ShipDataFast
Discuss a reconciliation
Menu
Data Reconciliation Migration Assurance Data Validation & Engineering Products Recon Check AI Agent QA Audit Rule Packs Financial Services About Blog Discuss a reconciliation
Home / Blog
BLOG

Notes on getting financial data right.

Longer write-ups of the reconciliation, migration and ETL regression topics we discuss on LinkedIn: what goes wrong, why, and the checks that catch it.

  1. What went wrong: 220 million trade reports that didn't match the trades

    The FCA fined Goldman Sachs International £34.3m in 2019 after 220.2 million transaction reports were wrong over 9.5 years. What the final notice says broke in the data, and 5 reconciliation checks that catch it.

  2. Green means it ran. Reconciled means it's right.

    A green pipeline run tells you the job finished, not that the numbers are right. Four things green doesn't prove, and where to start checking.

  3. It's 2am and the totals don't match. Who calls the rollback?

    Cutover plans detail moving the data but go vague on what happens when it doesn't reconcile. Four things to agree in writing before the weekend.

  4. The data moved. What was it translated into?

    Migration tests check that data moved, rarely what codes were mapped to. Four checks on value mapping to do in writing before cutover.

  5. Reconciled is not evidence: 4 things to see before sign-off

    A green tick can't be rerun or defended. The 4 things I want to see before signing off a reconciliation, including the one teams skip.

  6. Every pipeline or report change is a mini migration

    A change that only touched one join can still drop or double rows. Five steps to compare old and new output before you release a pipeline change.

  7. Most false breaks are type problems: normalise before you compare

    Leading zeros, date formats, rounding, trailing spaces and blanks cause most false breaks. Normalise these 5 before you compare two systems.

  8. Same cut-off, both sides: why most reconciliation differences aren't real

    Compare two systems at different cut-offs and most of the differences won't be real. What to agree and write down before the comparison.

  9. Most migrations don't fail on load day: 5 questions to answer before sign-off

    Migrations stall at sign-off when nobody can explain the differences. The 5 questions to answer in writing before your next mock load or parallel run.