What the model was asked
An examination answers one question at one date. The question has to
be written down before the model runs — otherwise the answer is
whichever number the output happens to flatter.
Here the question is a condition, written down before the model runs
rather than chosen after it: “Cash balance below zero at month-end”.
One line, one date — the Anchor, 2027-07-31, nineteen months into a
plan that begins on 1 January 2026. Everything else the model
produces is scenery; this is the thing it has to answer.
That is deliberately narrow. The examination does not ask whether
DemoCompany is a good company, whether the plan is ambitious, or
whether the team can sell. It asks how often, across a thousand runs
of the same business, the cash balance is below zero when the month
closes on that date.
What came back
The condition was met in 32% of the runs at the Anchor — 324 runs
met it, 676 did not.
Read it plainly: one future in three closes that month with an empty
account. That is not a doomed company and it is not a safe one — it
is a plan with a one-in-three failure mode sitting inside it,
nineteen months out, visible now while there is still time to act on
it.
No plan in a spreadsheet can show that. Not because the arithmetic
is harder — because a cell holds one value, so the file holds one
future: the one somebody typed in. Change three assumptions and it
does not put a second outcome beside the first, it replaces it. The
other futures never appear, which is not the same as not existing.
The plan is one of those futures — a real one, and the one worth
writing down first. The 324 where the account is empty at the Anchor
came out of assumptions nobody could have ruled out either, and they
were there before anybody modelled them. What changes is that they
now show up on the radar while the decision is still open, and
specific enough to act on: which hire, which threshold, which month.
For an investor that is a measure of how much of the plan's value
rests on things going better than the middle case. For a founder it
is a figure to argue with instead of a mood to manage.
What separates the two cohorts
Split the runs into the 324 that met the condition and the 676 that
did not, then compare the two groups input by input. What comes back
is a ranking of the inputs that took visibly different values on
each side of the line. Two stood clear of the rest.
The daily signup rate is first. The median run draws 0.46 signups a
day and sits at the overall 32% rate. Among the runs that ran above
0.64 signups a day, only 15% met the condition — the frequency
roughly halves. It is also the harder of the two to move from inside
the company: the lever behind it is top-of-funnel, which means
acquisition spend, partnerships, and word of mouth.
Free-to-paid conversion is second. At the bottom of the sampled
range, 3% a month, 69% of paths met the condition; above 8% a month
it falls to 19%. This one is internal — onboarding, activation, and
how the tiers are priced — which makes it the cheaper of the two to
work on even though it separates the cohorts less sharply.
Three more inputs separate the cohorts behind those two, and one of
them is not a growth number at all. The forensic report ranks all
five and puts a threshold on each.
What this does not say
This is the section to read twice. The comparison behind those
numbers is observational. It takes the runs the model produced,
sorts them by what happened at the Anchor, and reports how the two
groups differed. No intervention was tested: nothing was held fixed
while one input was moved and the rest of the future left alone.
That distinction decides what the numbers can be used for. A driver
can travel with the outcome without causing it. The runs with low
signup rates are also the runs where revenue arrives late, where the
raise is worth less by the time it lands, and where every new hire
meets a thinner base. The signup rate marks all of that at once, and
telling the marker apart from the mechanism takes a different test
than this one.
So the honest reading of the first driver is this: runs in which
signups ran above 0.64 a day were runs in which the condition was
met far less often. The reading it is not: push signups above 0.64 a
day and the risk halves. The second may well be true. This
examination does not establish it, and a report that implied
otherwise would be selling a conclusion it had not earned.
Establishing it is separate work — the same futures cloned, one
input changed, everything else left where it was. That step belongs
to an engagement, with a human deciding which change is worth
testing and what the result means. It is not what this page shows.
The forecast that survives a slipped month
Ask a founder what happens when a September order lands in October.
The answer is usually a new version of the plan — a rebuilt sheet, a
fresh set of numbers, and a story explaining the gap. The order
slipping was never the surprise. Rebuilding the plan around it is
what makes it look like news.
That does not happen here. The possibility that the order slips is
already in the model, so the month it lands in is one of the things
a thousand runs disagree about. A probable event arriving does not
call for a new plan; it narrows the one that exists.
Which is why the forecast half of the pair is not a projection. Its
P&L and cash-flow statement are one of the thousand runs — a
single coherent future, picked because its cash balance sits in the
middle of them all — and the report draws around it where the other
999 went. That middle future ends the horizon with €168k in the
bank. The bottom fifth of runs end €98k in the red, the top fifth
with €568k. Revenue growth runs at 14% a month in the middle, and
between 10% and 18% across those same fifths.
The range also says where to look. Insolvency risk peaks in July
2027, at 32% — which is the month the forensic study takes apart,
and the reason the Anchor sits where it does.
And because the range is written down before the months happen, it
can be checked once they have. Actuals land somewhere inside it:
near the middle, or out toward an edge. That is a sharper question
than whether the company hit its plan, and an investor can ask it
every quarter.
Where the rest of it is
Every number above is read out of that pair, and the pair is what a
client receives at the end of an engagement. DemoCompany's copy of
it is the sample.
Request the sample report and
both documents arrive by email.
