1. Brand Consistency (Flavor Similarity)
Brand DNA
The badge on a brand page compares the flavor fingerprints of all its expressions. The closer the fingerprints, the higher the percentage. It needs at least two expressions with flavor data.
85% and higher: Reliable Core Portfolio
A clear house style. If you like one bottle, you will probably like the rest.
70–84%: Consistent Brand DNA
Recognizable style with some room for variation between expressions.
55–69%: Diversified Selection
A broad range. Check the individual bottles before assuming one represents the brand.
Below 55%: High-Variance Specialist
Expressions differ strongly, often small batches or experiments.
A low percentage is not a quality verdict: it describes how varied the range is. The Deviation label on the card says the same in words (Low, Moderate, High, Very High).
The radar on the card shows the house style: the average share of each flavor family across all expressions of the brand. Below it, Most typical is the bottle closest to that house style and Most atypical the one furthest from it, each with its match percentage.
2. Data Confidence
The Data confidence bar does not show flavor. It shows how well the brand's data is supported by reviews. For how flavors are grouped, see Master Flavor Data.
Decoding the Colors
The color of the bar follows the percentage next to it.
- Deep blue (80% and higher): Well supported. Many reviews, many from our community, and reviewers agree.
- Light blue (60–79%): Fairly supported. Reliable enough, but a few more reviews would help.
- Amber (40–59%): Limited data. Few reviews or few real reviewers, so the picture can still change.
- Red (below 40%): Thin data. Very few reviews, mostly automated data, or reviewers who strongly disagree.
Why a bigger range needs more reviews
The score is measured per bottle: roughly 20 reviews and 5 from the community for each reviewed expression give full marks. A brand with one bottle therefore reaches deep blue much sooner than a brand with seven. Automated reviews alone can never make a brand stable: without community reviews the score stops at 60%.
Your Role in the DNA Loop
When you see an amber or red bar, the brand needs more reviews from the community. A review from you directly raises the score.
Community reviews are worth up to 40 of the 100 points and are the only way to pass 60%. Reviewers who agree with each other add the last points.
3. How Data Confidence Adds Up
The data confidence percentage (for example 88%) adds up three parts. From 75% a brand counts as stable; below that the brand page invites you to help stabilize the DNA.
Data Maturity
Total number of reviews compared with the number of bottles. More reviews reduce the impact of outliers.
Community Reviews
Reviews written by our own community instead of automated sources.
Consensus Factor
The more reviewers of the same bottle agree on its rating, the more points. It is measured per bottle with at least five ratings, so a range with a light and a heavy expression does not count as disagreement.
The Reliability value on the Brand DNA card (for example 7.9/10) is separate: the brand's average rating, lowered when reviewers of the same bottle disagree. It needs at least one bottle with five ratings.
4. Under the Hood
The models and software behind this page, and when they run.
-
spaCy
en_core_web_lgFlavor recognition: Finds the terms of our Flavor Dictionary in a review: exact matches, plus close relatives through 300-dimensional word vectors (a cosine similarity of at least 0.75).
When: Every review.
-
SciPy
Brand consistency: Compares the flavor profiles of a brand's bottles pair by pair (cosine similarity) and averages the result. No language model or generative AI is involved: the profiles come from flavor recognition, the rest is statistics.
When: Every time a brand page is opened.
Integrity Protocol
Reviews from our own bot account are kept apart from community reviews. They can add to volume and agreement, but never to the community part of the score.