1. Define one problem
Each guide addresses a specific combination of surface, ordinary soil, and user-level task. We separate similar-looking materials when their care limits differ. A ceramic-glass cooktop is not treated like window glass; cement grout is not treated like glazed tile; a washable metal filter is not treated like a charcoal filter.
2. Set the light-maintenance boundary
Before research begins, we identify conditions that make the task unsuitable for a general household guide. Cracks, loose parts, active leaks, electrical exposure, gas work, structural damage, major mold or contamination, hazardous materials, and unknown chemical residues are common stop points. The article must say when to use manufacturer support or a qualified local professional.
3. Prefer primary sources
Research starts with the current owner manual, product label, manufacturer care page, government safety guidance, or recognized standards organization. We may use a reliable secondary source to locate terminology or an original document, but it does not displace a more relevant first-party instruction.
Sources are checked for the exact claim they support, not merely added to make a page appear researched. An accessed date records when the linked material was opened for research; it does not by itself claim that Anna completed the page's final editorial review, and it is not a promise that the source publisher will never revise or move the page.
4. Resolve applicability and conflicts
We record the material, finish, coating, model, cleaner concentration, contact time, and tool limits described by each source. If two manufacturers give different directions, the article preserves the difference instead of blending the methods. The instruction for the reader's exact product wins.
A current safety notice, product label, owner manual, warranty instruction, and applicable local requirement take priority over Clean Home Lab. Stop when the product cannot be identified or the instructions conflict.
5. Draft the smallest useful method
The first method is normally the least abrasive and least chemically complex option supported by the sources. A complete guide names suitable surfaces, exclusions, a spot test, supplies, ventilation or protective measures, ordered steps, rinsing, stop conditions, maintenance cadence, and common questions. We do not recommend mixing cleaning products.
6. Apply the evidence label
Source reviewed is a publication-stage label applied only after Anna has checked the cited primary materials and edited the article against this methodology. It does not mean the method was personally performed or scientifically tested by the site. Field verified is reserved for a specific procedure supported by retained notes, conditions, images, and results. If those records do not exist, the page cannot use that label.
What a field record must contain
A hands-on note is useful only when another reader can understand what was actually done. Before a page can describe a procedure as field verified, Anna records the exact material or model, starting condition, product and dilution, tools, sequence, contact or drying time, relevant room conditions, observations, limitations, and the date of the work. Before-and-after images are tied to that same record and are labelled as observations, not a guarantee that another surface will respond identically.
A record that is missing the model, changes several products at once, or shows only a representative photograph stays source reviewed. We never infer a result from a stock image, a product description, or a reader's unverified comment. This keeps a useful experiment separate from a general instruction and gives corrections a concrete record to inspect.
7. Review text, visuals, and claims
Before publication, Anna checks the final page for unsupported performance claims, false precision, unsafe chemical combinations, scope drift, and citation mismatch. Images must have documented provenance recorded in the image licensing record. Licensed or representative photography can illustrate a topic, but it cannot be described as Anna's work or used as before-and-after evidence.
AI-assisted tools may support outlining, consistency checks, and copy editing. They are not accepted as sources and cannot supply experience, test data, quotations, credentials, or the final review. A page cannot pass the human release gate until Anna has opened the cited documents and approved it for publication.
8. Publish, monitor, and correct
Publication dates and substantive update dates are recorded separately where the content model permits. We revisit a guide when a source changes, a link breaks, a manufacturer issues new instructions, or a reader raises a supported concern. Material errors follow the public corrections process; quiet spelling or formatting repairs do not alter the evidence status.
What the method cannot do
Source review cannot inspect a home, identify an unknown substance from a photograph, confirm hidden construction, predict an allergy or material reaction, or guarantee a result. A spot test reduces uncertainty but does not eliminate it. The disclaimer explains these limits in more detail.
Methodology last updated: 3 September 2026.