DecisionOS — Decide · module

ROI

What this initiative is actually worth, as a range, and which single assumption the answer depends on.

When you reach for it

The week before a funding request, or when finance asks for the number and the honest answer is that nobody has built one.

What it asks you

5 blocks, in this order. Each one exists for a stated reason, and the run tells you that reason as you answer it.

01
The outside view
Asked first, and deliberately before anything else. Forecasts improve when the forecaster starts from how comparable past efforts actually turned out rather than from the specifics of this one. Optional, and worth two minutes.
02
The benefit mechanism
There is no generic benefit. The mechanism determines the arithmetic, and it determines which questions are worth asking next.
03
The decomposition
People estimate quantities they cannot judge holistically far better when the quantity is broken into parts they can judge. Every figure is low, expected and high — a single value is permitted, and is recorded as a single value.
04
The costs
Structured lines rather than a total, because the line that is missing is never the one anyone would have noticed missing.
05
Timing
Benefit that starts in month nine is worth less over three years than benefit that starts in month one, and business cases routinely forget it.

The method

How the result is produced

The benefit mechanism you chose determines the arithmetic, and the decomposition is multiplied out three times — once with every estimate at its low value, once at expected, once at high. Costs are summed by line, with run and supervision costs annualised across a 3-year horizon, and benefit is reduced for the months before it reaches its expected rate. Sensitivity is one-at-a-time: each estimate is moved to its low and then its high value with every other estimate held at expected, and the resulting swing in net benefit is ranked. The largest swing is the assumption the case depends on, and it is then solved by bisection for the value at which net benefit is exactly zero. This is deliberately not a Monte Carlo simulation: simulating correlated distributions would produce a smooth curve that looks authoritative and encodes independence assumptions nobody entered.

What this is not

These are your estimates, not measurements. Nothing here was observed, sampled or verified.

This is not a forecast with a confidence interval. The range is the range you gave, propagated — no distribution was assumed and no simulation was run.

This is not an approval. It is the arithmetic and the assumption it rests on, in a form a CFO can argue with.

Where it stops

Where you estimated a quantity, the result is only as good as the estimate. A narrow range is not the same as a confident one.

There is no peer comparison behind any statement in this report, and none is implied. We will not show one until there is enough real data to state an n.

A narrow range is not the same as a confident one. Where estimates were entered as single values, the true range is wider than the one shown.

Figures left blank are counted as zero rather than estimated, which flatters the case. The confidence grade names how many.

One-at-a-time sensitivity moves each estimate independently. Where two estimates are actually correlated — volume and cost usually are — the real swing is wider than the tornado shows.

Amounts are in your organisation's own currency; no conversion is applied anywhere and figures from different currencies are never summed.