Working Data skill
Decide with data
Build an honest comparison for one real decision.
Help someone decide something real that they are actually facing right now, with money or months at stake: which college, phone, laptop, job offer, apartment, plan. The product is a comparison sheet plus a short decision memo in the person's words. Be straight about the limit up front: the sheet organizes the decision and exposes hidden assumptions; it does not make the decision, and a spreadsheet that claims to is lying.
This skill supports a decision; it does not replace qualified medical, legal, tax, investment, or other regulated advice. For a high-stakes decision, use current official sources, record uncertainty, and route professional questions to an appropriate expert. The person chooses and owns the conclusion.
Frame it
Get the real option list — usually two to five. If one "option" is "do nothing", it goes in the sheet too; it has costs and scores like the rest. Then criteria, theirs and not a template's: ask what would make them regret a choice in a year, and what they are quietly already deciding on. Five to eight criteria; merge overlapping ones, because two criteria measuring the same thing is a hidden double weight. Each criterion needs a stated direction (cost: lower is better) and a unit or scale.
Fill it with real data
Options as rows, criteria as columns. For each cell, find the actual number: real tuition and aid figures, real commute time from the maps app, real battery benchmarks, the actual salary and hours from the offer. Keep a sources column or note per number. Some criteria are genuinely subjective ("I like the campus") — score those 1–5, but mark them as their judgment, not measurement; both belong in the sheet, labeled as what they are. Where a number is unknown, leave it visibly blank and note what finding it would take. A blank cell is often the output: it tells them which phone call to make before deciding.
Weight and score
- Normalize each column to a 0–10 score so units stop mattering; flip the direction for lower-is-better columns. Do the arithmetic in formulas they can inspect.
- They set weights before seeing totals, by asking which criterion would win a head-to-head trade-off. Weights sum to 100.
- Compute the weighted total per option. Then the important step: stress-test it. Shift the biggest weight up and down and watch whether the winner changes. A winner that survives says the decision is robust; a winner that flips on a small nudge says the options are genuinely close and the sheet has hit its limit.
- Watch for score-fitting: if they nudge weights until the option they secretly wanted wins, name it kindly — the sheet just told them their answer, and that is a legitimate result. Write it down as such instead of laundering it through fake weights.
Say what the numbers cannot
Finish with a half-page in their words: the top option, the margin, what the ranking was sensitive to, and the criteria that never fit in a column — gut feel about people, fear of missing out, family factors. Naming what the sheet excludes is what makes the rest of it honest. If the totals are within noise of each other, the finding is "either is defensible; decide on the unquantified part" — that is a successful analysis, and worth saying plainly.
---
name: decide-with-data
category: data
description: Build an honest comparison for one real decision. Criteria the person chose, weights, data per option, and what the numbers cannot settle. Use when they are stuck between options or ask "help me decide".
---
# decide-with-data
Help someone decide something real that they are actually
facing right now, with money or months at stake: which college,
phone, laptop, job offer, apartment, plan. The product is a
comparison sheet plus a short decision memo in the person's words. Be straight
about the limit up front: the sheet organizes the decision and
exposes hidden assumptions; it does not make the decision, and a
spreadsheet that claims to is lying.
This skill supports a decision; it does not replace qualified medical, legal,
tax, investment, or other regulated advice. For a high-stakes decision, use
current official sources, record uncertainty, and route professional questions
to an appropriate expert. The person chooses and owns the conclusion.
## Frame it
Get the real option list — usually two to five. If one "option" is
"do nothing", it goes in the sheet too; it has costs and scores like
the rest. Then criteria, theirs and not a template's: ask what would
make them regret a choice in a year, and what they are quietly
already deciding on. Five to eight criteria; merge overlapping ones,
because two criteria measuring the same thing is a hidden double
weight. Each criterion needs a stated direction (cost: lower is
better) and a unit or scale.
## Fill it with real data
Options as rows, criteria as columns. For each cell, find the actual
number: real tuition and aid figures, real commute time from the
maps app, real battery benchmarks, the actual salary and hours from
the offer. Keep a `sources` column or note per number. Some criteria
are genuinely subjective ("I like the campus") — score those 1–5,
but mark them as their judgment, not measurement; both belong in
the sheet, labeled as what they are. Where a number is unknown,
leave it visibly blank and note what finding it would take. A blank
cell is often the output: it tells them which phone call to make
before deciding.
## Weight and score
- Normalize each column to a 0–10 score so units stop mattering;
flip the direction for lower-is-better columns. Do the arithmetic
in formulas they can inspect.
- They set weights before seeing totals, by asking which criterion
would win a head-to-head trade-off. Weights sum to 100.
- Compute the weighted total per option. Then the important step:
stress-test it. Shift the biggest weight up and down and watch
whether the winner changes. A winner that survives says the
decision is robust; a winner that flips on a small nudge says the
options are genuinely close and the sheet has hit its limit.
- Watch for score-fitting: if they nudge weights until the option
they secretly wanted wins, name it kindly — the sheet just told
them their answer, and that is a legitimate result. Write it down
as such instead of laundering it through fake weights.
## Say what the numbers cannot
Finish with a half-page in their words: the top option, the margin,
what the ranking was sensitive to, and the criteria that never fit
in a column — gut feel about people, fear of missing out, family
factors. Naming what the sheet excludes is what makes the rest of it
honest. If the totals are within noise of each other, the finding is
"either is defensible; decide on the unquantified part" — that is a
successful analysis, and worth saying plainly.
## Done
- A comparison sheet: real options, their criteria, sourced numbers,
subjective scores labeled as such, inspectable formulas
- Weights they set before seeing results, plus a sensitivity check
- A half-page decision memo stating the person's choice or remaining options,
the margin, sensitivity, and what the numbers could not settle
- They can defend any cell out loud, source or judgment
Then: chart-the-truth if the comparison deserves one chart to show
the person they share the decision with.