Recursive self-improving decision intelligence for growth companies
Optimize your growth. By evidence, not opinions.
BayesBrain watches every number in your company, tells you what to improve, how, and learns from its effect. Getting recursively better by every decision.
Signal. Action. Outcome. Learning.
One loop, run continuously on everything you connect. Each turn of it leaves the company measurably closer to optimal, or tells you plainly that it did not, and the next turn starts from what this one learned.
Signal
BayesBrain watches every series across your tools and surfaces only the changes that are real, material and yours. Everything else stays quiet.
Trend and shift tests · false discovery control · materiality floorAction
Each signal comes with the action most likely to move it, and the reasons BayesBrain has for it, each one traceable to your own numbers.
Recommended action · the reasons · the evidenceOutcome
Take the action, and BayesBrain measures it day by day against what would have happened without it, in the metric and in dollars, until there is a verdict.
Counterfactual · day by day · a verdictLearning
Every measured outcome is kept, and the next proposal of its kind starts from it. Its range narrows, and its probability becomes your own.
Logged · measuring · outcome · learnedHistory
Every action leaves a trace. See what changed, when you changed it, and what happened to the metric afterwards.
What changed · when · what happened afterYour company, read as one system. Every tool on one daily history.
BayesBrain connects the tools you already use and turns their numbers into one continuous history of how the company changes, so each number is read against the others.
Your tools
One daily history
Illustrative values
Your data stays yours. Read, never changed. Never trained on.
BayesBrain only reads your tools, keeps what it reads encrypted and separate from every other company, and removes it when you disconnect.
We only read
BayesBrain never creates, changes or deletes anything in the tools you connect.
Encrypted, with our own key
HTTPS in transit and encrypted at rest. Access tokens are also encrypted with a key we hold outside the database.
Separate for every company
Every request is scoped to one company, and row-level security is on for every table.
Not used to train AI
Anthropic and Voyage AI, our AI providers, do not train models on what we send them.
Disconnect at any time
A tool’s access is deleted at once and its synced data within 30 days. Account deletion on request.
Stored in Frankfurt
Your data is stored in our database in Frankfurt, Germany.
Your company context, available to any agent or language model.
Connect BayesBrain through MCP so the AI you already use can retrieve your company's metrics, history, actions and measured outcomes before answering what to do next.
BayesBrain via MCPWhat should I focus on this week?
Looked through 148 series and 8 actions in BayesBrain · 4 used
Using BayesBrain via MCP4 relevant of 148 series and 8 actions
MRRMetric$37 300ARR growth, year on yearMetric+41%
Visitor to signupMetric7.8%Churn, monthlyMetric3.2%Trial to paidMetric31% · was 34%
CACMetric$28Ad spend, MetaMetric$5 300
Ad spend, TikTokMetric$8 000
Cash on handMetric$924 500
Trustpilot ratingMetric4.1
Newsletter conversionMetric2.4%Search trafficMetric12 400
App Store ratingMetric4.5MRRMetric$37 300ARR growth, year on yearMetric+41%Visitor to signupMetric7.8%Churn, monthlyMetric3.2%Trial to paidMetric31% · was 34%CACMetric$28Ad spend, MetaMetric$5 300Ad spend, TikTokMetric$8 000Cash on handMetric$924 500Trustpilot ratingMetric4.1Newsletter conversionMetric2.4%Search trafficMetric12 400App Store ratingMetric4.5Trial to paidChange31% · was 34%Focus on trial-to-paid conversion.Trial volume is up, but conversion has fallen from 34% to 31% Trial to paid 31%Trial to paid31%, down from 34% in 8 weeksTrial starts up 26% over the same weeksBayesBrain · /metrics, so I'd test the trial before raising acquisition spend Conversion 58ConversionScore 58, down 1 this weekTop signal: paid conversion is fallingBayesBrain · /overview. The bank sync status page is already lowering churn Action 006006 · Bank sync status pageMeasuredChurn −0.8 pts, likely −1.2 to −0.4Measured over 41 daysBayesBrain · /decisions/status-page, while the onboarding checklist is still being measured Action 007007 · Onboarding checklist v2MeasuringTrial to paid, day 12 of 28BayesBrain · /decisions/checklist.
Trial to paidLast 8 weeksVisual answers inside your assistant: coming soon
Illustrative values
Beliefs that move with evidence.
Thomas Bayes
c. 1701 to 7 April 1761
P(A|B) P(B) = P(A∩B) = P(B|A) P(A)
BayesBrain's engine is statistical inference, not machine learning. Detection is frequentist: trend tests corrected for autocorrelation, false discovery control across every series, and a materiality floor so only changes that are both real and large enough to matter surface. Measurement and memory are Bayesian: beliefs about what works start wide, and every observed outcome narrows them. The prior becomes the posterior, and the posterior becomes the next prior. The system does not get opinions. It gets evidence.
A rank-based test for whether a series is trending rather than wandering. It compares every reading with every later one and counts how often it rose against how often it fell, so it assumes no particular distribution and a single freak week cannot manufacture a trend.
Business metrics carry over week to week, so consecutive readings are not independent evidence. Left uncorrected that inflates significance and a flat series starts producing findings. The test's variance is corrected for the series' own serial correlation, which lowers the effective sample size to what the data actually supports.
Testing many metrics at once means some will look significant by chance alone. Rather than one threshold per test, the p-values are ranked together and the cut is placed so that only a small expected share of what we report is a false alarm.
Statistically real is not the same as worth your attention. A change also has to clear a size that matters for your business before it is written up, which is what keeps a technically significant but trivial move out of the brief.
An estimate starts from a prior drawn from comparable cases and is revised as your own data arrives. Early on the estimate leans on the prior and says so; as your evidence accumulates it takes over, and the stated uncertainty narrows with it.
FIG. 01 · BAYESIAN UPDATING
What would BayesBrain find in your company?
Give us your website. BayesBrain reads your company, connects read-only to the tools you already use, and shows how you are doing against similar companies, and what to improve.