Somebody felt forty percent.

That is a real piece of evidence and it beats a guess. It is also, on closer inspection, an answer to a question nobody asked.

Where the five holes were

Open the site plan and put the boreholes on it.

There are forty-two piles, spread across the whole footprint. There are five boreholes, and every one of them sits in the southern third of the site, within about thirty metres of each other, along the strip of existing hardstanding.

That isn't a coincidence or a mistake. It's where the drilling rig could get to without building a temporary road. Ground investigations are priced too, and the cheapest investigation is the one that doesn't need access works.

So two of five hit rock. What that tells you, precisely, is that rock exists under the southern strip of this site, at around eight to ten metres, in the area the rig could reach.

About the northern half of the site, where fourteen of the forty-two piles are going, it tells you nothing at all. Not “probably fine”. Nothing.

Forty percent of what?

Here is the question that should be asked of every probability on every register, and almost never is.

Forty percent of what?

There are at least three completely different statements hiding inside that number, and they have wildly different consequences.

It might mean there is a 40% chance we encounter rock anywhere on this site. But we already have. Two holes found it. On that reading the probability isn't 40%, it's 100%, and it stopped being a risk the day the report landed.

It might mean 40% of the piles will hit rock. That's not a probability at all, it's a quantity — and it's a quantity somebody has estimated without saying so.

Or it might mean on four jobs out of ten like this one, rock turns out to be a problem. That's a reference class, and it's a perfectly good way to think — but it isn't what two-out-of-five boreholes measures.

The register said 40%. It never said 40% of what, and the three readings are worth different amounts of money.

The number that was already inside the impact

Now do the same to the other column, because this is where it gets expensive.

The impact was $60,000. Where does $60,000 come from?

Work it back. Rock in a bore means a heavier rig, two extra days on that pile, and a knock-on to the cap sequence. Price that out and it lands at roughly $2,850 per affected pile. Divide $60,000 by $2,850 and you get twenty-one piles.

Twenty-one. Exactly half of forty-two.

So the estimator had already made a quantity judgement — half the site — and buried it inside the impact figure without writing it down. Then the register multiplied that figure by 40%.

$24,000. Which, at $2,850 a pile, is money for eight piles.

The same guess got made twice and discounted twice. Nobody did anything dishonest. Two people each made one reasonable judgement, in two different columns, and the columns multiplied.

What the evidence actually supports

Take the only hard information there is. In the tested area, two of five locations have rock above toe level. If that ratio held across the site — and there is no evidence either way for the north, which is the whole problem — then around seventeen of the forty-two piles are affected.

Seventeen piles at $2,850 is $48,450.

The register is carrying $24,000. The gap is $24,450, against a margin of $48,163.

And that's the central case, not the bad one. The bad one is that the north is worse than the south, which is entirely possible, because nobody has been there.

High, medium and low mean nothing

Which brings us to the scales themselves.

Most registers score impact as high, medium or low, and the words do no work at all. High to a project manager on a $1,000,000 job and high to the same person on a $40,000,000 job are different by a factor of forty, and the register does not know which one it is looking at.

A scale has to be anchored to money and days, and written down where everybody can see it. That part is uncontroversial and most people who bother do it.

The part almost everybody gets wrong is what they anchor it to.

Anchor to the margin, not the contract

Open any impact scale you have ever been handed and it will be a percentage of contract value. Low is under 1%. Medium is 1 to 5%. High is over 5%.

On this job, 5% of contract value is $50,000.

The margin is $48,163.

So a risk scored as medium on the standard scale is a risk that takes every penny the company was going to make. The scale isn't just imprecise, it is actively pointing the wrong way, and it will keep telling you things are fine right up to the point where they aren't.

Anchor the bands to the margin instead and the same numbers change meaning completely. Negligible is under 5% of margin. Critical is anything that takes all of it. Suddenly the rock, at $48,450, isn't a medium anything. It is the job.

Practical insight

Two things, and neither takes an afternoon.

First, go through your register and write, next to every probability, the words of what. Then finish the sentence. Half of them will turn out to be quantities pretending to be probabilities, and a few will turn out to be things that have already happened.

Second, find your impact scale and check what it is a percentage of. If it is contract value, work out what your medium band is in real money, and compare it to the margin on the job. On most construction projects the answer is uncomfortable, and it takes ninety seconds to find out.

Then go and find out where the boreholes were. Not how many — where. A sample taken where the rig could park is a sample of the car park.

Key takeaways

✔ Two of five boreholes is real evidence, but only about the ground the rig could reach.
✔ All five holes sat in the southern third. Fourteen piles in the north have zero coverage — not “probably fine”, zero.
✔ Every probability needs the words of what after it. Chance of encountering rock, proportion of piles, and frequency across similar jobs are three different numbers.
✔ The $60,000 impact already contained a quantity assumption: twenty-one piles at $2,850, or exactly half the site.
✔ Multiplying it by 40% discounted a judgement that had already been made once. The register held money for eight piles.
✔ The evidence supports around seventeen affected piles, or $48,450 — a gap of $24,450 against a margin of $48,163.
✔ Anchor impact bands to margin, not contract value. At 5% of a million, a “medium” risk takes the entire profit.

What's coming next

So the scales are anchored, the probabilities have been asked what they are probabilities of, and the fourteen risks each have a number that can at least be defended in a room.

The obvious next step is to multiply them together, plot them on a five-by-five grid, colour it red, amber and green, and present it. Every organisation does this. It is the single most recognisable artefact in risk management.

Next week we take our fourteen risks and put them on that grid — and watch three risks worth $12,000, $30,000 and $60,000 land in exactly the same box.