Host Notes · research notebook · manual experiment designer

Design a research protocol before you measure anything

Record a question, a host and page, an observation window, a comparison, an intervention description, confounders and an evidence URL. Nothing is scored. No findings appear until you have measured them. The tool exports the protocol you typed, not a result.

This designer needs JavaScript to run. It has no server step and no form submission. With scripting disabled you can still read the guide on what the parasite SEO label hides, the pages on separating discovery, indexing and ranking, designing an interpretable comparison and keeping a reproducible evidence log. A printable protocol template can be kept on paper by copying the field labels below.

Evidence chain used by this notebook: no stage is skipped and no stage is reported as a finding until it has been measured.

Protocol designer

All fields are local. Nothing is stored, sent or saved automatically, and the designer makes no network request. Fields remain locked until the page finishes initialising the controls.

Protocol fields

Research question (required)

State one testable question. Avoid wording that presumes the answer.

Host / page (required)

URL of the host page under observation, including protocol.

Evidence URL (required)

A public, primary source or capture that supports the observation. Prefer first-party policy text or an archived measurement.

Observation window, in whole days (required)

Between 1 and 365 days. Record the end date in your log separately.

Observation start date (required)

Calendar date the window opens.

Comparison or control (required)

What is being compared with the host page, and why is that a fair contrast? Describe the counterfactual.

Intervention description (required)

State exactly what changed, when, and what did not change. Do not describe intent as if it were an effect.

Confounders (required)

List plausible alternative explanations: seasonality, site-wide changes, template edits, query mix, algorithm updates, measurement limits.

No field asks for a score, a ranking claim or an outcome. This designer produces a protocol, not a result.

Controls are locked until the designer finishes initialising.

No protocol built yet. Nothing here is a finding.

What this tool will not tell you

It does not score hosts, rank tactics, benchmark anything or predict outcomes. Google's March 2024 announcement introduced three spam policies — expired domain abuse, scaled content abuse and site reputation abuse — and stated that some third-party reputation or authority scores do not correspond to its own signals. Bing's Webmaster Guidelines describe crawl, index and ranking behaviour and state that SEO and generative-experience guidance do not guarantee rankings, grounding or citations. Read both before treating any label as a measurement.

Google Search Central: spam policies — policy definitions and illustrative factors; no prevalence or effect-size data.
Bing Webmaster Guidelines — crawl, index and visibility guidance; states no ranking-weight data and no guarantees.