How does Fire Your SaaS decide whether to cancel a subscription?

A deterministic rule engine evaluates nine conditions in a fixed order against your supplied inputs and a verified playbook. The first condition that fires sets the verdict. No LLM, price scraper, or vendor claim overrides it.

What are the five exit verdicts?

REPLACE_CANDIDATE (a verified playbook supports a bounded pilot), AUGMENT (run the playbook beside the incumbent), UPGRADE (test the incumbent native AI first), KEEP (a blocker or unfavorable economics outweighs the exit), and NEEDS_REVIEW (mandatory evidence is missing).

Does a Replace candidate verdict mean I should cancel now?

No. It means the supplied inputs and a verified playbook support a bounded replacement pilot. Cancellation still requires the acceptance checks, an export, and rollback evidence.

What happens if I do not enter my actual bill or contract term?

The report withholds the numbers it cannot justify. With no bill it shows the replacement range but withholds savings, payback, and ROI. With no contract window it shows stable monthly net but withholds first-year net and payback.

Where is my report stored, and is it private?

The report is stored anonymously in this browser’s local storage for 30 days under a 32-character random ID. There is no account and no email attached. The URL will not resolve on another device or after storage is cleared.

METHODOLOGY / RULE VERSION exit-rules-2026.11.1

予言者はいません。明示的なルールだけです。

A SaaS exit decision is a workflow decision, a data decision, and a billing decision. This page shows exactly how the report reaches its verdict, what it refuses to estimate, and where the evidence can go stale.

Fire Your SaaS decides whether a software subscription can be exited with a deterministic rule engine, not an AI model. It checks 9 conditions in a fixed order — verified playbook coverage, export readiness, data sensitivity, native AI, system of record, capability gaps, contract window, and cost parity — and returns exactly one of 5 verdicts: REPLACE_CANDIDATE, AUGMENT, UPGRADE, KEEP, or NEEDS_REVIEW. Missing inputs are withheld rather than estimated, and a Replace candidate verdict never approves cancellation.

01 The five verdicts

One state per report. Each state carries its own limits.

置き換え候補 REPLACE_CANDIDATE / RC: A verified playbook covers every required capability and no blocking constraint was found. 「置き換え候補」は解約の承認ではありません。プレイブックを実行し全検査に合格してから解約してください。
併用 AUGMENT / AU: One or more must-have capabilities are not covered, or the work is too sensitive for an automatic replacement commitment. No cancellation is assumed. The incumbent bill remains in the model.
アップグレード UPGRADE / UP: You confirmed the current plan already includes usable native AI for this workflow. Native AI coverage is self-reported. Test it against your must-haves before changing plans.
維持 KEEP / KP: A blocking constraint, an unexportable record, a contract window, or cost parity makes cancellation irresponsible on the supplied inputs. Keep the incumbent and work the listed constraints. A weak migration is not a saving.
要確認 NEEDS_REVIEW / NR: The workflow is not covered by a verified playbook, the evidence is out of date, or a critical input is unknown. No savings are estimated. Collect the missing evidence or ask for a human review.

02 Rule order

Precedence matters. The first condition that fires sets the verdict.

01 No verified playbook or expired evidence → 要確認
If the workflow has no published, verified playbook, or the playbook is older than 90 days, the report stops. No replacement is recommended.
02 A critical input is unknown → Needs review / Keep
Unknown export status or data sensitivity produces Needs review. A confirmed inability to export key data produces Keep.
03 Native AI is confirmed by the user → アップグレード
If the user reports that the current plan already includes usable native AI, the report recommends testing that path first. It does not assume an upgrade price.
04 The incumbent is the system of record → Augment / Keep
The playbook decides whether the record can stay while one workflow moves, or whether the whole tool must stay.
05 Sensitive data → 併用
Sensitive content is never moved automatically. Only a bounded step with human review and an approved data agreement is modeled.
06 A must-have capability is missing → Keep / Augment
A blocking gap produces Keep. A non-blocking gap produces Augment. The missing capability is named in the report.
07 The contract leaves no room → 維持
If 12 months remain, there are no avoidable subscription months in the next year. A replacement would add cost without removing the bill.
08 The replacement is not cheaper → 維持
If the lowest modeled replacement cost is at or above the supplied bill, the report does not claim a saving.
09 Coverage complete and no blockers → 置き換え候補
Only then does the report recommend a bounded replacement pilot. It still does not approve cancellation.

“Replace candidate” is not cancellation approval. It means the supplied inputs and the verified playbook support a bounded pilot. Cancellation still requires the acceptance checks, export, and rollback evidence.

03 Reason codes

Codes are stable enough to audit. The report prints the same code the rule used.

Code	Meaning
`WORKFLOW_NOT_COVERED`	The workflow has no verified playbook.
`EVIDENCE_MISSING`	The matched playbook lacks a verified date or source set.
`EVIDENCE_OVERDUE`	The playbook is outside the 90-day review window.
`REQUIREMENTS_MISSING`	No must-have capabilities were selected.
`EXPORT_BLOCKED`	Key data cannot be exported or reacquired.
`EXPORT_UNKNOWN`	Data portability has not been tested.
`NATIVE_AI_REPORTED`	The user reports native AI coverage; it is self-reported.
`SYSTEM_OF_RECORD_KEEP`	The incumbent is the record and the playbook says keep it.
`SYSTEM_OF_RECORD_AUGMENT`	The incumbent stays while one workflow moves.
`SENSITIVITY_UNKNOWN`	The data classification is unknown.
`SENSITIVE_DATA`	Sensitive data requires a reviewed control path.
`BLOCKING_CAPABILITY_GAP`	A required capability is not replaced.
`CAPABILITY_GAP`	A non-blocking capability remains with the incumbent.
`CONTRACT_WINDOW`	The remaining contract leaves no avoidable months.
`COST_PARITY`	The replacement floor is at or above your bill.
`COVERAGE_COMPLETE`	Every selected must-have is covered.
`ALL_CAPABILITIES_COVERED`	The coverage map has no gaps.
`NO_BLOCKERS`	No export, data, record, contract, or coverage blocker was found.
`EVIDENCE_CURRENT`	The playbook is inside its review window.
`COST_ADVANTAGE`	The modeled monthly net is positive; one-time costs excluded.
`COST_RANGE_UNCERTAIN`	The modeled saving range crosses zero.
`SPEND_UNKNOWN`	No actual bill was provided.
`CONTRACT_UNKNOWN`	The remaining contract term was not provided.
`PARALLEL_WINDOW`	A pilot may overlap the contract and add cost.

04 コスト計算式

No default bill. No hidden human-time savings. Missing inputs stay missing.

Term	Formula
Current monthly	`current_monthly = the amount you enter`
代替案の月額	`verified playbook range (min–max)`
一度きりの移行	`verified playbook range (min–max)`
Avoidable months	`12 − remaining contract months`
Replacement months	`assumed equal to avoidable months`
安定した月次純額	`current_monthly − replacement_monthly`
初年度純額	`current × avoidable − replacement × replacement months − migration`
回収期間	`migration_once ÷ stable monthly net`

No current bill: the report shows the replacement range but withholds savings, payback, and ROI.
No contract window: stable monthly net can be shown; first-year net and payback are withheld.
Ranges stay ranges: the low and high assumptions are reported separately. The report never presents a midpoint as a fact.
Payback is conditional: it appears only when the worst-case monthly net is positive and there are avoidable months.
Human time is separate: reclaimed operator hours are described but never added to subscription savings.
Augment and Keep do not model cancellation: the incumbent bill remains; replacement ranges describe a bounded automation budget, not savings.

Published prices are list prices and usage allowances, not quotes. Your bill, taxes, annual prepayment, usage, and vendor discounts can change the result.

05 根拠の基準

A playbook is verified only when the workflow was run and the evidence was recorded.

Sources with dates

Every price source has a URL and a checked date. A vendor claim is labeled as a claim, not a test.

請求の前に検収

The playbook includes accuracy, permissions, exception, and rollback checks. A pilot is not a proof until those checks pass.

90-day review window

After 90 days, a playbook is marked Review overdue and cannot support a replacement verdict.

06 データ・保存・プライバシー

Minimum collection, local reports, and no billing credentials.

Reports stay in the browser

The report snapshot is stored in this browser’s local storage for 30 days. It is not attached to an email or an account. A high-entropy report ID prevents sequential guessing, but it will not resolve on another device or after storage is cleared.

No secrets or client records

Do not enter billing screenshots, API keys, passwords, or customer data. The audit asks only for the software name, the actual bill, seats, contract window, workflow, capabilities, and constraint answers.

Analytics without PII

Funnel events use workflow and verdict labels only. Email addresses, free-text notes, bills, and other amounts are not sent to analytics. The site uses Google Tag Manager; event names are listed in the source.

Lead intake is consent-gated

The contact form is the only place email is requested. If a configured intake endpoint is unavailable, the UI must show a failed or unconfigured state rather than pretend a request was received.

07 Limits

What this product is not allowed to claim.

The report is not an audit, a procurement approval, or a legal/compliance opinion.
It does not scan bank, inbox, or software accounts and does not cancel anything.
It does not claim a migration will preserve quality, integrations, or client acceptance.
It does not add human-hours savings to subscription savings or guarantee a first-year return.
It does not treat a vendor's native AI claim as verified. The Upgrade state is based on the user's own confirmation.
Estimates are directional. Only your invoice, contract, and tested acceptance results are authoritative.

08 Frequently asked questions

Short answers to the questions the report itself cannot answer.

How does Fire Your SaaS decide whether to cancel a subscription?: A deterministic rule engine evaluates nine conditions in a fixed order against your supplied inputs and a verified playbook. The first condition that fires sets the verdict. No LLM, price scraper, or vendor claim overrides it.
What are the five exit verdicts?: REPLACE_CANDIDATE (a verified playbook supports a bounded pilot), AUGMENT (run the playbook beside the incumbent), UPGRADE (test the incumbent native AI first), KEEP (a blocker or unfavorable economics outweighs the exit), and NEEDS_REVIEW (mandatory evidence is missing).
Does a Replace candidate verdict mean I should cancel now?: No. It means the supplied inputs and a verified playbook support a bounded replacement pilot. Cancellation still requires the acceptance checks, an export, and rollback evidence.
How long is a playbook considered verified?: 90 days. After that window the playbook is marked Review overdue and cannot support a replacement verdict until it is re-checked with dated sources.
What happens if I do not enter my actual bill or contract term?: The report withholds the numbers it cannot justify. With no bill it shows the replacement range but withholds savings, payback, and ROI. With no contract window it shows stable monthly net but withholds first-year net and payback.
Where is my report stored, and is it private?: The report is stored anonymously in this browser’s local storage for 30 days under a 32-character random ID. There is no account and no email attached. The URL will not resolve on another device or after storage is cleared.

Test one real subscription.

No login, no bank connection, no auto-cancel. Bring the invoice amount and the workflow you need to keep.

退出監査を始める

ルール版: exit-rules-2026.11.1
Review window: 90 days
Report retention: 30 days
Decision model: Deterministic, no LLM
Last revised: 2026-11-09