Rigour for strategic decisions
Companion № 03.3 · Simulate
№ 03.3
Companion
guide
A plain guide to the Simulate tool

Simulate, in plain English

A confident single number is the most comfortable way to be wrong. Here is what the tool does instead, and how to read every chart it draws you, no statistics required.
Ian Hargreaves Companion to: Simulate Reading ~6 min Annotated, plain-language

Open the simulator and, in effect, you are looking at thousands of possible versions of the same decision at once. That sounds heavier than it is. This is a tour of what the tool is actually doing, why it beats a single confident guess, and how to read each number and picture it gives back, in plain language, with every chart labelled.

We’ll use one small, everyday decision the whole way through, because the machinery is identical whether the stakes are a Saturday or a five-year contract. Once you can read it on the small example, the big one holds no surprises.

The running example

You’re thinking about taking a stall at the weekend market. It costs £150 for the pitch. You’ll make money if enough people stop, enough of them buy, and they spend enough each. None of those three things is knowable in advance, but you have a feel for the rough range of each. The question: is it worth giving up your Saturday?

iOne number is a comfortable lie

Asked to plan, most of us collapse every guess into one tidy figure. “About 150 people will stop, maybe a third will buy, they’ll spend around £12, so I’ll clear roughly £240.” It feels precise. It is the comfortable lie, because each of those inputs was really a range, and quietly picking the middle of every range throws away the one thing you actually wanted to know: how likely is this to go wrong?

A simulation refuses to throw that away. Instead of one value per guess, you hand it the honest range, and it keeps the uncertainty alive all the way to the answer.

iiSwap each guess for a range

For every uncertain input, you give three numbers instead of one: a centre (your best guess) and the two ends of a range you’d be surprised to fall outside. The tool calls that the 95% range, meaning you’d only expect to land outside it about one time in twenty.

One guess, told honestly
Fig. 1 · how an input works
90 150 220 PEOPLE WHO STOP → Your best guess the centre The 95% range low and high you’d be surprised to miss
You give three numbers, not one. “Probably 150 people, but it could be as few as 90 or as many as 220.” That honest spread is the raw material a simulation works with. You do the same for each uncertain input, here, how many buy, and how much they spend.

If you have no data at all and only judgement, that’s fine, turning a gut feel into an honest low/best/high is exactly what the companion Elicit tool is for. Either way, you end up with a range for each input.

iiiLet the computer play out the weekend, thousands of times

Here is the whole trick, and it really is this simple. The tool picks one plausible value from each of your ranges at random, say 130 people, 28 of every 100 buying, £14 each, and works out the result for that imagined weekend. Then it does it again with a fresh roll. And again. Twenty thousand times. (Statisticians call this a Monte Carlo simulation; the deeper article below uses that name.)

Each run is one believable version of your Saturday. Pile all twenty thousand results together and you get a picture of everything that could plausibly happen, not one guess, but the full spread, with the common outcomes piled high and the rare ones thin at the edges. That picture is the answer.

The picture the tool gives back
Fig. 2 · the outcome chart, annotated
£0 £150 £300 £500 PROFIT → The hump: the most likely outcomes where most imagined weekends land How high the curve sits = how often that result came up The long tail: rare, very good days YOUR TARGET MEDIAN · £240 Red = the weekends you fall short, about 1 in 4
£240
Median, the middle outcome
£30 to £460
Where 9 in 10 weekends land
25%
Chance you don’t clear £150
How to read it. Left-to-right is profit. The height of the curve at any point is how often that result turned up across the twenty thousand runs. The dashed red line is your target; everything to its left is shaded red, the imagined weekends that fell short. The three numbers below are the whole story: the typical result, the believable range, and the one that matters most: how often you miss.

Notice what the shape tells you at a glance. It isn’t a neat bell: it leans, with a long arm stretching to the right. That’s normal when you multiply uncertain things together, a few inputs landing high at once can produce a surprisingly big day, while the downside is capped. A single average would have hidden that lean completely.

ivWhich guess should you actually worry about?

Some of your three guesses barely move the result; one or two decide it. Before you fret over all of them, it’s worth knowing which is which, because that’s where it pays to do a little homework, ask around, or run a small test. The tool answers this with a second chart, the tornado.

Read it as: “If this one guess swung from its low end to its high end, and everything else stayed put, how far would the bottom line move?” One bar per input. The longest bar wins your attention.

Which uncertainty moves the money
Fig. 3 · the tornado, annotated
£0 £150 £300 £450 PROFIT FROM THE DAY → TARGET · £150 Average spend £120 £380 People who stop £160 £320 Share who buy £190 £300 Longest bar = biggest swing the input to pin down first
Where to spend your effort. Each bar runs the model twice, once with that input at its low, once at its high, holding the rest steady, so the bar’s length is that input’s own pull on the result. Here, average spend dwarfs the rest: nailing down what people actually spend would tighten the answer far more than counting footfall ever could.

This is the quietly useful part. A simulation doesn’t just tell you how risky a plan is, it tells you what to go and find out to make it less risky. Your next hour of homework should go to the longest bar, not to whatever was easiest to look up.

vThe one number that decides it

For a small, reversible choice, a single Saturday you can simply not repeat, the middle outcome is enough: £240 typical, sounds worth it, done. But the moment a decision is hard to undo, the number that matters is the red one: the chance you fall short. Twenty-five percent means one weekend in four ends below your target. Whether that’s fine or frightening is a judgement only you can make, but now you’re making it with honest odds in front of you (as good as the ranges you gave it), not a single hopeful average.

A simulation doesn’t tell you what will happen. It tells you the odds, which is what a real decision can actually use.

viMake it yours

The tool opens on a business example rather than a market stall, leads, a conversion rate, a deal size, and a target measured in the millions, but it is the very same machine: a handful of uncertain inputs, combined, with a target to clear. Clear the boxes, type your own low / best / high for each thing you’re unsure about, set the target that would make the decision a yes, and read the same three numbers back. If you’re working from judgement rather than data, build your ranges in Elicit first and bring them across.

Now try it on your own decisionTwo to five uncertain inputs, a target to clear, and the full spread of what could happen, computed in your browser, nothing saved or sent.
Open Simulate →

A note on honesty. A simulation is only as good as the ranges you feed it. It will happily turn confident-but-wrong guesses into a confident-but-wrong picture. The value isn’t false precision, it’s that the method keeps your uncertainty visible instead of quietly deleting it, and shows you where a little more knowledge would buy down the most doubt.

Want the deeper version, with the real business case and the maths running live? Read From sensitivity to simulation, the third piece in the Decisions under uncertainty series. And on why the gut mis-multiplies the ranges you feed in here, see Heuristics & biases.