Every risk on this job was already on paper.

That's not risk identification. It's a memory test, marked by the people sitting it.

And the problem isn't the facilitator. A room can only give you what's in the room. The rock in the pile bores wasn't in anybody's memory. It was in a document, in a folder, that four of the people at that table had access to and none of them had opened since the tender.

Six documents, in order of yield

Risk identification on a construction project isn't a creative exercise. It's a reading exercise, and the reading list is always the same.

They're worth doing in order, because the order is roughly how much risk each one gives you per hour spent.

The ground investigation report. Boreholes, trial pits, laboratory results, and the sentence at the front where the geotechnical engineer tells you how confident he isn't.

The tender query log. Every question you asked the client before you priced the job, and every answer you got back.

The drawing register. Not the drawings — the register. Revision status, issue purpose, date.

The subcontract enquiries. Specifically the boundaries: what package A stops doing and package B starts.

The programme. Every activity where somebody typed a duration instead of calculating one.

The site and the statutory file. Access, working hours, neighbours, consents, diversions.

None of these are risk documents. That's exactly why they work. Nobody was editing them to look good in front of a client.

The borehole log: five holes and a shrug

Open the ground investigation report for this job. Turn to the borehole records.

There are five. On a site with forty-two piles.

Two of them hit rock — one at 8.2 metres, one at 9.7. The other three were terminated in dense gravel above that depth, which tells you nothing about what is underneath. And on page three, in the paragraph everybody skips, the geotechnical engineer writes that the investigation was limited by access and that conditions between boreholes should not be assumed to be uniform.

That paragraph is a risk. It is the geotechnical engineer, in writing, telling you he doesn't know.

Now look at what else is in the same report. The groundwater standpipe readings, taken in August, showing a water table below formation level. Nobody dug in August. The piling starts in February, and nobody has asked what that standpipe reads in February.

Two risks, from one document, in twenty minutes of reading. One of them was already on the register. The other wasn't.

The query log, line by line

This is the highest-yield document on the list, and almost nobody reads it after the tender is submitted.

The rule is simple. Read every answer, and mark every one that doesn't actually answer anything. On this job there are four.

Query 9 asked whether the client had secured the wayleave for the site access road. The answer was to be confirmed. It was never confirmed. The tender was priced on the assumption that lorries can reach the site.

Query 14 asked for the extent of rock in the pile bores. The answer was contractor to satisfy himself.

Query 22 asked which reinforcement drawings were for construction and which were still for tender. The answer listed four drawing numbers and said the rest were subject to review.

Query 31 asked who was providing temporary power before the substation was energised. The answer was by others.

Four phrases, four risks, all of them written down and signed before a single tonne of concrete was ordered. This isn't a filing problem. It's the client transferring uncertainty to you in a document you then filed and stopped reading.

The gaps between two subcontracts

Here's a category that never comes up in a workshop, because it isn't a thing that happens — it's a thing that doesn't.

Lay the piling enquiry next to the substructure enquiry. The piling contractor's scope ends at the top of the pile. The substructure contractor's scope starts at the pile cap. Between those two sentences sits pile trimming, and neither of them priced it.

Do the same for the waterproofing package against the blinding. For builder's work in connection with the mechanical package. For testing and commissioning, where every enquiry assumes somebody else is providing the temporary water.

Four gaps on this job. Each one is a risk with a very specific shape: it will surface as a variation request from a subcontractor at exactly the moment you have no leverage.

Nobody in a workshop names these, because no individual is thinking about the seam between two other people's scopes. You only find them by putting two documents side by side.

Five becomes twenty-two

Read the six documents properly and the register on this job doesn't grow a little. It roughly quadruples.

Two from the ground investigation. Three more from the query log — the rock turns up there as well, but a risk only counts once no matter how many documents it appears in. Three from the drawing register — the reinforcement drawings still at revision B, the embedded items nobody has coordinated, the rebar clash between the core wall and the transfer slab. Four from the subcontract seams. Three from the programme, where durations were typed rather than built. Four from procurement and the market. Three from site, access and statutory consents.

Twenty-two.

Of those, exactly five were already on the register. Seventeen were sitting in documents the company already owned, before the job started, at a cost of zero.

And notice what this list is grouped by: documents. Not categories. Right now it's a pile — and a pile of twenty-two is harder to manage than a list of five, which is the honest reason most registers stay short.

Practical insight

You don't need a two-hour meeting. You need two hours on your own with the door shut.

Print the tender query log. Go through it with a highlighter and mark every answer containing to be confirmed, contractor to satisfy himself, subject to review, by others, or as required. Those five phrases will find you more risk in twenty minutes than any workshop you have ever sat in.

Then open the drawing register and count the drawings still marked preliminary or for tender that you are already building against.

Then take the two largest subcontract enquiries and read only the exclusions. The gap between two sets of exclusions is where your next variation is coming from.

If you do nothing else from this track, do this one. It costs an afternoon, and it is the highest-return two hours available to a planner.

Key takeaways

✔ A workshop can only produce what is already in the room. It is a memory test, not an identification process.
✔ Six documents carry almost all the risk on a construction project, and none of them are risk documents.
✔ The sentence where the geotechnical engineer admits uncertainty is itself a risk. Read page three.
✔ Five phrases in a query log find more risk than a two-hour meeting: to be confirmed, satisfy himself, subject to review, by others, as required.
✔ Scope gaps between two subcontracts are invisible to individuals — they only appear when you read two enquiries side by side.
✔ On this job, reading six documents took the register from five lines to twenty-two.
✔ Seventeen of those risks were already owned, written down and paid for before the job started.

What's coming next

Last week we found that the contingency was built on a five-line list nobody had checked. This week we read the documents that were sitting there the whole time, and the list became twenty-two.

But twenty-two items in the order you happened to find them isn't a register. It's a pile. There's no way to tell whether you have covered the whole job or only the parts you thought to look at, and no way to spot that an entire area of this project has nothing written against it at all.

Next week we give the pile a structure — the same thing a WBS does for scope, done for everything that can go wrong. And that structure will show us something uncomfortable: even at twenty-two, there are two branches of this job with nothing in them.