Somebody read the six-section assessment and produced this for the people who were not going to read it. It is the length a summary should be and it reads like one. Nine statements. Five of them should not go out.
Read the assessment first. Once, at normal speed. Then read the nine numbered statements below and mark every one you would not be willing to send to the Board under your own name.
Write your marks down before you scroll past the summary. The debrief underneath is worth nothing if you read it first, and the answer key underneath that stays hidden until you press a button, which is the only protection available.
To: Board
From: Strategy & Commercial
Date: 22 April
Subject: Eastern Region berthing — one-page summary of the 14 April
assessment
Full assessment attached. Section numbers in the assessment correspond to the order above.
There are three factual errors and two unsupported inferences in those nine statements. Five bad statements out of nine, in a document that reads perfectly well.
Here is the prediction this material was built to test, and you can now check it against what you actually wrote down.
The three factual errors are all contradictions. A number, a source name, a consent status — each one sits next to something in the assessment that says otherwise, and finding them is a matching exercise. Tedious, but the method is obvious and it terminates.
The two unsupported inferences are not contradictions. Nothing in the assessment disagrees with them. They are built from parts that are individually true and joined by a claim that is not in the document at all — and there is no matching exercise that finds a sentence which isn't there. You have to be looking for a different thing entirely: not "does this conflict with the source" but "where in the source does this come from".
That is the whole lesson in one specimen. Checking a summary against a document catches the contradictions and is structurally blind to the additions. The additions are usually the ones that read best, because they are the sentences that sound like analysis rather than transcription — and they are the ones a Board will quote back at you.
Answer key. Before you press the button: write down your five, or however many you have. Then check. The gap between your list and this one is the thing you came here for — not the score.
Statement 2 — wrong source. The 14% demand figure and the eleven-month waiting list both come from the CMTC press release (assessment, section 3), not from the Office of Regional Statistics. This is the expensive one. The assessment states in section 2 that ORS does not count berths, moorings or occupancy at all, so the attribution is not merely sloppy — it moves a self-selected trade-body number under the name of a compulsory statistical return, which is exactly the swap that makes a weak figure survive a Board meeting. "Member marinas" should have given it away: a statistical agency does not have members.
Statement 6 — wrong number. Harbourview's blended average rate rose 7.6% to £3,390 (section 5). £3,600 is CMTC's regional average, and the assessment says in the same section that the two are not comparable, because CMTC's basis is premium coastal marinas and excludes the river and tidal moorings that make up most of Kingsmill. Both numbers are real and both are in the document; they have been welded together. Note that the sentence does not even hold internally — 7.6% on £3,150 is £3,390, not £3,600 — so this one was catchable without the source at all.
Statement 8 — wrong status. The 48 berths have planning consent, granted in November. The marine licence is still outstanding, and what is expected in the first quarter of next year is the regulator's decision, not the berths. Construction is a further fourteen months from licence, with no start date set or budgeted. The assessment spells this out: "consented on land and unconsented on water." The summary has turned a decision date into an opening date.
Statement 5 — the causal link is invented. Every component is true and traceable. Kingsmill is four nautical miles up-river (section 1), tidal access is restricted to about five hours either side of high water (section 1), and occupancy is 81% (section 4). The word doing the damage is reflects. The assessment nowhere connects the location to the occupancy, and section 6 says so explicitly: "We have not investigated the difference, and no explanation for it is offered or implied here." The explanation is plausible. It might even be correct. It is not in the document, and a Board that accepts it will stop asking the question.
Statement 7 — the conclusion is invented. The eleven-month waiting list is in the source (via CMTC, and see statement 2). Everything after the comma is not. "Customers are absorbing increases without resistance" and "there is room to move rates further" are claims about price elasticity, and section 6 states in terms: "We hold no data on customer switching between marinas and no basis on which to estimate elasticity. Nothing in this assessment supports a view on how far rates could rise before occupancy falls." This is the most dangerous sentence on the page, because it is the only one that recommends an action.
Statements 1, 3, 4 and 9 are accurate and properly sourced. Statement 3 is the one worth noticing: it carries the CMTC caveat forward faithfully, including the 38-of-210 response base and the 6% enquiry gap. If you flagged it, ask yourself why. It is the single most useful sentence in the summary, and a habit that flags careful sourcing alongside genuine errors will make you slower without making you righter.
The pattern to carry out of here. Errors contradict the source and you find them by matching. Inferences extend the source and matching cannot find them. For anything that reads like a conclusion, ask for the line it came from — and treat "there isn't one" as the answer, not as a failure of the search.