Certamus · Method

How we work a case

Reduce the list to its causes, set the rule before the analysis, size what the case leaves out, and show the ranking rather than the decimals.

Count the incidents, then stop counting them

A case brief usually hands you a pile of events. The instinct is to answer each one, and a deck with seventeen fixes in it is the result.

SEVENTEEN RISKS, AS THEY ARRIVEDO1O2O3O4O5O6O7F1F2F3F4F5C1C2C3R1R2FOUR GROUPS, AND ONLY THREE ARE FAULTSRisk that was never priced6 of 17153 crSingle points of failure5 of 1778 crBlind spots5 of 1740 crMarket pressure, which failed our own test1 of 1720 crMarket pressure had no internal cause, so it was left outside rather than forced into a fourth box.
Seventeen incidents from the Sankriya brief, grouped by the internal fault that made each one worse. Three groups held. The fourth had no internal cause and was left out of the answer rather than rounded up to a tidy number.

That deck loses. It costs more than the problem, and the reason each event hurt is still sitting there untouched. So the first pass is to line the incidents up and ask what is common underneath them, then present the causes rather than the list. Three fixes that address why the damage kept happening beat seventeen that address what happened.

Set the test before you go looking

Grouping symptoms into causes is exactly the kind of analysis that finds whatever shape you wanted. The guard against that is a rule written down before the grouping starts.

One that has worked: a cause only counts if fixing it would have made at least three of the past incidents shorter or cheaper. Candidates that fail the rule get left out rather than argued in, even when they are interesting. Competitive pressure has been left out on exactly those grounds.

Size what the case does not give you

Briefs routinely omit the financials the recommendation depends on. The options are to avoid the number, or to build it in the open.

Build it. State the assumed revenue and margin, show the arithmetic step by step on the slide, and say plainly that the finding is the ranking rather than the rupees. Then test whether the ranking survives the guesswork: if the top of the list only changes when an input moves by a third, say so, because that is the fact a panel actually wants.

Count the money once

One incident can belong in several categories at once. A systems breach is an operations problem, a compliance problem and a reputation problem simultaneously.

Score it in each, because the exposure is real in each. Count its cost once. Totals that double-count are the fastest way to lose a panel that is paying attention, and they are very easy to produce by accident.

Separate a one-off from a pattern

A strike is bad luck. A site that flooded twice in one year is not. Treating those the same way either inflates the plan or misses the thing that is actually going to recur.

The distinction is usually available in the brief, and it changes which fixes earn their cost.

Say the weak part out loud

Every one of these steps produces something a panel can push on: an assumed revenue base, a grouping rule, a sensitivity that moves. Those go in the talk, in our own words, before anyone asks. That is its own habit, and it is set out in say the weak part first.

Set the test before you do the grouping

Any pile of incidents can be sorted into three tidy causes if you are willing to push hard enough, and a panel can tell when you have.

So fix the rule first and let it fail. In Sankriya a fault only counted if fixing it would have made at least three past incidents shorter or cheaper. Competitive pressure failed that test, so it was left outside the answer rather than made into a fourth box, and the slide said so.

Announcing the test before the result is what separates a classification from a rationalisation.

Decide what you are not doing

A brief usually contains more than can be answered in eight minutes, and the teams that try to cover all of it produce a deck of headlines with nothing underneath.

Choose, and say you chose. Naming the part you deliberately left out costs one sentence and buys the panel's trust for everything you did cover, because it demonstrates the judgement they are marking.

Write the recommendation before the analysis is finished

Not to lock it in. To find out what the analysis needs to support.

A draft recommendation written on day one tells you exactly which numbers are load-bearing, and therefore which of the twenty possible analyses are worth the time. Without it, teams analyse what is available and discover at the end that none of it decides anything.

It will change. That is the point: when it changes, you know something.