Blog
raid log templateraid lograid log example

RAID Log Template: Risks, Assumptions, Issues, Dependencies

Copy a RAID log template with columns for risks, assumptions, issues, and dependencies, a worked example, and an AI prompt to fill it in.

Portrait of Luis Chavez-Mattos
Edited by Luis Chavez-Mattos, Director, Product··RSS

For Operations, Managers and Engineering

A bearded man in a vintage sports jersey sits in profile in a dark room, visually sorting scattered paper notes into four neat piles under the glow of an unseen screen.

What is a RAID log?

A RAID log is a single document where a project or program team tracks four things: Risks (things that might go wrong), Assumptions (things you are treating as true but have not confirmed), Issues (things that have already gone wrong), and Dependencies (things you need from someone else, or that someone needs from you). Each entry has an owner, a status, and a next action. The team reviews it on a regular schedule, usually weekly. Below is a template you can copy, a filled-in example, and an AI prompt that builds the first draft from your meeting notes.

TL;DR

  • A RAID log has four sections with different columns, because a risk needs a probability score and a dependency needs a "needed by" date. One generic table for all four hides that.
  • Every entry needs one named owner and one next action. An entry without both is a worry, not a log item.
  • A risk is something that might happen. An issue is something that has happened. When a risk occurs, move it to Issues and keep the ID.
  • An assumption is only useful if it has a date and a person assigned to confirm it. Unconfirmed assumptions are risks waiting to be promoted.
  • Review the log weekly for 15 to 20 minutes, and close or update every open item that is past its date.
  • AI is good at turning messy notes into draft rows. It is bad at knowing who actually owns something or when it is really due, so you check those fields yourself.

What does a RAID log template look like?

Here is the full template. Copy each table into a spreadsheet tab (one tab per section works well), or paste all four into one document. Fields in brackets are instructions you replace.

Header block (top of the log)

FieldEntry
Project / program name[Name]
Log owner[One person who keeps the log current]
Sponsor[Name]
Review cadence[Weekly, day and time]
Last reviewed[Date]
Scoring scaleProbability 1 to 5, Impact 1 to 5, Score = P x I

R: Risks

IDDate raisedRisk description (If [cause], then [event], resulting in [impact])CategoryProbability (1-5)Impact (1-5)ScoreResponse (avoid, reduce, transfer, accept)Next actionOwnerAction dueStatusLast reviewed
R-01Open

A: Assumptions

IDDate raisedAssumption (We are assuming that...)Why it mattersWho can confirmConfirm by (date)If wrong, what changesStatus (Unconfirmed, Confirmed, Disproved)OwnerLast reviewed
A-01Unconfirmed

I: Issues

IDDate raisedIssue description (What happened, and what is it blocking)Severity (Low, Medium, High, Critical)Impact on scope, schedule, or costNext actionOwnerAction dueEscalated toStatusDate resolved
I-01Open

D: Dependencies

IDDate raisedWhat is neededDirection (We need it, or They need it from us)Other party (team or vendor)ContactNeeded byCurrent status (Not started, On track, At risk, Late, Delivered)Impact if lateOwner on our sideLast reviewed
D-01

That is the whole template. The rest of this post explains how to fill it in so it stays useful past week two.

How do you fill in each section?

Risks

Write each risk as a sentence with a cause, an event, and an impact: "If the vendor misses the March 14 delivery, then testing starts late, resulting in a two-week slip to launch." That format forces you to be specific. "Vendor risk" tells nobody anything.

Score probability and impact from 1 to 5, then multiply. A 5 x 4 gets discussed every week. A 1 x 2 gets a glance. The numbers are a conversation tool, not math. Two people will score the same risk differently, and that disagreement is the useful part.

For the response, use one of four words:

  • Avoid: change the plan so the risk cannot happen.
  • Reduce: take an action that lowers the probability or impact.
  • Transfer: move the exposure elsewhere, such as a contract clause or insurance.
  • Accept: do nothing now, and say so on purpose.

"Accept" is a legitimate answer. Writing it down means someone decided, instead of nobody noticing.

Assumptions

Phrase these as "We are assuming that..." and add the date by which someone will confirm. An assumption like "the finance team will have budget approval by Q2" is a risk if you have no plan to check it. The "If wrong, what changes" column is what makes the entry worth keeping: if the answer is "nothing," delete it.

When an assumption is disproved, do not delete the row. Change the status to Disproved, then create a new risk or issue that references it (for example, "see A-03"). That trail is useful in a post-project review.

Issues

An issue is already happening. It does not get a probability score because it is no longer uncertain. It gets a severity and an owner. If the issue came from a risk, keep a note like "Was R-04" so the history stays connected.

Add an "Escalated to" column and use it. If an issue is High or Critical and nobody above the project team knows, the log is not doing its job.

Dependencies

Track both directions. What you need from others is obvious. What others need from you is the part teams forget, and it is where you get surprised by a call asking why your deliverable is late. For each dependency, name a real contact on the other side, not just a team name. "Needed by" should be a date, not "ASAP" or "Q3."

What does a filled-in RAID log look like?

Here is a worked example for a made-up program: migrating a customer support team from one ticketing tool to another at a mid-sized company. The entries are illustrative.

Risks

IDRiskPIScoreResponseNext actionOwnerDueStatus
R-01If agents get under one day of training before cutover, then handle times rise and customer wait times grow, resulting in a dip in satisfaction scores in the first month4312ReduceSchedule two 90-minute sessions per shift before cutoverDana (Support Ops)Mar 3Open
R-02If the data export from the old tool drops attachments, then agents lose context on open cases, resulting in repeat questions to customers3412ReduceRun a test export on 200 closed tickets and compare attachmentsPriya (IT)Feb 21Open
R-03If the vendor's support queue is slow during cutover week, then we wait on fixes, resulting in a longer parallel run236AcceptConfirm named contact and escalation path in writingMarcus (Procurement)Feb 28Open

Assumptions

IDAssumptionWho can confirmConfirm byIf wrongStatus
A-01We are assuming the new tool's email integration works with our current shared mailboxesPriya (IT)Feb 18Mailbox setup moves into scope; add about a weekUnconfirmed
A-02We are assuming no major product launch lands in the cutover monthLena (Product)Feb 25Ticket volume spikes; cutover date needs to moveUnconfirmed

Issues

IDIssueSeverityImpactNext actionOwnerDueEscalated toStatus
I-01Security review of the vendor has not started and is blocking the contract signatureHighContract delay pushes the whole timelineRequest a review slot and ask the sponsor to prioritize itMarcus (Procurement)Feb 17SponsorOpen

Dependencies

IDWhat is neededDirectionOther partyNeeded byStatusImpact if lateOwner
D-01Sandbox environment for the new toolWe need itVendor onboarding teamFeb 14At riskTesting cannot startPriya
D-02Final list of ticket categories and routing rulesThey need it from usVendor implementation leadFeb 24On trackVendor cannot configure routingDana

Notice a few things. Every row has a name and a date. The risks are written as cause, event, impact. The accepted risk (R-03) still has a next action, because "accept" means you have chosen to monitor it. And the issue (I-01) was once a risk that nobody acted on, which is the usual story.

How can AI help you build a RAID log?

AI is useful here for the blank-page problem. Most teams have the raw material already: meeting notes, a kickoff deck, a project charter, email threads. Turning those into structured rows is tedious. A chat assistant does a decent first pass in under a minute.

Paste this prompt, then paste your notes below it:

You are helping a program manager build a RAID log (Risks, Assumptions, Issues, Dependencies).

Read the notes below. Extract every item that fits one of the four categories and output four markdown tables.

Rules:
- Risks: things that might happen but have not. Write each as "If [cause], then [event], resulting in [impact]". Score probability (1-5) and impact (1-5) and multiply for a score. Suggest a response: avoid, reduce, transfer, or accept.
- Assumptions: things the notes treat as true without confirmation. Write as "We are assuming that...". Add "If wrong, what changes".
- Issues: things that have already happened and are causing a problem now. Add a severity: Low, Medium, High, Critical.
- Dependencies: things we need from others, or others need from us. Mark the direction and the other party.
- Use only names and dates that appear in the notes. If an owner or date is not in the notes, write "TBD" in that cell. Do not guess.
- After the tables, list up to five items that look like they belong in the log but are too vague to write down, and tell me what question I should ask to clarify each one.

Columns:
Risks: ID, Description, Probability, Impact, Score, Response, Next action, Owner, Due
Assumptions: ID, Assumption, Why it matters, Who can confirm, Confirm by, If wrong
Issues: ID, Description, Severity, Impact, Next action, Owner, Due
Dependencies: ID, What is needed, Direction, Other party, Needed by, Impact if late, Owner

NOTES:
[paste meeting notes, charter, or email thread here]

The "use only names and dates that appear in the notes" rule matters. Without it, assistants tend to fill empty cells with plausible-sounding owners and deadlines.

A second prompt is useful once the log exists, before your weekly review:

Here is my current RAID log. Review it as a skeptical project sponsor would.

1. Flag any entry with no owner, no date, or no next action.
2. Flag any risk that reads like an issue (it has already happened) or any issue that reads like a risk.
3. Flag any assumption that has no confirm-by date, or whose date has passed.
4. Point out dependencies that are likely to affect more than one risk or issue.
5. List the five questions I should ask in the review meeting.

LOG:
[paste the log]

What should you check in the AI's answer?

  • Owners and dates. If it filled any in that were not in your notes, remove them. A made-up owner is worse than a "TBD."
  • Scores. Treat the probability and impact numbers as a starting point. You and your team know the context. Change them.
  • Risk versus issue. Assistants often file things in the wrong bucket. Check each one against "has this happened yet?"
  • Missing items. The AI only sees what you pasted. It cannot know about the hallway conversation where someone said the budget might get cut.
  • Duplicates. Overlapping notes often produce two rows for one concern.

Where should you not trust AI?

Do not let it decide who owns something, or whether a risk is acceptable. Those are decisions that need a person. Also check your company's policy before pasting project notes into a public AI tool. Notes can contain customer names, contract terms, or personnel matters. If your company has an approved assistant, use that one, or remove sensitive details first.

How often should you review a RAID log?

Weekly for most programs, for 15 to 20 minutes, with the log on screen. Go through it in this order:

  1. Issues first. They are already costing you something.
  2. Dependencies that are due or late. Late dependencies become issues quickly.
  3. Top-scoring risks. Check whether the next action happened and whether the score changed.
  4. Assumptions past their confirm-by date. Confirm, disprove, or push the date with a reason.
  5. New items. Ask the room what is missing.

After the review, update the "Last reviewed" column. If a row has not been touched in a month, either close it or ask why it is still there. Fast-moving programs, such as a launch week or a cutover, may need a short daily check on just issues and dependencies.

What is the difference between a risk and an issue?

A risk is uncertain. An issue is certain. "The vendor might deliver late" is a risk. "The vendor told us today that delivery slips two weeks" is an issue. The distinction matters because they need different handling. Risks call for prevention and monitoring. Issues call for resolution and, often, escalation.

Assumptions sit between the two. An unchecked assumption is a risk you have not named yet. A common habit that works: for each assumption, ask "what is the risk if this is wrong?" If the answer is big, add a matching risk and cross-reference the IDs.

Dependencies are different in kind. They describe a relationship, not a problem. But a late dependency is an issue, and a dependency that looks shaky is a risk. Many teams link them by writing the dependency ID in the risk description.

What are common RAID log mistakes?

  • Treating it as a one-time kickoff exercise. A log filled in at kickoff and never reopened is a document, not a tool.
  • Vague entries. "Resource constraints" has no owner and no action. "Two of the four testers are on another project until March 10" does.
  • Too many items. Fifty open risks means none are managed. Close the low scorers or merge them.
  • No owner, or a team as the owner. A team is not a person. Pick one name.
  • Hiding bad news. If the log only shows green, people stop trusting it. Make it safe to add a risk without it being read as blame.
  • Never closing items. Closed items are a record of what the team handled. Keep them on an archive tab.
  • Skipping the sponsor. If High issues and top risks do not reach the sponsor, the log is a private diary.

Where should you keep a RAID log?

Use whatever your team will actually open every week. A shared spreadsheet is the most common choice because it is easy to sort, filter, and color-code by status or score. A table in your team wiki works if people already live there. Some teams keep the log inside their project tracker, using a custom item type for each of the four categories. Pick one location, put the link in the project charter and the weekly meeting invite, and avoid copies. Two versions of a RAID log is how items disappear.

For programs that span several projects, keep one log per project and a short program-level view of the top risks and issues across all of them. Don't merge every project's entries into one giant list.

Could you build your own RAID log app instead of using a spreadsheet?

If your spreadsheet RAID log has outgrown its tabs, you might want something with real owners, overdue reminders, and a view for the sponsor. Remy builds a full-stack app from a plain-English description. You could describe the four sections, the fields in the template above, sign-in with roles (editors for the team, read-only for sponsors), and a weekly email listing overdue actions per owner. Remy deploys it to a live URL, with a database and sign-in included, and a typical first build takes about an hour. It costs a flat monthly fee plus AI usage at cost, so check the pricing page and compare with what you pay for tools now.

Frequently Asked Questions

5 questions

What does RAID stand for in project management?

RAID stands for Risks, Assumptions, Issues, and Dependencies. Some organizations use the same letters slightly differently (for example, the "D" can stand for Decisions), but the four-part version above is the most common. If your company uses a variant, add a fifth tab rather than overloading the four.

Is a RAID log the same as a risk register?

No. A risk register tracks only risks. A RAID log covers risks plus assumptions, issues, and dependencies. Many teams find the wider view more useful because those four things interact: an unconfirmed assumption creates a risk, and a missed dependency creates an issue.

How many items should a RAID log have?

There is no right number, but you should be able to discuss every open item in a 15 to 20 minute weekly review. If you cannot, consolidate, close, or accept the low-scoring ones. A small program will have fewer open items than a large one, but only the top-scoring ones should reach the sponsor.

Who owns the RAID log?

One person keeps the log current, usually the project or program manager. That person does not own every item. Each entry has its own owner who is responsible for the next action. The log owner's job is to chase updates, keep the format clean, and run the review.

Can AI keep my RAID log updated automatically?

AI can draft entries from notes and review the log for gaps, but it cannot know what changed in the real world unless you tell it. Use it to save time on writing and checking, and keep a person responsible for owners, dates, and decisions.

How we source, write, and correct our articles. Editorial standards

Related articles

Straight to your inbox.
New pieces as they publish. No noise.