How we work
Contents
At The Lama Nest we apply the LAMA Method™, our in-house system to analyze products and reviews rigorously and clearly. We combine AI-assisted automation with human review. Our goal: help you make better decisions with less noise.
Our Principles
- Usefulness over hype: technology is a tool; judgment is human.
- Transparency: we explain the methodology, publish the update date, and cite sources when appropriate.
- Rigor & simplicity: robust metrics with easy reading and actionable recommendations.
- Independence: if there is any affiliate relationship, we disclose it; our conclusions are not for sale.
The Process, Step by Step
- 1) Scope & categories — We define the product, variants, and evaluation categories (e.g., perceived quality, ease of use, support, etc.).
- 2) Data collection & preparation — We gather public reviews from different markets (when available), remove duplicates, and detect noise (spam, bots, copies).
- 3) Normalization & aggregation — We harmonize scales, languages, and formats; we balance review weights by period and market to avoid bias.
- 4) LAMA metrics computation — We apply four core metrics and generate visualizations for quick reading.
- 5) Human review — We validate outliers, cross-check with technical sheets, manufacturer support and, when applicable, regulatory documentation.
- 6) Publication & monitoring — We publish with an update date and re-evaluate if reviews or the product change substantially.
LAMA Method™ Metrics
What it measures: the minimum percentage of good reviews (4–5★) we can guarantee with 95% confidence.
How it’s calculated: Wilson interval on the % of 4–5★; we use the lower bound (LB95).
How to read it: higher = better. Indicates a solid base of satisfied customers.
What it measures: the maximum percentage of bad reviews (1★) that could exist, with 95% confidence.
How it’s calculated: Wilson interval on the % of 1★; we use the upper bound (UB95).
How to read it: lower = better. Signals a low likelihood of mass “haters.”
What it measures: a practical risk that the product causes issues due to technical or regulatory topics.
How it’s calculated: we detect risk patterns in descriptions and reviews and summarize them as LOW / MEDIUM / HIGH.
How to read it: the higher the risk, the more caution. With few reviews, the score may rise due to uncertainty (the model’s caution).
How We Interpret Results
- Quick read — High Happy Llamas + Low Grumpy Llamas = confidence in overall satisfaction. High consistency = cross-country alignment. Low disappointment risk = little post-purchase friction.
- Context matters — New products or small datasets may show wide intervals; we disclose that and update when volume permits.
- Practical conclusions — We close each analysis with clear recommendations based solely on the objective information collected (who it’s for, pros/cons, what to check before buying, etc.).
Quality Controls
- Temporal sampling — We monitor “spikes” (campaigns, new versions) that can skew perception.
- Noise detection — Signals of inauthentic or coordinated reviews are minimized.
- Traceability — We log dates, sources, and relevant product changes.
- Editorial review — A person validates the final judgment and the explanatory text.
What We Publish (and What We Don’t)
- We publish: conceptual methodology, metric definitions, visualizations, and conclusions.
- We don’t publish: prompts, code, additional internal thresholds, or detailed pipelines (trade secret).
Contact & Suggestions
Are you a manufacturer who wants to provide technical documentation or clarifications? Are you a user who has noticed a change? Write to us. We improve the LAMA Method™ with you.