Planning Artifact design intermediate

RAID Log

Risks, assumptions, issues and dependencies in one place, each with an owner and a date — the artifact that makes the difference between a surprise and a decision.

Four registers people keep separately and should not. The distinction that matters most is risk versus issue: a risk might happen, an issue already has. Mixing them makes it impossible to tell what needs mitigating from what needs fixing.

The shape

ID Type Description Impact Prob Owner Response Due Status
R-01 Risk Legacy data quality worse than sampled; migration fails validation High Med Data Lead Full-volume profiling in wave 1, not a sample 12 Sep Open
R-04 Risk Key mainframe SME retires in Q4 High High Programme Shadowing from Aug; record interface walkthroughs 30 Aug Mitigating
R-07 Risk Partner cannot test EDI before Nov Med High Integration Build a conformance simulator; contract clause 20 Sep Open
A-02 Assumption Legacy remains available read-only for 30 days post-cutover Ops Validate with infra by 05 Sep — cutover plan depends on it 05 Sep Unvalidated
A-05 Assumption Peak is 4× average, as in last 3 years Product Confirm against this year's forecast 15 Sep Validated
I-03 Issue Test environment unavailable 9 days; wave 2 slipping High Platform Ephemeral environments funded; interim shared slot 08 Sep In progress
I-06 Issue Two systems disagree on customer count by 4% Med Data Lead Root cause: duplicate merge rule; reconciliation running 14 Sep In progress
D-01 Dependency Network team must deliver the hybrid link High Network Committed 30 Sep; blocks wave 2 30 Sep On track
D-04 Dependency Security sign-off on the landing zone High Security Review booked 22 Sep 22 Sep At risk

When you produce it

At programme start, and it is reviewed weekly for the whole programme. A RAID log written once at initiation and never revisited is worse than none, because it gives the appearance of risk management.

Who reads it

The steering committee, weekly. Workstream leads, who own rows. Audit, on a programme that goes wrong — which is a good reason to keep it honest as you go.

What good looks like

  • Assumptions have a validation date and an owner. An unvalidated assumption is a risk that has not been written down yet, and this is where most programmes actually fail.
  • Risks have a response, not just a score. "High/High, monitor" is not a plan.
  • Issues have a resolution date, and slipping dates are visible rather than quietly moved.
  • Dependencies name the external party and what they block.
  • Closed items stay in the log with their outcome, which is where the lessons are.

Common mistakes

  • Conflating risks and issues, so the register cannot be acted on.
  • Assumptions never validated, then discovered to be false at cutover.
  • Scoring theatre — elaborate probability-impact arithmetic with no corresponding action.
  • Owned by "the programme", which means nobody.