How these numbers are calculated

Published 2026-07-25 · Updated 2026-07-25

This site does two calculations. It divides a current Amazon price by the number of servings in the package to get price per serving, and it compares that figure to the median of similar products to get a value score from 0 to 100. Everything below documents exactly how each step works, including what gets excluded and when a number is withheld rather than estimated.

Where the prices come from

Prices come from the Amazon Creators API. Each record carries the price, the product title, the brand, the listing attributes, and the Amazon detail page URL. Nothing is scraped, and no price is typed in by hand. Every price displayed on the site carries an “as of” timestamp showing when it was captured.

How often prices refresh, and what happens when they do not

A scheduled job refreshes the full catalog every 12 hours. Freshness is then evaluated on every single page request rather than at build time, so a cached page can never outlive its data:

• Under 18 hours old: prices display normally with their capture timestamp.
• Between 18 and 24 hours: prices still display, and the page marks the data as ageing.
• Past 24 hours: every price table is replaced with a link-only product list. No price is shown at all.

No price history is retained or published. Each refresh replaces the previous snapshot rather than appending to it.

How price per serving is calculated

Price per serving is the current price divided by the total servings in the package, where total servings is the servings per container multiplied by the pack count. A two-pack of 30-serving tubs is treated as 60 servings, not 30. The result is rounded to four decimal places. When the serving count cannot be determined at all, price per serving is left empty rather than guessed.

How serving counts are read from listings

Serving counts are parsed from the listing title and its attribute text. The parser records how much it trusts each result, and that confidence controls what the site is willing to publish:

• High confidence. The listing states servings explicitly, as in “60 servings” or “60 servings per container”. Only these products are eligible for a value score.
• Low confidence. The listing states a unit count instead, such as capsules, tablets, gummies, softgels, scoops, or a plain count. The site treats the count as servings so the product can still be listed, but withholds the value score, because one serving is often several capsules.
• Unparsed. Neither pattern appears. The product has no price per serving and no value score.

Pack multipliers are detected separately, from phrasings like “pack of 3”, “3-pack”, and “3 pack”, and are applied before the division.

How the value score is calculated

The value score answers one question: is this product cheap or expensive compared to products like it? It is computed in four steps.

1. Build a cohort. Products in the same category and the same form (powder, capsule, gummy, liquid, drink) are grouped together. If that group has fewer than 5 members, the whole category is used instead. If the category still has fewer than 5, no score is produced.
2. Take the median. The median price per serving of the cohort becomes the baseline. A median is used rather than an average so a single unusual listing cannot drag the baseline around.
3. Measure the distance. The product is expressed as a fraction cheaper or more expensive than that median, then capped at 50 percent in each direction.
4. Map to 0 to 100. A score of 50 is exactly the cohort median. Each point above 50 is one percent cheaper than typical, and each point below is one percent more expensive, so the scale runs from 0 to 100.

Only products with a high-confidence serving count enter a cohort or receive a score. Products with low-confidence counts are still listed and still sortable by price, but they do not influence anyone else’s score.

What gets excluded from the catalog

Three filters run before a listing is eligible to appear anywhere on the site:

• Prescription drug name claims. Listings whose titles use GLP-1 medication brand or molecule names as marketing claims are removed, along with constructions like “nature’s” plus a drug name.
• Manual blocklist. Specific products can be excluded by ASIN when review finds a listing is miscategorised or misleading.
• Implausible per-serving values. Each of the 8 categories has an upper and lower bound on price per serving. Values outside the bounds almost always indicate a mislabelled bulk listing or a sample size rather than a real bargain, so they are dropped rather than published.

What this methodology does not do

• Price per serving is a cost normalization, not a quality score. A cheaper product is not necessarily a better one.
• No ranking on this site reflects efficacy, ingredient quality, third-party testing, or manufacturing standards. None of those are measured.
• Rankings are computed purely from price per serving and are not influenced by affiliate commissions.
• Serving sizes are taken from the listing as written. A product that states a large serving will look more expensive per serving than one that states a small serving, even for the same amount of powder. Read the label.

How this page is maintained

The thresholds described here are read directly from the same code that runs the site, so this page cannot quietly drift out of date when a number changes. It is reviewed whenever the pricing or scoring logic changes, and the “updated” date at the top reflects the last such review. The catalog currently spans 8 categories and is paired with 9 buying guides. For what the site is and who it serves, see About.