Glossary — How This Works

This dashboard uses several classification systems side by side — a star-rating split, an AI sentiment score, safety/authenticity flags, complaint severity, action priority, and momentum. They measure different things on purpose and are not required to agree with each other. This page defines each one and where you'll see it.

Star-rating metrics (per product)

The plain Amazon star split — no AI involved. Positive + Neutral + Negative always add to ~100%.

Positive
% of reviews rated 4★ or 5★.
Neutral
% of reviews rated 3★.
Negative
% of reviews rated 1★ or 2★.
Avg Rating
The plain average of all star ratings for the product.

Sentiment Score & label

A separate AI read of what review text actually says — tone, emotion, how strongly customers feel — scored 0–100. It's a different lens from the star split above and can land higher or lower than it: lukewarm 5★ praise scores lower, a 2★ review with one minor complaint can score higher.

Sentiment Score
0 (all-negative tone) to 100 (all-positive tone), from AI analysis of review text.
Positive / Mixed / Negative label
Derived from Sentiment Score only: 70+ = Positive, 45–69 = Mixed, under 45 = Negative. Not derived from the star-rating %s above.

Safety & Authenticity Watch (product-level flags)

Set when at least one Amazon review raises a self-reported issue — not a confirmed incident, lab finding, or decision already made. Surfaced for your team to triage.

Safety flag
At least one review describes a possible physical reaction or injury (e.g. rash, burn, allergic reaction).
Authenticity flag
At least one review describes a possible marketplace-integrity issue (e.g. unsealed/tampered packaging, formula mismatch, suspected counterfeit).
Isolated report / Recurring pattern
A confidence indicator next to each Watch write-up, separate from its tone: "Isolated report" means one customer account; "Recurring pattern" means multiple independent reports of the same or a similar issue.
Added because the write-up language (e.g. "immediate recommendation to halt sales") is the same urgent register regardless of whether it's evidence from 1 review or 15 — this badge is what actually tells the two apart.

Complaint severity (per feedback bullet, on product pages)

Each individual complaint under "Customer Feedback" is tagged with one of four categories, from most to least serious:

Safety
Physical reaction or injury risk (rash, burn, allergic reaction).
Authenticity
Marketplace integrity — tampering, counterfeits, unsealed product.
Product Failure
The product itself malfunctioning or underperforming (e.g. pump breaks, formula doesn't tan).
Experience
Everything else — packaging preference, scent, price, shipping speed.

Action item priority & team

Every action item is assigned an urgency and the team best placed to act on it.

HIGH / MEDIUM / LOW
How urgently the item should be worked, independent of complaint severity above — a LOW item can still originate from a Safety complaint if the fix itself isn't time-critical.
Formulation & Safety
Ingredient/formula changes, allergen investigation, dermatological review.
Product & Packaging
Dispenser/pump mechanisms, bottle design, shipping protection.
Marketing & Positioning
Claims accuracy, product titles/descriptions, category placement.
Customer Service
Response protocols, refund/replacement handling, review monitoring.

Momentum

Improving / Declining / Stable
Compares the average rating in two ~90-day windows, as of when this report was generated — catches a product trending up or down before it shows up in the all-time average. Shown as an absolute date range, not "last 90 days": this report is regenerated periodically, not live, so a relative label would go stale.
Requires 5+ reviews in each of the two windows; below that, shown as "insufficient data" instead of a direction, since a couple of reviewers can swing a thin window.
Category Trend (Brand Insights)
The same idea applied to a whole product category, in 6-month buckets instead of 90-day windows (categories rarely have enough monthly volume for a stable shorter read).

Catalog

External Brand
A third-party/competitor product (e.g. Podz Eyewear, Ed Hardy) kept in the dashboard for benchmarking. Excluded from every Devoted Creations brand-performance number so it can't drag down or inflate "our" figures.
DHA Level
1–5 scale for self-tanning intensity: 1 = lightest, 5 = darkest concentration of DHA, the active self-tanning ingredient.