The programme is green. There are 186 items in the RAID log. Nobody has opened it since the last steering committee.
A spreadsheet can hold a remarkable amount of trouble without reducing any of it.
The short answer: a RAID log records risks, assumptions, issues and dependencies, each with an owner and a next action. Use it to trigger decisions, not merely document concern. An item without an owner or response is a note, not a control.
Keep the four categories distinct
A risk is an uncertain event or condition that could affect objectives. An assumption is something the plan treats as true but still needs validation. An issue is a problem that already exists. A dependency is something one piece of work needs from another party or activity.
Some organisations use different meanings for the letters, including actions or decisions. Agree your convention at the start. This template uses risks, assumptions, issues and dependencies, with actions recorded against each item.
Link related entries rather than duplicating their history. An assumption can become an issue when disproved; a dependency may carry a risk of delay. Keep the connection visible.
Write entries that tell someone what to do
| Type | Weak entry | Useful entry |
|---|---|---|
| Risk | Supplier risk | If the interface arrives late, integration testing loses its planned window |
| Assumption | Environment ready | The test environment is assumed available before integration; confirmation needed |
| Issue | Access problem | The agreed test account is unavailable and today’s tests are blocked |
| Dependency | Partner input | Build completion needs an accepted interface specification from the partner |
A useful description names the event or condition and its consequence. The next action should have a person and date, not simply “monitor”.
Use the log to run a short review
Review changed items, overdue actions and matters needing a decision. Ask owners what evidence has changed, whether the response remains appropriate and whether the consequence now exceeds their authority.
Separate the operational review from the sponsor’s decision queue. A sponsor does not need to read every row, but does need a concise account of material exposure, choices and the decision required.
The APM risk-management guidance provides broader context. The template here is a suggested working record, not an official APM format.
Handle an assumption that fails
Suppose the plan assumes a partner can provide test data by the agreed date. The owner checks and learns that approval will take longer. Update the assumption as disproved, create or link the resulting issue, and assess the schedule options.
The response might be a sanitised substitute dataset, resequenced work or a revised milestone. Record who can approve the choice. Leaving the original assumption green because it once sounded reasonable serves nobody.
Keep closed items with the resolution and date. The history helps explain later decisions and prevents the same unresolved point reappearing under a new ID.
Define escalation thresholds
Use thresholds appropriate to the programme: impact on a critical milestone, spending beyond delegated authority, material service risk or an unresolved cross-team decision. Specify the receiving person or forum and the evidence needed.
Avoid a ritual in which everything is escalated and nothing is decided. The entry should say what is needed: approve an option, allocate capacity, accept a consequence or change the plan.
For UK and UAE delivery teams, record a named dependency owner and agreed timing rather than relying on “end of day”. Time zones and working calendars can turn a casual phrase into a missed handoff.
Keep it small enough to use
Archive resolved entries, merge genuine duplicates and focus reviews on material change. Do not delete uncomfortable history to improve the apparent health of the programme.
The test of a RAID log is whether it changes action. If it only changes the colour of a report, simplify the process until the people responsible can use it.
Sources and scope
Reviewed 1 November 2026. The examples and worksheets are my suggested working methods, not measured client results or official standards.
