Unit tests

Automated checks of one function or class in isolation.

Mandatory for 18 of 46 product types. Usually run by a developer.

What it catches

• wrong calculations
• edge cases in logic
• regressions in pure code

What it misses

• integration faults
• UI
• configuration
• performance

Which product types need it and why

Done conditions used on playbook pages

Each mandatory cell has its own done condition, written for that product type.

• accounting and invoicing SaaS (Fintech): VAT, sales tax and rounding on invoice lines match a reviewed table for each tax rate and currency offered, including credit notes.
• buy now pay later service (Fintech): Instalment split, rounding of the last instalment and late-fee caps match a reviewed table for each plan length and currency offered.
• clinical trial management software (Healthcare): Every combination of randomisation strata allocates from the expected table, and dose calculations match the protocol table at each weight boundary.
• crypto exchange (Fintech): Margin, liquidation and fee rules return the expected result for each test case, and no account flag bypasses liquidation without a recorded approval.
• data pipeline (Infrastructure and developer products): Each transform returns the documented output for empty files, oversized files, null keys, duplicate keys and timezone boundaries.
• desktop application (Infrastructure and developer products): In the test report, the cleanup and repair code refuses to delete the home folder, the root folder, an empty path or any folder the app did not create.
• e-signature service (B2B SaaS): In the test report, every unchanged sample document passes signature validation and every sample with a changed byte or page fails it.
• email marketing platform (B2B SaaS): The suppression step drops every unsubscribed, bounced and complained address from a fixed send list before the send queue is built.
• ERP system (B2B SaaS): Each pay rule returns the expected amount for fixed edge cases: mid-period hire, retroactive raise, unpaid leave and termination.
• expense management tool (Fintech): Currency conversion, per diem and mileage amounts match a reviewed table for each rate and country the tool supports.
• identity and access management service (B2B SaaS): Credential and cache-key checks refuse a wrong password for usernames of every length in the fixed sample set, including the longest the directory allows.
• insurance app (Fintech): For a reviewed set of test drivers and homes, the rating engine returns the same premium as the filed rate tables, factor by factor.
• lending platform (Fintech): APR, amortization schedule and late-fee results match a reviewed spreadsheet for every loan product, including month-end and leap-day due dates.
• loyalty program app (Commerce): For a fixed set of sample members, redemption prices for points earned before a terms change stay at the old award price, and earn rules for each tier and promotion produce the points listed in the program terms.
• machine learning model service (Infrastructure and developer products): In the test report, every missing, out-of-range and low-quality sample input gets its documented value, and every sample score comes back with at least one reason.
• software as a medical device (Healthcare): Dose and threshold inputs written as ".5", "0.5", "5" and with a comma decimal each produce the intended value or a rejection, never a different dose.
• SDK or client library (Infrastructure and developer products): In the test report, every server-sent config and response field in the fixed sample set passes with a null, wrong-type and missing value and no crash.
• stock trading app (Fintech): Buying power, margin and options collateral calculations match a reviewed table for cash, margin and options accounts, including after a partial fill.

Tools

Alphabetical. Not recommendations.