Decision Cockpit

What this tool does

CM8-344 is a weighted decision matrix that also handles the four things a matrix normally hides: how risky each option is, how much you trust your own figures for it, how much of the score is coming from it simply being cheap, and whether the winner survives when somebody reasonable disagrees about what matters most.

That last one is the reason the tool exists. A weighted matrix is an argument written down, and its weakness is that the weights are set by the same people who prefer one of the options — usually after they have seen the scores. This tool re-runs the whole decision under four different weightings and tells you whether the answer holds.

Everything runs inside this single file: no account, no upload, no network request, which matters when the contents are competing quotations and a candid view of which supplier you do not trust.

The six dimensions

• Benefit or return — how good the outcome is if it works.
• Strategic fit — how well it matches where the organisation is going, and what it makes possible or forecloses. The dimension most often left out of a matrix and most often decisive afterwards.
• Risk — chance and severity of it going wrong, combined. Note the direction: this is the only scale where a high number is bad, and the tool inverts it before scoring.
• Cost — total over the period you care about, including running costs.
• Effort — internal person-days. Deliberately separate from cost, because an option that is cheap in money and expensive in the attention of the three people who could be doing something else is not a cheap option.
• Speed to value — how soon the benefit arrives.

Before scoring anything, anchor the scales: agree out loud what a 10 and a 3 look like for this decision. Without that, the numbers record enthusiasm rather than judgement, and the person who scores most generously wins.

How cost and effort are scored

Cost, effort and time are compared within each decision: the cheapest option scores 10, the dearest scores 0, and everything else sits proportionally between them. Absolute amounts never enter the score.

This has a consequence worth understanding. Adding a wildly expensive option to the list compresses every other option's cost score towards the top, which makes cost look like less of a differentiator than it is. If you want a decoy option in the list for negotiating purposes, keep it out of the matrix.

Because the comparison is per decision, one file can hold several decisions safely — as long as the decision name is spelled identically on every option belonging to it.

Confidence

Confidence is not a criterion; it is a discount. An option scoring 8 on figures nobody has checked is not the same proposition as an option scoring 7 on measured ones, and by default the tool scales the score accordingly.

This is what stops the best-sounding option from winning simply because nobody has looked into it yet. In the sample data the cheaper supplier is genuinely attractive and scores 5 for confidence because nobody has visited a reference site — and that alone costs it the top position. The correct response is not to argue with the arithmetic; it is to go and visit the reference site, then raise the score.

Turn scaling off only when every option has been researched to the same standard.

The sensitivity test

Four weightings are applied to every option, and all four are defensible positions a reasonable colleague might hold:

• Balanced — roughly even, weighted slightly towards benefit.
• Value-led — benefit and fit dominate; cost barely matters.
• Cost-led — cost and effort dominate.
• Risk-averse — safety dominates.

An option that wins under all four is a genuinely strong choice, and the tool says so. An option that wins under one is a preference with arithmetic attached — which does not make it wrong, but it does mean the real decision is about the weights, and that is the discussion to have.

The most useful outcome is often that different options win under different weightings. That is not a failure of the method; it is the method telling you the truth, which is that the group has not yet agreed what it is optimising for. No amount of further scoring will resolve that.

Reversibility decides how hard to think

Reversibility is recorded and deliberately kept out of the score, because it does not tell you which option to pick — it tells you how much analysis the decision deserves at all.

An easily reversed decision should be made quickly and corrected if it turns out wrong; spending three weeks and a committee on it is itself the mistake, and the cost of the delay usually exceeds the cost of being wrong. A decision that is effectively permanent deserves the opposite treatment: more evidence, more challenge, and a genuine attempt to find a way to make it reversible before committing.

Ask of any hard-to-reverse option whether a smaller version could be tried first. A trial, a pilot, a single machine rather than two, a six-month subcontract instead of a five-year lease — converting a one-way door into a two-way one is often worth more than any improvement in the score.

Always include doing nothing

The sample data opens with "do nothing — keep repairing it", and it is there on purpose. A matrix without the do-nothing option will always recommend action, because every option in it is an action.

Doing nothing has real costs and they should be scored honestly: continuing repairs, continuing downtime, continuing risk. Sometimes it wins. More often it loses clearly, and the decision to act becomes much easier to defend for having been compared against the alternative everybody was actually living with.

The assumption column

Every option rests on something that, if false, turns it from good to bad. Writing that sentence down is frequently more valuable than the entire score.

It gives you three things: a list of what to check before committing, an early-warning list to watch afterwards, and — when the decision is reviewed in two years — an honest record of what was believed at the time rather than a reconstruction of it.

The formulas

FAQ

Two options are within a few per cent. The matrix has not separated them, and the tool says so. Decide on something the scores do not capture — who you would rather depend on, what you would learn, which one leaves more options open — and record why in the notes.

Can I add my own criteria? Not as extra columns, deliberately. Six dimensions cover most decisions, and matrices with fifteen criteria are almost always worse: the weights become impossible to argue about, and the extra criteria correlate with each other so the same consideration gets counted three times. Fold a specific concern into the nearest dimension and put the detail in the notes.

Should the weights add up to 100? They are normalised, so it makes no difference to the arithmetic. Keeping them near a hundred makes them easier to discuss, which does make a difference.

What if an option is disqualified outright? Then it is not an option. Mark it rejected with the reason — a hard constraint is not a low score on a criterion, and treating it as one lets a disqualified option outscore a viable one on everything else.

Everything scores 7 or 8. The scales were not anchored before scoring. Pick the best and worst option on one dimension, agree they are 9 and 2, and rescore the rest against them.

Saving your work

Options, weights and the report header are written to this browser's local storage as you type. Treat Export .json as the real save, which Import .json restores anywhere. Export CSV gives you the cockpit for spreadsheet work. Reset asks twice, then erases everything. There is no undo.

The backup may contain supplier prices and candid opinions about them. Keep it out of anything a supplier could receive.

Accuracy & disclaimer

A weighted score is an argument written down, not a result. The weights are chosen by the same people who prefer one of the options, and a small change to a weight can reverse the ranking without anybody noticing — which is why the sensitivity test matters more than the headline score.

Cost and effort here are estimates, and estimates about options nobody has committed to are the least reliable kind. Nothing in this tool is financial or professional advice, and no score should be presented as a business case on its own.