# Downstream work supplement v1.0.0 Publication date: 2026-09-07. Methods only; no results collected or drafted. Canonical page: https://ranch.bot/field-report/200-ewe-record-keeping ## Relationship to the original protocol This additive supplement accompanies [protocol v1.0.0](protocol-v1.0.0.md), finalized 2026-08-25, SHA-256 `74b2c651fa53c4c309dcb42320b318f0c8c1e2f48737f33d7c4da9311596fac7`. It does not replace or alter that file's bytes, hash, population, eligibility, observation window, scheduled retrieval checks, primary metrics, or planned report window. The original October 1 through November 30, 2026 observation and all its exclusions still apply. The only enrolled operator remains the operator defined in the original protocol. These additional observations ask what happens when saved information is needed for ordinary work. They cannot establish causal effects, universal savings, customer validation, reconciled inventory from generic movement notes, or improved livestock or financial outcomes. The operator also owns and develops Ranch.Bot; that conflict applies to this supplement too. ## Downstream task and private linkage A `downstream_task` is a real need to retrieve, assemble, or communicate information from one or more eligible recorded events: for example, finding a movement date or assembling the recorded history to answer another person's question. Log a task even if its intended record was never saved, cannot be found, lacks the needed detail, or the task is abandoned. Use private study IDs to link each task to the original eligible event IDs, attempts, reviews, and saved records where they exist. A task can link to several events; an event can support several tasks. Missing links have a reason, never an invented record ID. Mixed tasks mark the eligible portion separately; if its effort cannot be separated, label the comparison unknown. Tasks involving only excluded events do not enter the cohort; count them separately with an exclusion reason. Neither a new task nor a missing record expands the original event population. For each task retain its consecutive ID, date, trigger, intended question or deliverable, linked event IDs, product version, steps attempted, review help, outcome, evidence source, active effort, separate delays, and reason for any missing value. Keep exact identifiers, questions, timestamps, recipients, and raw links private. ## Previous method and evidence strength Ask what the operator actually used for the last comparable task: notebook, spreadsheet, messages, memory, another person's help, or another method. Record whether that task was performed, avoided, or unnecessary. Do not assume a retyping step or assign time to work the producer did not perform. A task previously avoided that becomes practical is newly useful work, not measured time saved. An unnecessary task has no assumed benefit. Label every timing and outcome source as observed ordinary use, participant recollection, or unknown. Demonstrations receive a separate demonstration label. Recollection may explain the previous method but recalled estimates cannot enter measured savings, be filled into missing stopwatch values, or be pooled with observed timings. Quantitative comparisons require observed comparable tasks, with the same information scope, completion standard, and accounted steps. Record differences in event count, complexity, operator assistance, and product version; show unmatched tasks separately. Report pair counts, individual differences and descriptive summaries, missing values, failures, and limitations. Do not infer causation from the comparison or extrapolate to annual savings. ## Effort accounting Record active person-seconds for capture, navigation, review, clarification, correction, retrieval, report preparation, and operational duplicate bookkeeping. Include unsuccessful attempts, retries, abandoned work, and time spent helping the operator review. A producer choosing to keep a notebook alongside Ranch.Bot during evaluation incurs real extra effort. Count that operational duplication even if it was chosen for reassurance. Use one private effort ledger with an effort ID, category, performer role, observed duration or unknown, and linked event/task IDs. Capture and review shared across later tasks are counted once at the event level. Task totals contain only additional task effort; period totals add event effort once plus task effort once. Do not charge the same capture/review seconds to each later lookup, or omit them when discussing the overall burden. Overlapping activity by one person cannot count twice; help by another person is separately identified person-time. Report setup, onboarding, and import effort once at the period level, separate from ordinary operation. Recording that effort does not make imported events eligible. Keep system waiting, deferred-review delay, and unrelated interruptions in separate fields from active effort. Mark incomplete timing as missing or partially observed; never replace it with zero or an estimate. Preserve the original protocol's timing definitions for its primary metrics. Research overhead is a separate ledger category: independent source logging, photographing and hashing source pages, worksheet administration, adjudication, and researcher-scheduled retrieval/export checks. Mandatory research source notes are not operational duplicate bookkeeping. If an activity serves both purposes, document which incremental portion belongs to each; if inseparable, report it as mixed/unknown without counting it twice. Report research burden visibly, outside ordinary operational effort and any measured operational difference. ## Consecutive observation and missing days Log naturally occurring tasks consecutively throughout the unchanged observation window, including failures, abandonment, missing records, and tasks with no useful answer. Scheduled research checks remain in the original protocol and the research ledger; they are not natural use or activation evidence. Complete a daily log with date, observation status, task count, and reason for gaps. A completed zero-task day explicitly states that no natural task occurred. A missing or incomplete daily log is missing observation data, not zero tasks and not evidence of no meaningful use. Report completed days with tasks, completed zero-task days, incomplete days, and missing days separately. ## Outcomes and reporting Report these outcomes separately, with task counts, denominators, evidence labels and unknowns: - Work reduced: an observed comparable task required less active operational effort. - Work unchanged: an observed comparable task required the same active effort at the timing resolution used. - Newly useful tasks: a previously avoided task became practical; no assumed time savings. - Added burden: additional active operational work, including review help and duplicate books. - No meaningful use: an observed attempt yielded no useful information or deliverable; describe failure or abandonment. Completed zero-task periods are reported explicitly, not as failures. - Unknown outcomes: missing evidence, incomparable tasks, or incomplete timing prevents a conclusion. Separate task utility from effort: a useful task can also add burden. If more than one category applies, report the overlap and do not sum those categories as exclusive totals. Use exact observed seconds and describe the instrument's resolution; no post-hoc threshold for a favorable result. Preserve failed attempts in descriptive effort summaries and show incomplete timings separately. Report period setup and shared event effort alongside task effort so a faster lookup cannot conceal a larger overall burden. Producer walkthroughs are separate from the enrolled field population. Keep demonstration, participant-report, and observed-ordinary-use evidence distinct, with separate denominators. A narrated demo or successful scheduled check is not evidence of unassisted ordinary use. ## Privacy and versioning Only the blank worksheet, methods, and permitted sanitized summaries may be committed. Completed worksheets, source records, effort ledgers, and raw linkage data remain private. Any summary or trace release follows the original protocol's consent, redaction, reconciliation, and review rules. No results are drafted before the observation closes. This version's bytes and SHA-256 companion are immutable after publication. A subsequent material change receives a new versioned file, actual publication date, hash, and explanation; changes after observation begins are also dated deviations. The original protocol remains available.