A response with no price is a wish.

So this week the register has to stop describing the project and start changing it.

Four things you can do, and a fifth

There are only four things you can do to a threat, and they are worth naming by what they do to the numbers rather than by the words in a textbook.

Avoid drives the probability to zero, and on a construction project it almost always means changing the design or the method. Move the building off the rock. Use a raft instead of piles. It is the most powerful response and the one least available, because by the time the register exists somebody has usually already bought the design.

Reduce moves the probability, the impact, or both. Order the switchgear early. Hire a standby pump. Put a weekly slot in the diary for design release. This is where most of the work is.

Transfer moves who pays without touching whether it happens. Insurance, a bond, a subcontract clause, a term in the main contract. It looks like the cheapest response on the page and it is the one most often misunderstood, which is why the next two weeks are about nothing else.

Accept means hold the money and get on with it. That is a legitimate answer, and for genuine variability it is usually the only honest one. You cannot manage your way out of February.

And then there is the fifth, which does not appear on the standard list and which Week 8 earned us. For an epistemic risk — one where there is an answer today that nobody has gone and got — the response is not any of the four. It is investigate. Drill. Ask. Write the letter. Nine of the fourteen risks on this job are in that category and between them they carry $63,880 of exposure.

Priced, owned, dated

Whatever the strategy, three things have to appear next to it or it will not happen.

A price, because a response is work and work costs money. If the cost is genuinely zero, write zero — but write it, because a blank means nobody has thought about it.

An owner, and it must be a person rather than a department. “Commercial” is not an owner. “The QS” is not an owner. A name is an owner.

A date, which is not the same as the closing date from Week 4. That was when the risk stops being possible. This one is when the response stops being possible — and it is almost always much earlier. You cannot investigate the ground after the piling rig has left.

Without those three, what you have written down is not a response. It is a wish with a reference number.

The boreholes, properly

Week 8 said two boreholes in the north would cost about $6,800 and would collapse a large part of the biggest exposure on the job. It is time to check whether that was true, because it is easy to say and it is a real six thousand eight hundred dollars.

Lay out the decision. Two choices, two states of the world.

If we do not investigate, we find out during piling. Rock is discovered bore by bore, the rig stands while a heavier one is brought in, and the sequence gets rearranged around it. That is the $2,850 per affected pile we have been using since Week 5.

If we do investigate, we know before the rig arrives. The right equipment comes once, the affected bores are planned into the sequence rather than interrupting it, and the cost per affected pile falls to about $1,900 — plus a mobilisation of $6,000 if the rock runs into the north, or $4,000 if it turns out to be confined to the south.

Put a probability on the state of the world. Given two of five southern holes found rock, and given the geology, call it 55% that it extends north.

Now run it. Three hundred thousand times, because the number of affected piles is itself uncertain in both branches.

Expected cost if we react: $34,323. Expected cost if we investigate first: $34,778.

Investigating is $456 worse.

Expected value says no

That is a genuinely awkward result and it would be dishonest to bury it. On the arithmetic that most people are taught to use, the boreholes do not pay for themselves. The saving from planning instead of reacting almost exactly cancels the cost of finding out.

Take that to a finance director and the answer is no, and the answer is defensible.

Now look at the rest of the distribution.

At P80, reacting costs $53,431 and investigating costs $48,397. Investigating is $5,033 better. At P95 it is $67,532 against $57,878, which is $9,654 better. The standard deviation falls from $19,226 to $13,649, down twenty-nine percent.

So the boreholes do not buy a lower average. They buy a shorter tail.

And that is exactly what a contractor is in the market for, because of what we established last week. Nobody funds a project at its mean — you would run out half the time. The company holds money against a percentile. If the percentile is where the money is held, the percentile is where responses have to be judged.

A response appraisal done on expected value alone will reject the best trade on this project. It rejects it correctly, by its own logic, and the logic is looking at the wrong part of the curve.

Every response creates something new

Two things get missed on almost every register.

The first is residual risk. The response does not empty the line. Drill the three holes and you still do not know what is between them — you know a great deal more, and something is left. That remainder stays on the register with a smaller number against it, and the smaller number is what the contingency now covers.

The second is secondary risk, and it is the one that bites. A response is work, and work has its own risks.

Take the switchgear. The response is to order early, which reduces the delivery risk and costs $3,200 in early payment. But ordering early means committing to a specification before the electrical design is frozen. If the load schedule changes in March, you own a large piece of switchgear that is wrong, and nobody will take it back.

That new risk did not exist last week. It exists because of a decision we just made, and it belongs on the register with the same cause, event and effect discipline as everything else. The rule is simple: when you write a response, write the risk it creates. If you cannot think of one, look harder.

Practical insight

Open your register at the response column and read only that column, with the risks covered up.

Count how many entries are verbs with a person and a date attached — Ahmet to issue the wayleave letter by 14 March. Then count how many are nouns and intentions: monitor, manage closely, liaise with client, ongoing review.

On most projects the second list is longer, and every entry on it costs nothing, changes nothing and can never be marked complete. That is the tell. A real response has a completion date, and on that date somebody either did it or did not.

Then take your three largest risks and price the response properly. Not the impact — the response. If the price is more than the exposure, say so and accept the risk instead, deliberately. Deciding to do nothing is a decision. Not getting round to it is not.

Key takeaways

✔ Four responses to a threat: avoid, reduce, transfer, accept. Each does something different to the numbers.
✔ Epistemic risks get a fifth — investigate. Nine of the fourteen risks here, carrying $63,880.
✔ A response needs a price, a person and a date the response stops being possible. Without those three it is a wish with a reference number.
✔ The boreholes cost $6,800 and are $456 worse on expected value. On the arithmetic most people use, the answer is no.
✔ At P80 they save $5,033, at P95 $9,654, and they cut the standard deviation by 29%.
✔ Money is held at a percentile, so responses must be judged at a percentile. Expected value alone rejects the best trade on the job.
✔ Every response leaves residual risk and creates secondary risk. Ordering the switchgear early buys a delivery date and sells a specification freeze.

What's coming next

Four of the five responses are things we do. The fifth is different, and it is the one this project has been quietly relying on since Week 1 without ever writing it down.

Transfer does not reduce a risk. It decides who pays for it — and that decision was made before any of us arrived, in a document nobody on this register has opened. Week 3 found two empty branches on the risk breakdown structure and diagnosed why: every document we read was about the physical works, and the contract was never on the reading list.

Next week we open it. Three standard forms, one clause, and the difference between them is whether the rock under those forty-two piles is our problem or the client's.