Ranch.Bot
Sign In
Protocol registered before observation

How Ranch.Bot record keeping will be observed on approximately 200 ewes as of August 23, 2026

This protocol fixes the units, denominators, exclusions, timing, and failure rules before a two-month observation on a working Alberta sheep operation. No result has been collected or drafted for this page.

Fixed observation boundary

Observation window
October 1 through November 30, 2026 (Mountain Time)
Report window
December 15 through 31, 2026
Operator
Jerremy Lewis, the developer behind Ranch.Bot and the flock operator. He is the only enrolled operator.
Eligible work
Consecutive real sheep events the operator personally observes or performs and logs on the independent source record before the first Ranch.Bot input.
Excluded work
Tests, demonstrations, retrospective cleanup, imports, non-sheep records, and events observed or entered only by another person.

This is an on-farm observation by Ranch.Bot's developer, not a customer case study or controlled trial. The same person owns the product and operates the flock. The final report will repeat that conflict and state what the results cannot support.

One event stays linked through every stage

Event, attempt, proposal, review, saved-record, and export measures stay separate. A retry cannot disappear inside a saved-record success rate.

Eligible real-world event
A real sheep event the enrolled operator personally observes or performs, would normally preserve, and logs independently before the first Ranch.Bot input.
Input attempt
One SMS intended to capture one linked eligible event. A retry or rewritten note is another attempt.
Proposal and review
A structured write proposal and the inspection that ends in confirm, cancel, or cancel followed by a corrected input.
Saved and exported record
The persisted record after confirmation, then its representation in the end-of-window JSON and simplified CSV exports.

The product cannot grade itself

A consecutively numbered paper note starts before each first input. Daily page photos receive SHA-256 hashes in a private append-only ledger. After the window, a reviewer who did not perform the input checks each disagreement against that note and the product artifacts. Unresolved evidence stays unresolved and remains in the denominator.

Missing product output after an eligible input is a failure or missing outcome. It is not an exclusion. A missing source note is an exclusion under the rule fixed above and will be counted with its reason.

Current Ranch.Bot conversation screen showing a synthetic movement note and the review controls
Pre-window product example with synthetic movement data. It verifies the current proposal and review controls, not a field observation event or result.

Counts keep their own denominators

Event
Capture, first-attempt success, retries, abandonment
Attempt
Clarification and proposal yield
Proposal and review
Identifier match, required fields, correction
Saved record
Save completion, required fields, unintended duplicates
Retrieval and export
Lookup success and JSON/CSV reconciliation
Timing
Capture, proposal latency, review, retrieval, and export

Every metric will show its numerator, denominator, missing count, and product-version stratum. Timing reports count, median, minimum, maximum, and interquartile range. Material workflow releases split the results instead of being pooled.

Failures remain in the report

Successful, corrected, abandoned, failed, and unresolved examples stay in the private study archive. A failure may have more than one class.

  • Capture or carrier failure
  • Delayed or missing reply
  • Clarification did not resolve
  • Wrong event type, animal, or group
  • Missing, added, or changed field
  • Review or confirmation failure
  • Unintended duplicate
  • Retrieval or export failure
  • Operator error or unresolved evidence

Aggregate release, private raw records

Raw notes, messages, animal identifiers, internal IDs, phone numbers, precise routine timestamps, and row-level exports remain private. The public release will contain aggregate metrics, a data dictionary, and up to three redacted traces selected without excluding failures.

The observation cannot establish improved lamb survival, conception, health, profitability, productivity, safety, veterinary outcomes, or universal time savings. If privacy review or reconciliation is incomplete, the report will be delayed and the reason will be stated.

Immutable protocol v1.0.0

The versioned file will not be overwritten. A material change gets a new file and hash. A post-start change is reported as a dated deviation. Typographical fixes use errata that preserve the original bytes.

SHA-256
74b2c651fa53c4c309dcb42320b318f0c8c1e2f48737f33d7c4da9311596fac7