The Clarity Score — Methodology
One number, taken seriously.
Five pillars. A score from 0 to 100. Every weight, boundary, and default is printed on this page — what is published cannot be a gimmick.
Verified against the production engine · July 2026
01 — What the score is
A mirror, not a grade.
The Clarity Score measures how you handle money — not how much of it you have. It is computed entirely from your own answers: nineteen questions if you hold no investments, twenty-eight if you do, plus a short life-context section that sets the curve you are graded on. No accounts are linked. No documents are uploaded. Nothing is checked against a bureau. The score is only as honest as you are — which is the point of a mirror.
One structural fact matters more than any formula below: the scorer sells nothing at scoring time. No products, no leads, no cross-sell waiting behind the reveal. A measurement with nothing to sell has no reason to flatter you — and no reason to frighten you either.
The arithmetic at the top is deliberately plain. Each of the five pillars — Spend, Save, Debt, Plan, Invest — is scored from 0 to 100. The Clarity Score is their equal-weighted average: 20 percent each. If you hold no investments, the Invest pillar is excluded and the score is the average of the remaining four — an empty portfolio is a fact worth recording, not a reason to drag down everything else.
02 — The five pillars
What is measured, and by how much.
The weights below are those of the production engine’s primary tier — the one that scores the assessment you can take right now. Sub-signals are scored on published curves, then combined at the stated weights. A deeper tier with additional signals exists in the same engine; where it needs an answer the assessment has not yet asked, the engine does not guess.
Spend
Spend (pillar, noun) — discipline relative to income, not to aspiration.
| Signal | Weight | How it is scored |
|---|---|---|
| Savings rate | 45% | Percentage of income actively set aside, on a non-linear curve: saving 20% of income scores 60; above 25% the curve flattens and each further five points of rate earns five points of score. 50% or more scores 100. |
| Bill payment consistency | 25% | Every bill on time, every time scores 100; about half on time scores 20. |
| Non-essential to essential ratio | 30% | Inverted curve. Non-essential spending at half your essential budget scores 75; parity — non-essentials equal to essentials — scores 0 on this signal. |
Save
Save (pillar, noun) — the number of months your life continues if the income stops.
| Signal | Weight | How it is scored |
|---|---|---|
| Emergency runway | 40% | Months of essential expenses you could cover from savings alone, scored against a personal target rather than a universal three months — see the personalisation section below. Meeting your target scores 65; holding one and a half times it scores 85; anything beyond that scores 100. |
| Net position | 35% | Total net assets as a multiple of annual income — a ratio, so it works in any currency. Roughly zero scores 5; one and a half times annual income scores 50; ten times scores 100. In the deep tier this becomes the Stanley wealth-to-age benchmark: target wealth = age × annual gross income ÷ 10, adjusted by age cohort (×0.6 under 35, ×1.3 over 50). |
| Long-term goal confidence | 25% | Self-assessed, from no confidence (0) to a clear plan being executed (100). |
Debt
Debt (pillar, noun) — the cost of earlier decisions, measured against present income.
| Signal | Weight | How it is scored |
|---|---|---|
| Payment-to-income ratio | 60% | Share of monthly income going to debt payments, on an inverted sixteen-point curve: 0% scores 100, 20% scores 75, 35% scores 50, 50% or more scores 0. |
| Revolving credit behaviour | 40% | Always clearing the full balance inside the grace period scores 100; missing minimums scores 0. Holding no credit card at all carries no penalty — the signal is scored as clean. |
The assessment also asks four credit-history questions — past denials, thirty-day misses, whether you know your bureau score, how many revolving accounts you hold. They are context, not weight.†
Plan
Plan (pillar, noun) — five arrangements that either exist or do not.
Plan is an additive checklist, 20 points per item, no partial credit: a budget or expense tracking; automatic saving; life and disability cover on the household’s primary earner; health cover for everyone in the household; and savings or investment accounts with specific goals attached. Five yeses score 100. The pillar rewards paperwork nobody sees, which is precisely why it predicts so much.
Invest
Invest (pillar, noun) — a position, as distinct from an account.
| Signal | Weight | How it is scored |
|---|---|---|
| Participation | 35% | Breadth of instruments held, multiplied by an allocation gate on the share of take-home actively invested. The engine ranks instrument quality — international exposure 25, broad funds 20, single stocks 15, down to crypto at 2 — and the gate is strict: instruments held while investing under 1% of take-home score 0 on this signal. Full credit arrives at 20% or more. |
| Goal alignment | 30% | Whether the investing serves named goals. Not yet asked in the current assessment; scored at a neutral midpoint until it is.‡ |
| Risk alignment | 35% | Whether holdings match the risk profile. In the deep tier, actual equity share is scored against a published target per profile — Conservative 30% through Aggressive 87% — alongside a home-country bias curve on which a fully domestic portfolio scores 0 and 30% or less domestic scores 100. |
If you hold no investments the pillar is not scored at zero — it is excluded, and the Clarity Score becomes the average of the other four pillars.
Asked, not weighted
The assessment collects more than the primary engine weighs. Your saving time horizon, credit history, self-described risk profile, equity share, home-country share, and diversification counts are scored by the engine’s secondary scorer and reserved for the deep tier — in the primary score they currently carry no weight. This is printed here because a methodology page that omitted it would be marketing.
03 — The bands
Six bands. The boundaries are printed because they do not move.
| Band | Range | The reading |
|---|---|---|
| Crisis | 0–20 | Things are moving without you. That can change. |
| Vulnerable | 21–40 | What is missing is a plan — not ability, not income. |
| Developing | 41–60 | Managing, not yet building. The gap is specific. |
| Aligned | 61–75 | The foundation is there. The structure above it is forming. |
| Thriving | 76–90 | Sound fundamentals. What remains is the difference between having investments and having a position. |
| Mastery | 91–100 | Everything in order — rarer than it sounds. |
Beneath the six display bands, the engine resolves seven finer internal segments — 0–20, 21–39, 40–52, 53–65, 66–78, 79–89, 90–100 — used for calibration and for setting the register of the written verdict.§ The five-question quick read on the landing page is a deliberate approximation of this scale and uses coarser boundaries.‖
04 — Personalisation
A 34-year-old renter and a 44-year-old parent are not graded on the same curve.
The clearest example is the emergency runway target. The engine starts at three months of essential expenses and builds a personal target from your life context: one month is added for renting, one for a mortgage, one for a partner who does not work, one per child up to three, one or two for other dependents, and one when retirement is near — over 55 for men, over 50 for women; if you prefer not to say, the engine uses the more conservative threshold. The deep tier adds a month each for a rare profession, no second income, an unstable industry, or an unstable country. The target never falls below one month.
Worked example. Both people report three months of runway. The 34-year-old renter with no dependents has a target of four months and scores 65 on that signal. The 44-year-old homeowner with a mortgage and two children has a target of six months and scores 25. Same savings, different lives, different grades — which is what a personal target is for.
Two calibrations exist in the engine but are deliberately neutral today. The engine carries country profiles — gross-up factors and median incomes for eight markets plus a default — and an effort adjustment that grades a given savings rate 15 percent kinder below half the median income and 10 percent stricter above twice it. The current assessment asks for neither your income nor your country, so it scores everyone at the neutral median default rather than guessing. When those questions ship, this page will say so.
05 — Benchmarks
Where the peer numbers come from.
Two kinds of peer figures appear in the product, and they have different provenance — stated here plainly.
The peer percentile shown with your score comes from a synthetic cohort model: a normal-distribution estimate anchored to the CFPB Financial Well-Being Scale — population mean 54, modelled standard deviation 15 — adjusted by age and income bracket. It is a model calibrated to published survey data, not a live distribution of Wellfy users. As the user base grows, live distributions replace the model.¶
The per-answer percentages — “x% of respondents chose this” — are static distributions shaped on the CFPB Financial Well-Being Survey (a 2016 national survey of 6,394 respondents), structured for replacement by live data. Sensitive questions, such as your net position, show no peer figures by design.
What we will not do is present a modelled figure as a measured one. The model is honest about being a model; the moment it is live data, it will be honest about that instead.
06 — The perception gap
You feel behind. Are you?
The research case for a mirror is straightforward: how people feel about their money and how their money is actually arranged are two different measurements, and they disagree in both directions. The CFPB built its Financial Well-Being Scale precisely because well-being is not income in disguise — people at the same income report very different financial lives, and behaviour explains more of the difference than earnings do.
“Many individuals are over-optimistic about their prospects, including their financial health, with a significant difference between perception and reality.”
The gap runs the other way just as often: people with sound arrangements who feel permanently behind. Financial anxiety is not financial incompetence — it is usually the absence of a picture. A score built from behaviour, not balance, closes the gap from both sides: it tells the over-optimistic what is actually missing, and the anxious what is actually fine.
This is why the score is descriptive rather than prescriptive, and why the product’s voice gets quieter at low scores, not louder. A mirror that shouts is a billboard.
07 — Engineering
The same answers produce the same score. Every time.
The Clarity Score is computed by a 590-test deterministic engine‡‡ — a pure function with no randomness, no model temperature, and no silent revisions of the math. The suite covers the canonical formulas, adversarial and boundary inputs, and full user journeys. The output is always a whole number from 0 to 100: inputs are clamped, malformed values degrade to defined defaults, and there is no path to a NaN, an infinity, or a negative score.
The score is computed in your browser, on this domain, by the same engine files described on this page — there is no server round-trip to produce the number and no black box behind an API. Your answers persist to Wellfy’s database only if you create an account to keep your history; without one, they stay on your device.
In embedded (partner) mode the posture is stricter still: no answers, no scores, and no identifiers are sent to Wellfy’s servers — the default persistence layer is a null store, and results reach the host page exclusively through an event callback that carries the score, band, archetype, and pillar scores, never individual answers and never personal data.
Finally, a score is a photograph, not a portrait. The engine defines a staleness rule — after ninety days, a stored score regresses ten percent toward the population mean — so that an old number cannot quietly pass itself off as a current one.
08 — For institutions
Banks know behavioural scoring works. The question is whose score.
The category evidence is public: Discovery Bank built a five-behaviour money score into a banking product, and in the UNSGSA’s case study its top-band clients are 97 percent less likely to be in arrears and hold eleven times the deposits of its lowest band.§§ Building that took them a bank. Embedding this takes a sprint: the Clarity Score ships as a white-label widget — a sandboxed iframe with a validated eight-key theme contract, so it renders in your palette, light or dark — and in its default mode no user data ever reaches Wellfy’s servers. This page is the diligence packet: the weights, the boundaries, the defaults, and the caveats are the same ones your analysts will find in the code. The commercial case — packaging, integration, the live demo — is at Wellfy for institutions. See the working demo, or write to hello@getwellfy.com.
Notes
- † The credit-history answers are consumed by the engine’s secondary scorer and reserved for the deep tier; they do not move the primary score. ↩
- ‡ A neutral midpoint, not a hidden assumption: every respondent receives the same value on this signal until the question is asked, so it cannot separate one score from another. ↩
- § Seven segments, six bands: the display bands are what you see; the internal segments set tone and calibration. Both sets of boundaries are fixed in code and covered by tests. ↩
- ‖ The quick read scores five answers at up to 20 points each and labels the result with wider bands (its Aligned runs to 79). It exists to show you the shape of the instrument before you sit for the full assessment; it is marked as an approximation in its own code. ↩
- ¶ Consumer Financial Protection Bureau, Financial Well-Being in America (2017) — the scale’s published national mean is 54. The standard deviation and the age and income adjustments are Wellfy’s modelling choices, stated as such. ↩
- †† Discovery Bank, Banking for a Stronger South Africa, the research paper behind Vitality Money. Cited as evidence for the perception gap, not as an endorsement of Wellfy. ↩
- ‡‡ 590 passing tests at the time of writing (July 2026). The count drifts in one direction — up — as signals ship; the number here is refreshed when the page is. ↩
- §§ UNSGSA case study on Discovery Bank’s Vitality Money. This is category evidence — proof that behavioural money scoring predicts outcomes — not a Wellfy result. We label our evidence, and we do not borrow other people’s. ↩