Transparency
How we score dashcams
Every score on this site is produced by a documented, repeatable process — not editorial opinion. Here is exactly what we do, why we do it, and what the numbers mean.
What this is — and isn't
This is meta-analysis. We do not physically test dashcams. Instead, we systematically process real YouTube reviewer experiences and synthesise them into a score. Think of it as aggregating 15 independent product tests rather than running one ourselves.
This is not a spec sheet aggregator. Anyone can list resolution and sensor type. What we extract is lived experience — what happens to the night vision after 8 months in Chennai traffic, not what the spec sheet claims.
We have no commercial relationship with any brand. Affiliate links earn a small commission from purchases — not from scores. A brand cannot pay to improve their score.
The pipeline
Video discovery
We use the YouTube Data API to find the top 10–15 review videos for each dashcam model. We filter for videos over 5 minutes, over 1,000 views, and in English. Brand-owned channels and paid review disclosures are excluded.
Transcript extraction
We extract the auto-generated or manually-uploaded transcript from each video via the YouTube Transcript API. No audio processing — we work only from official transcripts.
LLM attribute extraction
Each transcript is sent to Gemini 2.0 Flash with a structured prompt. The model extracts reviewer sentiment for each of our 6 scoring attributes, produces a score out of 10, a confidence level (0–1), and a direct evidence quote from the transcript.
Confidence-weighted aggregation
Scores are aggregated across all transcripts. Each individual score is weighted by confidence — a reviewer who spent 3 minutes specifically testing night vision carries more weight than one who mentioned it in passing. The final score per attribute is a weighted average.
Overall score calculation
The 6 attribute scores are combined using the weights below. The result is scaled to a 0–100 overall score. This score is stored with the date, review count, and source channel list.
Scoring attributes & weights
Weights are intentional, not arbitrary. Night vision is weighted highest because India and Southeast Asian road conditions at night are the primary safety use case for this audience. This is a deliberate, defensible methodology choice.
Night / Low-light Video
25%India and Southeast Asian roads are most dangerous at night. Poor streetlighting, motorcycles without lights, and high-beam glare are the primary safety risks. Night vision performance is the single most important differentiator for this audience.
Day Video Quality
20%Resolution, sharpness, and colour accuracy in normal driving conditions. Number plate legibility at typical Indian highway speeds is the key test.
Build Quality
15%India and Southeast Asia see cabin temperatures of 45–60°C in summer. Supercapacitor vs battery, heat resistance, and mount durability are critical — failures here are permanent.
App & WiFi Connectivity
15%Most modern dashcams require a smartphone app for footage review. App reliability, WiFi stability, and Android/iOS compatibility are commonly flagged in Indian reviewer communities.
Value for Money
15%Assessed in Indian market context. A camera priced at ₹8,000 is judged differently than the same model in the US at $150 — purchasing power parity and feature availability in India matter.
Install & Setup
10%Ease of DIY installation, cable routing, and initial configuration. Indian buyers often install dashcams themselves rather than using a workshop.
What the scores mean
Strong
Consistent positive sentiment across most reviewers. Worth serious consideration.
Mixed
Some positives, some concerns. Worth considering if it addresses your specific use case.
Weak
Significant recurring issues across reviewers. Proceed with caution.
Limitations — be honest about them
Sample size varies by model. Popular models like 70mai A800S have 13+ reviews available. Less popular models may have only 3–5 suitable videos. Scores from fewer reviews are less reliable — we show the review count prominently for this reason.
English-language bias. We process English transcripts only. Indian and SEA reviewers who make videos in Hindi, Tamil, Bahasa, or Tagalog are not represented. This skews toward urban, English-speaking reviewer demographics.
Recency. Scores reflect the reviews available at analysis time. Firmware updates, quality control changes, or new model variants may not be reflected until we re-run the pipeline. We show the last-updated date on every page.
LLM imprecision. Gemini 2.0 Flash interprets reviewer language well but is not perfect. We include confidence scores to surface uncertain attributions. Low-confidence scores are labelled accordingly.