The field notebook · Proposed protocols
A method is worth testing when the decision, the measure and the risks are clear. No completed results are claimed here.
Protocol 01 · Not yet run
Does a structured recap help both sides identify the problem and agree the next decision?
Compare similar conversations using the current recap and a structured version. Record buyer corrections and confirmed next steps; avoid selecting only easy opportunities.
Buyer-confirmed problem, misunderstood commitments and time spent clarifying. Keep a record of missing or misleading information.
Stop if the format encourages assumptions to be presented as buyer-confirmed facts.
Protocol 02 · Not yet run
Can the reader identify the scope, client obligations and decision without another meeting?
Keep offer and commercial terms comparable. Where opportunities cannot be fairly compared, run a comprehension review rather than claim a conversion experiment.
Accurate understanding of scope and responsibilities, clarification rounds and explicit decisions. A faster rejection may be a useful result.
Reject a shorter format if important commitments disappear or misunderstandings increase.
Protocol 03 · Not yet run
Can assistance reduce total preparation effort without lowering factual accuracy?
Use similar public-source research tasks and a pre-defined rubric. Counterbalance order where practical; blind the reviewer to the method if feasible.
Total effort including verification; material errors, stale facts, unsupported inferences and usable sources. Assess useful questions separately from fluent prose.
Do not adopt a workflow that saves drafting time but fails the agreed accuracy threshold.
Adapted from the supplied manuscript’s experiment handbook. These are protocols, not evidence that the interventions work.
Read the evidence & experiments collection →The publisher’s field guide · Twelve proposed tests
Choose one question, an eligible cohort, a fixed window and a stop condition before collecting results. When traffic is too low for a fair comparison, observe real tasks and report descriptive counts rather than declare a winner.
Guide protocol 01 · Not yet run
Does a task-specific promise improve understanding of who the publication serves?
Measure: Audience and benefit comprehension; then confirmed subscription over a fixed window if volume permits.
Guardrail: Same destination and product; monitor immediate opt-out and useful action. No early stopping at a favourable result.
Full protocol · Guide page 73 ↗Guide protocol 02 · Not yet run
Does a task-based route help people find appropriate help and begin using it?
Measure: Successful task completion and beginning the intended tool, not clicks alone.
Guardrail: Keep content equivalent when testing navigation; distinguish a new content package from a navigation-only change.
Full protocol · Guide page 74 ↗Guide protocol 03 · Not yet run
Which offer leads to a useful first action, not merely more signups?
Measure: Confirmed signup and defined early use for equivalent eligible cohorts.
Guardrail: Inspect delivery and opt-outs. Different acquisition channels make the comparison observational.
Full protocol · Guide page 75 ↗Guide protocol 04 · Not yet run
Does an example and follow-up help a subscriber use the promised resource?
Measure: A defined use action in the same observation period after assignment.
Guardrail: Handle delivery failures explicitly; avoid duplicate messages, complaints and excess contact.
Full protocol · Guide page 76 ↗Guide protocol 05 · Not yet run
Does an annotated example improve the reader’s decision on a realistic task?
Measure: A preset task rubric, preferably scored without knowing the treatment.
Guardrail: Confidence and enjoyment are not accuracy. Label illustrative documents and preserve the same recommendation.
Full protocol · Guide page 77 ↗Guide protocol 06 · Not yet run
Can readers identify both what a finding supports and what it does not?
Measure: Correct interpretation and ability to locate the source.
Guardrail: Hold the claim constant. A small usability exercise cannot establish a quantified trust lift.
Full protocol · Guide page 78 ↗Guide protocol 07 · Not yet run
Do availability labels help readers find usable and relevant material?
Measure: Actually opening, playing or downloading prominent resources; successful retrieval tasks.
Guardrail: Keep unknown distinct from failed. Do not infer paid retention from retrieval success.
Full protocol · Guide page 79 ↗Guide protocol 08 · Not yet run
Does concrete wording generate relevant downstream action on the same issue?
Measure: Reliable action in a predefined window; repeat across issues.
Guardrail: Account for automated clicks, mismatch, complaints and opt-outs; opens are secondary diagnostics.
Full protocol · Guide page 80 ↗Guide protocol 09 · Not yet run
Can a narrowly scoped clinic meet a predeclared economic and usefulness threshold?
Measure: Purchases, attendance, task completion, refunds and total workload.
Guardrail: Include staff time and disclose enrolment/refund conditions. Payment does not prove sales effectiveness.
Full protocol · Guide page 81 ↗Guide protocol 10 · Not yet run
Does a relevant sharing invitation create subscribers who actually use the material?
Measure: Confirmed referred subscriptions and early-use quality.
Guardrail: Count reward cost, shared exposure and attribution ambiguity. Treat small pilots as exploratory.
Full protocol · Guide page 82 ↗Guide protocol 11 · Not yet run
Does a documented case provide useful transferable detail beyond an illustration?
Measure: Supported decisions and subsequent relevant use, alongside reporting cost.
Guardrail: Reporting changes several attributes; do not claim an isolated authenticity effect. Obtain permission and retain alternatives.
Full protocol · Guide page 83 ↗Guide protocol 12 · Not yet run
Does assistance reduce total effort to an acceptable standard?
Measure: Research, drafting, verification, editing, rework and failed assignments, using an agreed quality rubric.
Guardrail: Document configuration; blind review where feasible. Major errors cannot be averaged away by faster output.
Full protocol · Guide page 84 ↗