How we score
The Experience Index measures how a subscription actually feels to live with — not how it markets itself. Every score is reproducible from logged sources, and we publish the conflicts instead of hiding them.
Methodology version 2.0. The numbers on this page are the exact weights and thresholds the scoring engine uses — published here so the results are reproducible, not merely asserted.
The five dimensions
- Exit ease — how painlessly you can cancel.
- Customer support — whether help is reachable and effective.
- Account sharing — what the rules are and how they’re enforced.
- Multi-device — how well it works across your screens.
- Price stability — how often and how much the price rises.
Three evidence streams
Each dimension is scored by triangulating three independent streams: manufacturer (official pages and price history), expert (datable, hands-on reviews from an allowlist of outlets), and community (trailing-12-month sentiment from named platforms, attributed by platform — never by individual username). The streams are blended at dimension-specific weights.
Per-dimension stream weights
Every dimension blends the three streams at its own weighting (each row sums to 100%). Sharing and price stability lean on official rules and price history; support and exit ease lean on lived experience.
| Dimension | Manufacturer % | Expert % | Community % |
|---|---|---|---|
| Exit Ease | 25% | 35% | 40% |
| Customer Support | 20% | 30% | 50% |
| Account Sharing | 55% | 25% | 20% |
| Multi-Device | 50% | 30% | 20% |
| Price Stability | 80% | 0% | 20% |
Composite weights
A service’s overall Experience Score is a weighted blend of its five dimension scores. We weight the dimensions people most regret getting wrong — getting out and getting overcharged — the heaviest. The weights below sum to 100%.
| Dimension | % of Experience Score |
|---|---|
| Exit Ease | 30% |
| Customer Support | 10% |
| Account Sharing | 20% |
| Multi-Device | 15% |
| Price Stability | 25% |
Consensus, not averages
When the official story and lived experience diverge by four points or more, we mark the cell contested and resolve the published score toward lived experience — we never average a conflict away. A cell with thin or missing evidence can never be rated “high consensus.”
Thresholds
These are the exact cut-offs the engine applies. They decide when a score is allowed to publish, when a cell is flagged as contested or a big swing, and when a stream is trusted.
- Composite minimum cells:
4 - A service’s overall Experience Score is only published once at least this many of the five dimensions have a committed, source-backed cell. Below it, we show the individual cells but no composite.
- Contested divergence:
4 - When the official (manufacturer) story and lived experience diverge by at least this many points, the cell is marked “contested” and resolved toward lived experience — never averaged away.
- High-consensus spread:
2 - A cell can only earn “high consensus” when the present per-stream sub-ratings sit within this many points of each other. Wider disagreement caps the cell at moderate or below.
- Big-swing delta:
2 - If a re-score moves a cell by at least this many points versus the prior committed value, it is flagged as a big swing and routed to human review before it can auto-publish.
- Expert minimum reviews:
3 - The expert stream’s confidence cannot exceed “moderate” until it is backed by at least this many datable, hands-on reviews from allowlisted outlets.
Provenance gate
No number is published without a logged source. Scores are computed by a deterministic engine; our editors and tools only set per-source sub-ratings from evidence they actually read. Every change is versioned, and contested or low-confidence scores are reviewed by a human before they’re trusted.
Methodology version 2.0
We bump the methodology version only when the rubric, weights, or thresholds change — never for routine re-scoring of individual services. The full history is below.
Changelog
- v2.0 — current rubric and weights (2026). Per-dimension stream weights, composite weights, and thresholds as published on this page. Contested cells resolve toward lived experience; the provenance gate withholds any unsourced score.