Source Citation Policy

1. Scope

This policy applies to every numeric, factual, or quoted claim published on https://nutrogine.com — calorie figures, macro values, portion weights, app accuracy claims, pricing, revenue, headcount, and historical events. It does not apply to opinions, predictions, or framings, which are clearly labelled as such.

2. The four Source Badges

Each badge is clickable. The destination is the primary source (USDA FDC entry, brand PDF, Reddit thread, etc.) — not a paraphrase on a third-party site. The friendly explanation of each tier lives on the methodology page; this policy defines the formal rules for using them.

3. Precedence (most to least authoritative)

When more than one source is available for a given numeric value, we publish the higher-precedence number and disclose any meaningful conflict inline. Precedence order:

1. ●USDA verified USDA Foundation Foods. Lab analyses performed under USDA quality controls on specific, well-defined food items. ~395 entries. Highest authority for ingredient-level values (e.g., grams of protein in 100g of grilled chicken breast).
2. ●USDA verified USDA FNDDS. Composite dishes matched to dietary survey data. ~5,400 entries. Used for common composite foods where Foundation Foods lacks coverage (chicken burrito with rice and beans, latte with whole milk).
3. ●USDA verified USDA SR Legacy. The older Standard Reference dataset. ~7,800 entries. Used when newer USDA datasets do not cover the item. Slightly less current; flagged as "SR Legacy" in the badge metadata when used.
4. ●Brand-claimed Brand-claimed. The chain's own published nutrition (PDF, calculator, or website). Authoritative for the brand's idealised default portion only. Not authoritative for actual served portions, which vary by store, employee, and time of day.
5. ●User-reported User-reported. Aggregated first-hand reports from Reddit, Yelp, Lemon8, TikTok, and similar public-comment platforms. Used to surface portion variance, actual-vs-advertised macro deviations, and real-world serving outcomes. Subject to the sample-size threshold below.
6. ●Estimated Estimated. Computed from the cross-references of two or more higher-precedence sources when no single source covers the dish directly. Used sparingly; flagged on the page.

4. Exceptions to precedence

Brand-claimed values take precedence over USDA values when:

• The dish is a proprietary recipe (e.g., a specific Chipotle marinade, a Starbucks proprietary syrup). USDA cannot authoritatively measure something the public dataset does not model.
• The brand has published a recent (within 24 months) third-party lab analysis of the specific menu item, and we have access to the analysis details.

User-reported values take precedence over brand-claimed values when, and only when:

• We are reporting portion variance rather than nominal composition (i.e., "what's actually in your bowl" rather than "what the brand says is in a default bowl").
• The user-report sample meets the threshold in §5 below.
• The brand value is more than 24 months old without restatement.

5. User-report sample threshold

We do not publish a User-reported value as a primary citation unless it meets all of:

• Sample size: at least 10 distinct first-hand reports (one report per author; duplicate reposts are not counted).
• Platform diversity: reports come from at least 3 different platforms (e.g., Reddit + Yelp + TikTok). A 50-report Reddit thread alone does not meet the threshold.
• Time window: reports must fall within the most recent 24 months. Older reports are noted and weighted lower.
• Quantitative content: reports must contain a numeric measurement (weight in oz/g, calorie count from a tracker, etc.) — anecdotes without numbers are noted but not aggregated.

When a value misses the threshold but is editorially relevant, we publish it with explicit caveats ("based on N=4 reports; treat as directional only") and do not assign it a User-reported badge.

6. Conflict resolution

When two sources disagree on a value, we apply this procedure:

1. Higher-precedence source wins by default.
2. Conflict is disclosed inline in the affected piece. If brand-claimed says 660 cal and Wells Fargo's audit of 75 bowls suggests 410–820 cal, both numbers appear; we do not quietly suppress the lower-precedence figure.
3. Magnitude is reported. A conflict is flagged as "minor" (within ±5% of the higher-precedence value), "material" (5–15%), or "substantial" (>15%). Substantial conflicts trigger a paragraph of analysis explaining the divergence.
4. Resolution attempt. Where possible, we contact the brand for clarification. We disclose whether a response was received.

7. Confidence thresholds

We do not publish a number unless we can attach at least one source from tiers 1–4 (USDA, brand, or user-report-meeting-threshold). If only Estimated material is available and the editorial topic requires a number, we either:

• Publish the Estimated value with the explicit badge and inline methodology paragraph; or
• Decline to publish a single number and report a range.

We never publish a number marked as authoritative when it is actually the result of educated guessing.

8. Citation format

• Every numeric value in body prose has either a Source Badge immediately after it or an inline citation link.
• Tables of values use a Source column or a per-row badge.
• Where multiple sources support one value (e.g., USDA confirms brand-claimed), the badge stack shows both. Click-through goes to the highest-precedence source by default; the others are listed in the page's Sources section.
• Direct quotes are linked to the original publication or post. We do not paraphrase a quote and present it as verbatim.

9. Source freshness and re-validation

• USDA: we re-pull the FoodData Central dataset quarterly. Each USDA-badged number carries an FDC ID linking to the live record.
• Brand-claimed: we re-check the brand's published nutrition page on every page review (target cadence: twice a year) and immediately when a brand publicly announces a recipe change.
• User-reported: aggregations are timestamped at publication. Pieces older than 12 months trigger a re-aggregation if the value is contested by a reader correction.
• Stale flag: any page older than 6 months without a fresh review carries a "Last updated" date that is noticeably old. Readers are explicitly invited to flag discrepancies in older pieces.

10. What we do not do

• We do not synthesise from secondary aggregators. We do not cite other calorie-tracking sites or apps as primary sources, even when they are correct, because we cannot verify their methodology end-to-end. (We may cite them descriptively as "another aggregator reports X" with explicit framing.)
• We do not bulk-import third-party datasets. Every number is added by hand, with the source-of-record verified at import time. This is slow; it is also what makes the Source Badge claim defensible.
• We do not silently update numbers. When a value changes, the page is re-dated and (for material changes) gets a correction note.
• We do not synthesise customer reports we did not read. We are open to doing AI-assisted aggregation when the underlying posts are already in our reading queue, but the editorial conclusion attaches to a human who has read the underlying material.

11. Reporting a citation problem

If you find a number on Nutrogine whose Source Badge is wrong, whose linked source does not support the claim, or which conflicts with a primary source we should have cited:

• Email [email protected] with the URL and the issue.
• Or DM @AlecZakhary on Twitter.
• We acknowledge within 48 hours and correct verifiable errors within 7 days, per our Editorial Policy.

12. Related policies

• Editorial Policy — who writes, AI disclosure, conflicts of interest, corrections process
• Methodology — friendly explainer of the four Source Badges
• About Alec — the founder and sole editor