The Record · public status

Coverage & reliability.

These are dated operating numbers, not market estimates. The census, freshness clock, depth mix, and review state are generated from the private Record and published with their limits attached.

As of 2026-09-23 · data state data-e70453c369dbe72e

01 · census

The universe is a record too.

96 products across 147 families and 22 brands are known to the Record. The public number names only products at the approved depth and excludes crossover products.

CENSUS STATUS

PARTIAL, BLOCKED

Enumerated 2026-08-03. Public denominator: withheld until the census is complete.

PUBLIC-ELIGIBLE FAMILIES

97 / withheld

Format variants remain visible in the Record, but family counts prevent padding the headline.

Limit: First-party DTC enumeration across 19 covered brands' own pre-workout collections. 2 brand shelves could not be enumerated. The six retailer shelves in Coverage Strategy V1 section 3 remain disabled for want of verified rights rows, so this is not a retail shelf census and the enumerated-universe denominator remains unknown. No coverage percentage may be derived from these counts.

02 · operating clocks

Breadth first. Depth when it matters.

The offer loop is cheap and calendar-driven. The label loop is expensive and trigger-driven. Missing observations stay in the denominator.

OBSERVATION CURRENTNESS

141 / 147

0 aging, 0 stale, 5 suspended because their source could not be read, 1 never observed.

LAST MONITORING RUN

2026-09-23

1 of 6 products yielded a valid observation; 5 checks did not complete.

HASH BASELINE

145 / 147

Offer observations with both preserved page and claims hashes. Missing baselines stay visible.

HISTORICAL SWEEPS

27 / 147

12 offer states recovered; 0 formula states recovered. A sweep is not formula history.

FORMULA HISTORY

8 / 147

18 states have usable components in a comparable pair; 246 dated states are metadata-only.

CLAIMS + CERTIFICATIONS

0 / 147

Products with two dated captures carrying explicitly complete claims and certification inventories.

PUBLISHED SIGNALS

3 / 3

Reviewed change events that pass the publication gate, over reviewed change events in the Record. This is not a market rate.

REVIEW PASS RATE

0 / 31

Confirmed decisions divided by resolved submissions in this review ledger. This is not the share of all published Signals or a measure of product quality; confirmed events elsewhere in the Record are counted separately above.

AWAITING REVIEW

114

Unresolved evidence candidates, across the review ledger, the findings a monitoring run opened, and the historical packets. 1 more remain WATCHING, which is not a failure. These candidates may still need source acquisition or identity work before they can be reviewed for publication.

AUTOMATION PLAN

31 overdue

31 due, 37 planned, 0 blocked; oldest waiting 53 days. This is planned work, not run state or evidence.

ACQUISITION QUEUE

28 / 147

Covered products with planned acquisition work outstanding. The other 119 need no research task today; every one of them is still watched on its own cadence, which the currentness card above counts.

EVIDENCE CEILINGS

Open across 60 products, every one carrying the condition that would lift it. A ceiling is a statement about the evidence, never about the answer.

03 · evidence depth

The Record holds every stage.

Public analysis begins at NORMALIZED. Registration is retained because a permanent dossier is more honest than pretending an incomplete product was never observed.

Depth	Products	Meaning
REGISTERED	34	Identity and permitted first observation; not a public formula claim.
NORMALIZED	113	Current printed panel normalized against the ingredient record.
LINEAGE	0	More than one identity-resolved dated state.
REPORT_READY	0	Passed the full report release gates.

Strategy 1 · coverage-strategy-v1. Publication requires the applicable evidence and review gates. Agent review is not human review.

04 · for developers

Two different kinds of unhealthy.

The API being up and a manufacturer’s page being readable are separate facts, and only one of them is ours.

Question	Where it is answered	What a failure means
Is the API serving?	/api/v1/health, no key needed	A non-200 is our problem. Retry with backoff and quote the request id.
Is the MCP endpoint serving?	POST /mcp with tools/list	Same. It runs in the same Worker and shares its health.
Can we still read this product’s source?	Inside the answer: `status`, `confidence` and `limitations`	Not an outage. The Record holds a dated state and says it has stopped being able to refresh it.
How current is the Record overall?	This page, and `get_record_version`	A stale count moving is research state, not availability.

A source that stops answering halves that product’s monitoring interval rather than stopping it. Developer documentation is at formulasignal.com/developers.