Data integrity and migration testing

Verifying data stays correct through writes, migrations, imports and exports.

Mandatory for 27 of 46 product types. Usually run by a developer.

What it catches

• lost or corrupted records
• rounding
• duplicate rows

What it misses

• UI issues

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): Importing the same bank statement twice leaves one copy of each transaction, and the VAT report totals equal the sum of the posted ledger lines.
• clinical trial management software (Healthcare): After a protocol-amendment migration, a reconciliation report shows every subject, visit and value matches the pre-migration export, and audit trails are carried over unchanged.
• crypto exchange (Fintech): After each ledger close and each wallet movement, customer balances per asset in the ledger sum to no more than on-chain and custodial holdings, and no account other than approved market makers carries a negative balance.
• data pipeline (Infrastructure and developer products): For each run, row counts and control totals reconcile between source extract, staging and target, and a mismatch stops publication.
• dating app (Consumer apps): After a member deletes the account, a support lookup finds no profile, photos or messages, and matched members no longer see the chat.
• desktop application (Infrastructure and developer products): A test profile with known files keeps the same file list and contents before and after each update, including files saved in redirected folders.
• e-signature service (B2B SaaS): Editing a completed document makes signature validation fail, and editing a document mid-flow restarts signing for everyone who already signed.
• ecommerce store (Commerce): A daily report shows every cancelled or unshipped order with a card refund, not store credit, unless the shopper chose credit.
• electronic health record system (Healthcare): Medication lists and allergies for a sample of test patients match field for field between the sending and receiving records, and each required quality report produces a value.
• email marketing platform (B2B SaaS): After a list import or merge, previously suppressed contacts stay suppressed and every imported contact shows a consent source and date.
• ERP system (B2B SaaS): After migration, the trial balance, open payables and employee records match the legacy system to the cent and to the record.
• EV charging app (Infrastructure and developer products): For a sample of sessions, the kWh on each receipt matches the charger's meter values and the amount charged matches the price shown at start.
• expense management tool (Fintech): For a sample of approved expenses, each stored receipt image opens, shows a readable amount and date, and is linked to the same expense line after edits.
• fleet and logistics tracking (Infrastructure and developer products): After an edit to a driver's duty log, the original entry is still retrievable, the edit shows who made it, and the driver is asked to accept or reject it.
• insurance app (Fintech): After each rating data load, a reconciliation report shows no renewal whose premium differs from a recalculation with the filed rates.
• lending platform (Fintech): The credit bureau file for a set of test loans shows the same status, balance and days past due as the servicing screen, including a loan paid on its due date.
• loyalty program app (Commerce): After a ledger migration or program merge, every member's points balance and tier on a full export matches the old system at the published conversion ratio.
• online marketplace (Commerce): A daily ledger report shows, for every order, buyer charge equal to seller payout plus fees plus refunds, and lists every high-volume seller with verified details.
• machine learning model service (Infrastructure and developer products): On a held-out set labelled by people, accuracy and false-positive results are recorded per group and per input quality, and the release record shows them.
• mobile banking app (Fintech): After the overnight batch, balances, interest and statement lines for a seeded set of test accounts match the values worked out by hand in the reconciliation sheet.
• mobile game (Consumer apps): For a sample of sandbox purchases, every charge in the store's purchase records matches a purchase the player confirmed in the game, and each purchased item appears in the player's inventory once.
• neobank (Fintech): After each partner bank statement, the sum of customer balances in the app ledger equals the pooled account statement; a run that finds a difference lists it by customer.
• peer-to-peer payment app (Fintech): After each statement from the app's bank, the sum of user balances, pending sends and cash-outs equals the funds held there; a run that finds a difference lists it by user.
• pharmacy delivery app (Healthcare): A test controlled-substance refill past its prescription's refill limit, and one with no valid prescription on file, both stop in the pharmacist queue with a visible reason.
• ride-hailing app (Commerce): Every driver deactivation in a test week traces to a named human reviewer, and a sample of fraud flags rechecked by hand matches the recorded reason.
• subscription box service (Commerce): After each renewal run, the report shows no box shipped and no card charged for a skipped month or a cancelled or paused subscription, and no box queued for shipment on a subscription without a stored consent record.
• stock trading app (Fintech): Cash balance, buying power and position values on the app match the clearing firm's end-of-day records for a set of test accounts, including accounts with open options.

Tools

Alphabetical. Not recommendations.