Accuracy is the product.

Only official sources count

Every claim in a guide - each step, refund rule, pause option and data-retention statement - is backed by the service's own documentation: help centers, official docs sites, pricing pages, terms of service and privacy policies. For subscriptions billed through the Apple App Store or Google Play, we use Apple's and Google's own support pages.

We never source instructions from blogs, forums, Reddit, YouTube or other "how to cancel" sites, and we never write steps from memory of a UI. If the official documentation names a button, we quote that name. If it only describes where to go, our steps stay at that level of detail rather than inventing clicks.

What "verified" means here

Each guide lists every source it relies on, with the date it was last checked. A guide is only published with the status "verified", and every guide shows "Manually verified on [date] - based on official documentation" so you can judge its freshness yourself.

If we cannot verify how a service is cancelled from official sources, we do not publish a guide for it at all. An incomplete guide that looks complete is worse than no guide.

Guides expire after 60 days

Services change their flows constantly, so a verification date is a promise with a shelf life. Guides older than 60 days are automatically flagged in our tooling for manual re-review, and an automated link checker verifies that every official source URL still resolves.

When a source changes or breaks, a human re-reads the documentation and updates the guide. Nothing is rewritten automatically - not by the link checker, not by reader feedback, and not by AI.

Reader feedback is a signal, not an edit

Every guide has two buttons: "These steps still work" and "Something changed". Reports land in a review queue that a human works through - feedback never changes published content on its own. This keeps the site useful (you tell us when something drifts) without letting anyone inject wrong instructions.

Alternatives are comparisons, not rankings

We only publish an alternative when it has its own verified guide and a specific editorial reason to compare it. Generic category matches and external research leads stay hidden. Its plan, price, free-plan outcome and cancellation difficulty are read directly from that guide, so copied comparison facts cannot drift out of date.

A lower listed entry price does not mean a tool is universally cheaper or better. Seats, usage limits, features and billing terms can differ, so we show the exact plan basis and link to the official pricing source.

The Cancellation Difficulty Index is derived, not researched

Every guide carries a difficulty rating - Easy, Medium or Hard - set from the verified cancellation path. The Cancellation Difficulty Index turns that into one score from 1 to 10 for every guide, using nothing but the record itself: the rating fixes the band (Easy 1 – 3.9, Medium 4 – 6.9, Hard 7 – 10), and six weighted signals read from the same record place the service inside it: time & steps (25%), route to cancel (20%), obstacles (20%), findability (15%), money after cancel (12%), pause or downgrade (8%).

Signals come from the verified steps and timing, the billing route, the fact rows, the notice period and the refund, pause and downgrade text. Weights and thresholds are fixed constants; no score is set or adjusted by hand, and the score is recomputed from the record every time, so it can never drift from the guide it summarises. When a guide is re-verified and its facts change, its score changes with it.

Sponsors never touch the editorial side

Sponsor slots are clearly labelled, sold as a monthly subscription at a published ladder ($99, then +$100 per new advertiser, capped at $999), and physically separated from guides. Alternatives in guides are chosen and ordered editorially, before any slot is sold. If a sponsor asks us to change a guide, we end the placement.