pipeline above runs once. This one closes. Every lever you pull is logged, the outcome is checked a month later, and what actually worked changes what the engine recommends next time, for you and for companies shaped like yours.
the read names the tension and the one lever
you pull it, or you deliberately do not
the decision is logged with an owner and a review date
30 to 60 days later it asks what actually changed
the observed outcome updates the weights
Tension, lever taken, outcome observed. No competitor holds this table, and it only fills from real use.
Named, not numbered.
Ten of the 27, by name. Each one is a pair like the one above, with thresholds tuned to your business model, not a generic SaaS template.
Each carries the pair it watches, the threshold it fires on, and the lever that answers it.
What breaks in Q3 signals in Q1.
The months between the first signal and the visible damage feel like calm. They're the window, where the cheap fixes live. The engine reads that gap continuously, so you act while it's still a decision, not a loss.
Leaving clients show up in the read long before they show up in revenue.
A founder bottleneck slows sales long before any report says so.
An overloaded team reaches delivery first, margins after.
By the time the P&L notices, you've already acted.
AI helps. It never decides.
The part that flags your risks runs on deterministic rules: same numbers, same read, every time. AI works around the engine, never above it.
Rules, inspectable
Every tension shows its thresholds and evidence. Nothing is flagged by a guess.
AI around the engine
Drafting campaigns, building pages, explaining the why: acceleration, not authority.
Works without it
Switch the assist off and Monday's read still arrives. The engine never waits on a model.
Sensitive calls stay yours. The engine brings the evidence; you bring the judgment.
Get your first read.
Run the Stress Test with your numbers. The engine shows its work.
Run the Stress Test →