Know what the official record is missing.
Official records tell you what was notified. Zbaseline reads the independent traces around them — trade, procurement, product testing — and tests how often those traces moved first.
Illustrative interface · sample data · anonymised corridors
Trade, procurement, product testing and other public traces, standardised into indicators.
Every indicator run against past events, using only the information available at the time.
What led, what lagged, and what fired with nothing behind it.
The times an indicator was wrong are published alongside the times it was right.
Three times the reported figure was wrong by this much
Between an analyst's herd-loss estimate and China's official cumulative African swine fever cull.
Rabobank vs MARA · 2019
Countries graded "BSE likely" on import records alone, while each reported zero cases. Three confirmed within two years.
EU Scientific Steering Committee · 2000
US states where H5N1 turned up in retail milk before a single infected herd had been reported there.
Retail milk monitoring · 2024
The record is kept by the people it costs.
Each of these systems is complete on its own terms and blind at its edges. Evidence that falls between them is not recorded anywhere — and the reasons are mostly structural.
Records what was notified.
Records the seizure.
Records the consignment.
Why the gap opens
- Export status to protect
- Local reporting incentives
- Compensation below the loss
- Surveillance that waits to be asked
- Holdings outside the reporting net
- Crime — concealment, illegal movement
Four states, with the evidence behind each one.
The question is narrow on purpose: how far are conditions from their own normal, and how have configurations like this behaved before? What to do about it stays with you.
σ is standard deviations — how far a value sits from what is normal for that unit at that time of year. So +2σ means unusually high for this place and this season, not high compared with somewhere else.
Conditions sit inside their historical range. Nothing to explain.
One or more indicators have left their normal range and stayed out.
Several indicators from different domains are moving together.
The configuration resembles historical periods that preceded comparable events.
Six families of evidence, kept by people with no stake in the answer.
A customs officer and a procurement clerk have no view on whether a country has a disease problem. That indifference is what makes their records hard to suppress.
Each mark is a record arriving · spacing is reporting cadence
Commodity flow, origin mix, consignment records.
Culling, disposal and disinfection tenders, dated to publication.
The commodity itself, sampled across many holders at once.
When a thing became knowable, against when it was notified.
Interdiction and seizure records. Incursion pressure at a border.
Density, habitat and seasonal conditions. The baseline hazard.
Measured against, never mixed in
Notification and case counts. Scored against, never an input.
Your own operational data, standardised the same way.
Regions, species and supply chains you define.
People whose decisions rest on a number someone else had a reason to keep low.
Exposure monitoring
Your book is priced off reported figures. We tell you how much confidence each one deserves.
Sourcing & supply
Your supplier's country reports a number. Know when the evidence stops supporting it.
Commercial planning
Geographic and species-level risk conditions, measured independently of what has been declared.
Situational awareness
An independent cross-domain view alongside existing surveillance, never replacing it.
The private beta opens in stages.
We're working with a small number of organisations who carry this risk operationally. If that's you, the queue is short.
Methodology noteIndicator definitions, the sources behind each one, backtest results and measured false-positive rates are set out in full — shared under a mutual NDA, on request, before anyone is asked to rely on a number.
No newsletter, no mailing list, no tracking pixels. Your message is read by one person and answered by hand.