How SiftScore Works in Practice

Every score on ViaReviews comes from a published formula, not from an editor's mood. That matters because the score is used to rank products, compare them and decide which ones look legitimate. A ranking you cannot reproduce is just an opinion with a number attached.

This article walks through the formula with a concrete example, explains how missing data and red flags change the result, and — just as importantly — explains what the score is not allowed to tell you. The live version, computed from the same code that produces stored scores, is on the methodology page.

The shape of the score

SiftScore is a weighted sum of five components. Each component is scored from 0 to 100, then multiplied by its weight:

• Value — 25%
• Trust & legitimacy — 25%
• Evidence — 20%
• Market reception — 15%
• Transparency — 15%

The total is rounded to one decimal place and stored on the product. When data changes, the score is recomputed at write time by the daily pipeline, a manual re-fetch or an import. There is no separate "editor score" hidden behind it.

The five components in detail

Value (25%)

Value asks whether the commercial offer is reasonable relative to its category, not whether it is cheap in absolute terms.

• Start at 50.
• If the price is within ±30% of the category median: +20. Within ±60%: +10. Otherwise: 0.
• Refund window: ≥180 days +30, ≥90 days +20, ≥60 days +10, >0 days +5, unknown 0.

The median is computed from published prices in the same category, ignoring products with no price. That means value is a relative signal: a $79 product can score well in a category where $79 is typical, and poorly where the median is $30. It also means a category with very few prices produces a shaky median, which is a limitation, not a feature.

Trust & legitimacy (25%)

Trust starts at 50 and rewards a documented refund policy:

• Refund policy documented: +20.
• Each red flag: −15.

The red flags are specific and documented: auto-ship / negative-option clauses, fake countdown timers, no refund policy, undisclosed subscriptions and exaggerated claims. The score does not infer a red flag from a bad feeling; it uses the flags recorded on the product.

This component is deliberately blunt. A product with two flags loses 30 points before any other evidence is considered. That is the point: commercial behavior matters.

Evidence (20%)

Evidence is different from the other components because it can be unknown. If no ingredient list is published, the component sits at neutral 50 rather than being punished as if the formula were bad.

When data exists, the component is built from 0 rather than 50:

• Ingredient list published: +20.
• Each named ingredient: +5, capped at +20.
• Each ingredient with a cited evidence link: +20, capped at +60.
• The component is clamped to 0–100.

This rewards disclosure and sourcing, not the number of ingredients on a label. A short, fully disclosed formula can score better than a long proprietary blend with no doses.

Market reception (15%)

Market reception starts at 50 and adds two independent signals:

• Marketplace gravity: ≥200 +25, ≥100 +15, ≥50 +10, >0 +5.
• Latest known rating: ≥4.5 +25, ≥4.0 +15, ≥3.0 +5.

Gravity is a network signal of relative sales activity, not proof of quality. A rating is only used when a documented snapshot exists; a missing rating adds nothing. The component is context, not a popularity contest.

Transparency (15%)

Transparency starts at 50 and rewards disclosure:

• Price documented: +20.
• Refund documented: +20.
• Description published: +10.
• Each red flag: −10.

A product can be transparent about a weak offer; transparency measures whether the facts are visible, not whether the offer is good. That distinction is why the components are separate.

The unknown-data rule

Missing data sits at a neutral 50 in Evidence and Market, and the other components simply receive no points for the missing item. A product with no verified data at all — no price, refund, description, ingredients, gravity or rating, and no flags — scores exactly 50.0. The product page says "unknown" instead of inventing a value.

This rule is a design choice with a cost: a product can look "average" partly because it is undocumented. The alternative — treating missing data as zero — would penalize products for our inability to verify them, which is worse. The page shows the breakdown so you can see which parts are neutral.

A worked example

Take a product with these captured values:

• price $49 in a category whose median is $50;
• refund window 60 days;
• one published ingredient with no cited study;
• no gravity or rating data;
• a published description;
• no red flags.

Now compute each component.

Value: base 50. Price is within ±30% of the median → +20. Refund 60 days → +10. Value = 80.

Trust: base 50. Refund documented → +20. No flags. Trust = 70.

Evidence: list published → 20. One named ingredient → +5. No cited evidence. Evidence = 25.

Market: no gravity, no rating → neutral 50.

Transparency: price +20, refund +20, description +10 → 100.

Total:

• Value: 0.25 × 80 = 20.0
• Trust: 0.25 × 70 = 17.5
• Evidence: 0.20 × 25 = 5.0
• Market: 0.15 × 50 = 7.5
• Transparency: 0.15 × 100 = 15.0

SiftScore = 65.0.

What one red flag does

The same product with one documented red flag — say, an auto-ship clause:

• Trust: 70 − 15 = 55.
• Transparency: 100 − 10 = 90.
• Value, Evidence and Market are unchanged.

The new total is 0.25×80 + 0.25×55 + 0.20×25 + 0.15×50 + 0.15×90 = 59.8.

That is a 5.2-point drop from one flag. A second flag lowers Trust to 40 and Transparency to 80, taking the total to 54.5. The legitimacy verdict also changes: two flags trigger avoid, while one flag is caution (and one flag with no documented refund is also avoid). The score and the verdict are related but not identical: the score is a weighted summary, the verdict is a rule-based label.

What changes the score most

• Missing ingredient data pulls Evidence to 50; a product with a published, sourced formula can gain up to 50 points there.
• A missing refund policy costs 20 in Trust and 20 in Transparency — 8 points of the total before any red flag.
• A red flag costs 15 in Trust and 10 in Transparency — 5.25 points.
• A high refund tier adds up to 30 in Value (7.5 points of the total).
• Market signals are capped at 50 points of upside and cannot rescue a weak commercial offer.

The weights are public and deliberately favor trust and value over popularity. A product cannot buy its way to a high score with gravity alone.

What SiftScore is not

• It is not a measure of efficacy. A score can be high for a well-documented product with weak evidence, and low for an undocumented product that might be fine. Read the ingredient evidence separately.
• It is not medical advice. A score does not tell you whether a product is safe for you, your medications or your health conditions.
• It is not a probability of success. 65 does not mean "65% chance this works". It is a relative comparison of documented attributes.
• It is not a review score. Customer testimonials are displayed with sources, but they do not become a laboratory result.
• It is not influenced by commissions. The affiliate link is tracked separately from the data; a higher commission does not raise the score.

If a product has a high score, the honest interpretation is: "this product discloses more of what we can measure". It is not a promise.

How to reproduce a score

Every product page shows the component breakdown. You can check it against the rules above:

1. Read the methodology page for the constants and a live example.
2. Open a product, for example KeySlim Drops, and compare its price, refund, ingredients, red flags and market data.
3. Recalculate each component with the tiers in this article.
4. Multiply by the weights and compare the total.

If your calculation differs, the data or the code is wrong — and we want to know. The verification process explains where each value comes from, and the contact page is the way to report a discrepancy.

FAQ

Why is a missing ingredient list neutral instead of zero?

Because missing data reflects our ability to verify, not necessarily the product's quality. Treating it as zero would punish products for our access limits. Neutral 50 is the compromise, and the breakdown shows it.

Can a product with no data score above 50?

Yes, if it has other documented attributes — a price, a refund window, a description or a rating. The "exactly 50" case is a product with no verified data at all.

Is a higher score always a better product for me?

No. The score measures documented commercial and transparency attributes. Your goals, health, medications and tolerance are separate. Use the score to compare disclosure and terms, not to self-diagnose.

Why do two products with the same price have different Value scores?

Value is relative to the category median and includes the refund window. The same price can be typical in one category and above the median in another; the refund tier also differs.

Do red flags affect the legitimacy verdict?

Yes. The verdict is rule-based: avoid for two or more flags (or one flag with no documented refund), caution for one flag or no refund, and legit only when there are no flags and a documented refund. The score and the verdict can disagree at the margin.

How often does a score change?

Whenever the underlying data changes: a new snapshot, a price update, a new refund term, an ingredient edit or a red flag. The daily pipeline recomputes scores after the data refresh.

Bottom line

SiftScore is a transparent comparison tool with public weights, a neutral rule for missing data, and explicit penalties for commercial red flags. It is useful precisely because it is limited: it does not pretend to measure efficacy or safety, and it shows its work.

Read the methodology, check the breakdown on any product page, and use the comparison tool to see how the numbers behave across products. A score is a starting point for a decision, not a substitute for one.

Last reviewed by the ViaReviews Research Team on 2026-11-06. This article is informational and is not medical advice.