Most report errors are not wrong facts. They are two right facts from different dates, sitting in two documents that nobody compared. Here are the checks to run before every pack goes out, in the order that catches the most.
The ten checks
- Same period, same story. Put the Status Report, the Milestone Report and the RAID log side by side. The overall colour, the go-live date and the top three risks should match in all three.
- Dates against the plan. Every milestone date in the report is the one in the current plan. Not the one in last month’s report.
- Reds have plans. Every amber or red has a problem, a plan, a date and a named owner. If one is missing, the colour is not earned.
- Owners are people. No “TBC”, no “PMO”, no “vendor”. A name, and that person knows.
- Roll-ups add up. The overall RAG is not kinder than the worst item under it. The budget total is the sum of the lines. Percent complete comes from the rows.
- Decisions have outcomes. Every decision from the last meeting has an action and an owner. Every decision needed this time has a deadline.
- What changed, and why. For every number or date that moved since last period, there is a reason in plain words.
- Nothing is missing. The sections the reader expects are there. An empty section says “nothing to report”, not nothing.
- Adjectives have numbers. “Minor slippage” is a number of days. “Broadly on budget” is a percentage. Find it or cut the adjective.
- Someone else reads page one. Five minutes from a colleague who was not in the data. If they ask a question you cannot answer from the pack, the pack is not ready.
Do them in this order, every period, and write down what you found. A short note of “checked on [date], two mismatches fixed” is worth more to a reviewer later than a perfect pack with no trace of how it got that way.
Checks one and three lean on the two documents most packs get wrong. The rules for each are in What is a RAID log, and how do I keep it honest? and What should a SteerCo pack include?
A real example
A Status Report went out saying the programme was amber with “no critical issues”. The same week’s RAID log had a red issue: a supplier had missed a contract milestone. The PM had updated the log and not the report. A board member read both.
Nothing in the report was invented. The two documents were just from different moments. Check one would have caught it in a minute. If part of the pack was drafted by an AI tool, add the five-minute routine in Can AI make up numbers in a project report?
Where PMOpack fits
PMOpack runs a version of checks one, three, four, five and eight as fixed rules, inside the pipeline, every time a pack is generated. It compares the artefacts in the same period with each other, flags an amber or red without a path to green, flags a placeholder owner on a decision, and checks roll-ups against their rows. A critical finding blocks publish until a person fixes the data or accepts it with a written reason, and the acceptance is stored and cannot be edited later. A check skips, and says so, when the partner document is missing. It checks figures for agreement with each other, not against your finance system. Checks two, six, seven, nine and ten are still yours. The free scan on the home page is a completeness readout only: no AI, nothing saved, and it does not run the conflict checks.
Questions people ask
- What is the most common error in a project report?
- Two documents from the same period that disagree: a Status Report that says amber while the RAID log holds a red issue. Neither is invented; nobody compared them.
- How long should pre-issue checks take?
- About twenty minutes for a standard pack if you do them in a fixed order, every period.
- Can a tool do these checks for me?
- Some of them. Agreement across reports, roll-ups, missing owners and missing plans can be checked by fixed rules. Judgement checks like 'what changed and why' stay with you.
- What does PMOpack's free scan check?
- Completeness only: which expected sections and fields are present in one file. No AI runs, nothing is saved, and it does not run the cross-artefact conflict checks.