A Like-for-Like Supplement Comparison Worksheet

Write the comparison question before collecting products

A worksheet becomes unreliable when its columns are chosen after seeing which product looks favorable. Begin with a narrow question, such as comparing the label transparency and one-time delivered cost of packages that declare the same named ingredient and measurement basis. Define the market, observation date, package format, currency convention, and required evidence. This protocol determines which products qualify and which fields will be collected. It should be saved before data entry so later exclusions or changes can be explained.

The phrase like for like does not mean that products are medically or functionally equivalent. It means the worksheet uses a consistent documentary basis. Two packages may share an ingredient name while differing in source description, other ingredients, format, certification, or serving definition. Those differences deserve columns rather than silent normalization. If the research question requires an interpretation beyond label facts, transaction facts, or verifiable documents, narrow the question or obtain the appropriate qualified review before publication.

Create a fixed evidence schema

A practical schema includes product and package identifiers, manufacturer and seller, label URL, label revision or access date, serving fields, verbatim ingredient wording, declared amount and unit, other-ingredient link, allergen statement, warnings as printed, batch-document availability, named certification and directory result, package price, mandatory fees, currency, and observation date. Each field should define its allowed values and source priority. Verbatim fields preserve evidence; normalized display fields make the table readable without replacing the original wording.

Use structured states for documentation: verified in named issuer directory on a date, manufacturer-provided document found, claim shown but not independently located, not found after stated search, not applicable, and unknown. These states are more informative than yes and no. A blank cell is ambiguous because it can mean no information, no research, or no feature. Require a reason when a field is unavailable. Keep source links at field or record level so a reviewer can trace a statement without searching an entire product page.

Normalize only what can be normalized safely

Formatting normalization can make identical facts easier to compare. Standardize capitalization, whitespace, currency display, dates, and compatible SI prefixes while retaining the original text. Transaction arithmetic can produce cost per labeled serving under one declared convention. It should not collapse different ingredient descriptions, extracts, organisms, compounds, or measurement systems into an invented equivalence. When the basis is uncertain, display the values in their original units and mark the normalized field unavailable.

Serving definitions also remain visible. One product may declare one tablet as a serving and another may declare two capsules. A cost-per-serving column can still be calculated, but its heading and note must explain that the labeled servings contain different declarations. Never use a visual score to hide this distinction. Color should communicate data status, such as verified, seller-claimed, or missing, rather than suggesting that more ingredient or lower price is inherently preferable. Accessible text labels must carry the same meaning as color.

Treat missing and conflicting data as findings

A manufacturer page, retailer page, package image, and certificate may disagree. Record the values, source dates, and package identifiers before deciding whether they describe different revisions. Establish a source-priority rule for each field, but do not automatically assume the manufacturer page matches stock sold by a retailer. A current package photograph with an identifiable code may be more relevant to that listing, while an official certification directory is the authority for whether a named certification is currently listed.

Missing data should not receive an average, neutral score, or favorable interpretation. Show missing and explain the search performed. If a critical inclusion field is missing, keep the row in an excluded-record appendix with the reason. This prevents quiet selection bias and gives manufacturers or readers a precise correction path. Corrections should require a direct source and preserve prior versions. A visible correction log strengthens the worksheet more than pretending that the first collection pass was complete.

Present comparison without manufacturing a winner

The final table should answer the stated question through fields and filters. It can sort by an objective transaction value, such as observed cost per labeled serving, provided the default and limitation are clear. It can group records by document status or package format. It should not award a general rank because the worksheet does not determine personal relevance, product performance, or health outcomes. A narrative summary can identify patterns such as frequent missing batch identifiers without converting those patterns into unsupported conclusions.

Publish the protocol, field definitions, evidence dates, source hierarchy, calculation formulas, excluded records, correction route, and update schedule. Have a second reviewer reproduce a sample of records. Automated validation should catch duplicate identifiers, incompatible units, broken source links, missing dates, and arithmetic errors. Finally, state the method's limits near the table rather than burying them in legal text. The reader should immediately understand what was compared, what was not compared, and why a missing value remains missing.

• Write and date the comparison protocol before data collection.
• Define the inclusion rule and evidence source for every field.
• Store verbatim label text beside normalized display text.
• Normalize only compatible units and clearly defined transaction values.
• Use explicit unknown, conflict, and not-applicable states.
• Retain excluded records with reasons.
• Publish formulas, evidence dates, and correction history.
• Use a second reviewer to reproduce sampled rows.