What you said.
A policy, campaign or target gives us a claim to check. It does not show that anyone benefited.
THE HAPPYWORTH GUIDE
How it works.
What we believe, how we score, and how you can take part.
What is a brand worth to people?
Money shows what people spend. HAPPYWorth examines how a company relates to people’s wellbeing, relationships, health, security and ability to make choices.
Our scores are judgments about selected evidence. They do not directly measure happiness or cover everything a company does. We show the scope and uncertainty so you can decide what the score means for your question.
Evidence earns the score.
Payment, profile ownership and popularity never buy a higher rating.
The essentials stay open.
Scorecards, sources and these rules are free. Reports contain the full written interpretation in dated editions you can keep.
People help shape the questions.
Community input informs future proposals. It does not directly change a company’s score.
Five dimensions. One clear method.
Wellbeing, connection, vitality, security and agency each carry 20% of the overall score. We assess documented commitments, practices, outcomes and harms using the same published evidence rules.
Score describes what the reviewed evidence suggests. Confidence describes how strongly it supports that judgment. A high score can still have limited confidence. Missing evidence means no score, never zero.
Current standard: v1.1, published September 15, 2026. Existing v1.0 editions retain their original scores. The September assessments used AI assistance and have no recorded independent human research review.
Read the formatted scoring standard →
Read the full scoring standard
HAPPYWorth scoring standard v1.1
Version 1.1 · September 15, 2026
Fifteen indicators. More specific evidence.
Each dimension has three equally weighted indicators. Score each indicator using the evidence levels below. All fifteen indicators need sufficient evidence before we publish an overall score.
| Dimension | Indicators, in assessment order |
|---|---|
| Wellbeing |
|
| Connection |
|
| Vitality |
|
| Security |
|
| Agency |
|
Average the three indicator scores to one decimal place, then average the five dimension scores and round the total to a whole number. An adverse indicator caps the dimension at its lowest adverse score; positive activity cannot cancel it. A missing indicator leaves that dimension and total unscored. Outcome requirements apply to each individual indicator.
More granular values are not proof of greater accuracy. This method remains an editorial assessment. Compare scores produced under the same version and scope; earlier v1.0 scores are retained without recalculation.
Five dimensions. Equal weight.
Each dimension carries 20% of the score. This is an editorial choice. Purchases, profile claims and company size do not affect scores or weights.
| Dimension | What we assess | Weight |
|---|---|---|
| Wellbeing | Experienced wellbeing, work–life support and emotional health. | 20% |
| Connection | Inclusion, belonging and participation in communities. | 20% |
| Vitality | Physical health, occupational safety and product-related health. | 20% |
| Security | Economic stability, personal safety and protection of personal information. | 20% |
| Agency | Informed choice, accessibility, skills, voice and access to remedy. | 20% |
Evidence determines each score.
Each indicator receives one of six evidence scores, with a finding and source explaining why. A promise, a programme in place and a measured outcome receive different scores.
| Score | Basis | Rule |
|---|---|---|
| 0/100 | Extreme adverse evidence | Credible evidence of extreme, continuing harm in the assessed scope, without an effective response. |
| 20/100 | Material adverse evidence | A documented material adverse outcome dominates the evidence in the assessed scope. |
| 40/100 | Commitment or mixed evidence | A specific relevant commitment is documented, or material contrary evidence limits credit for a positive practice. The company-specific explanation identifies which applies. |
| 60/100 | Implemented support | A relevant practice or support programme is operating. Reach or implementation is documented; a positive human outcome is not established. |
| 80/100 | Measured positive outcome | A relevant favourable outcome is quantified for a stated population and period. Self-reporting, selection and causal limits still apply. |
| 100/100 | Strongly corroborated outcome | Sustained favourable outcomes are independently corroborated with adequate population coverage and no material unresolved contrary evidence in the assessed scope. |
Every dimension needs a supported score, a finding and linked evidence. Review dates, source provenance and the extra requirements for higher scores must pass the published checks.
Missing evidence means no score, never zero. Incomplete or invalid reviews cannot produce a published score. The last valid edition may remain available as history.
What earns 80 or 100?
A score of 80 needs a named measure, numeric result and unit, population, observation period and source. The outcome must fit the dimension below.
| Dimension | Qualifying outcome |
|---|---|
| Wellbeing | A measured experience. Engagement or satisfaction must be labelled as an indirect measure. |
| Connection | A measured experience of belonging, inclusion or connection. |
| Vitality | A health or safety outcome, such as a documented injury-rate improvement. |
| Security | A benefit people actually received: pay, financial aid or a defined protection. |
| Agency | Completed education, recorded career progression or an exercised right to have a say, such as coverage under a ratified collective agreement. |
Spending, enrolment, participation and available benefits show implementation alone. Graduation can support agency, but does not prove better wellbeing or earnings. A feedback channel does not prove that concerns are resolved.
A score of 100 needs all of the following:
- A qualifying measured outcome.
- Supporting evidence from at least two independently controlled publishers.
- At least two observation periods ending a year or more apart.
- Evidence covering at least 70% of a defined eligible population.
- A documented selection method.
- No material unresolved contrary evidence within the assessed scope.
These are editorial requirements, not scientific certification.
When the evidence shows harm.
We weigh benefits and harms within the same population, activity and period. Mixed evidence scores 40. Material harm that dominates the finding scores 20. Extreme, continuing harm without an effective response scores 0. Positive programmes cannot cancel a dominating harm in that dimension.
Each decision explains what happened, the evidence and the response. We distinguish allegations, findings and settlements, and record the dates of ended outbreaks, terminated orders and verified remedies. Past harm is not automatically continuing harm; a promised remedy is not proof of resolution. A targeted search can miss harm. Finding none does not prove none occurred.
A simple calculation.
Overall score = (wellbeing + connection + vitality + security + agency) ÷ 5
Scores of 60/100, 60/100, 80/100, 40/100 and 60/100 average to 60/100. All five dimensions need sufficient evidence; missing values never become a neutral score or zero. In v1.1, calculate dimension scores from their three indicators first. Small gaps may not be meaningful.
A balanced mix of company and external evidence.
Credible company disclosures may support individual indicators. Every new edition also needs substantive external checks that address its actual findings, population and evidence period. The review records source passages, publishers, contrary evidence and remaining gaps. A policy index verifies a disclosed commitment; it does not prove the policy works. A survey provider, union or programme partner may have a commercial or advocacy relationship, which must be disclosed.
A report can be published with explicitly unrated indicators. An overall score requires all fifteen indicators. Missing evidence is never zero and is never averaged away. These reports remain outside numeric rankings. Earlier editions keep their original scores, scope and sources.
Published does not mean certain.
Published means an assessment meets the evidence and review requirements. Confidence describes how strongly the evidence supports it. A published score can still have limited confidence.
| Confidence | Requirements |
|---|---|
| Limited evidence | Publication requirements are met, but stronger confidence requirements are missing. Company disclosures may support findings alongside substantive external checks. A targeted search, older observations or the absence of an independent human research review can still limit confidence. |
| Moderate evidence | An independent human review, a systematic search for contrary evidence, recent dated evidence, and independent evidence in at least three dimensions. |
| High evidence | All moderate-confidence requirements, plus independent evidence in all five dimensions. Each dimension also needs two independent publishers, periods ending at least a year apart, at least 70% coverage of a defined eligible population, a documented selection method and no material unresolved contrary evidence. |
A high score can have limited confidence; strong evidence can support a low score. The September 2026 assessments used AI assistance and have no independent human research review. Their evidence confidence remains limited.
Scores, colors and rankings.
A score is a number out of 100. Rank orders companies by score, highest first. Ties share a rank and skip the corresponding places after it; tied names appear alphabetically. Category filters recalculate ranks within that category. A name search preserves them.
Score colors
- 0–39 Concerning
- 40–64 Mixed
- 65–79 Positive
- 80–100 Strong positive
Red covers 0–39, amber 40–64, blue 65–79 and green 80–100. Colors help you read scores; they are separate from the scoring rules. Crossing a color boundary does not prove a meaningful difference.
Confidence is shown separately. Blue and green scores can have limited confidence. Gray means no current published score.
Sources and their dates.
Findings link to the public company reports, benefit and supplier policies, regulator records, research or external evaluations used. Sources are labelled company-controlled, independent or independence not established. Different websites with the same publisher do not provide independent corroboration.
The date we check a source is separate from the period it describes. Checks must fall within 30 days of the full review. Evidence older than two years needs a specific reason for inclusion. Unknown observation dates stay unknown. Reopening an old report does not make it current.
We examine funding, survey selection, eligibility, reporting boundaries and corporate changes. New reviews seek representative worker and customer evidence, comparable health and injury measures, adequate-pay evidence, progression, independent evaluations and regulator follow-up. A company disclosure is not independent verification.
Scope, corrections and change history.
Workforce, supplier and education-programme assessments cover different people. A brand does not inherit its parent’s score. Employees do not represent every contractor, customer or franchise worker. Averages can hide unequal experiences; disclosure quality and evidence selection also affect the result.
Use the correction link on any assessment. Include the company, dimension, source, period and people affected. Each new edition keeps its evidence and review dates, names the methodology, explains changes and links to the previous edition. Corrections may change the score, confidence or scope.
Changes to weights or scoring rules require a new published methodology. Community proposals follow the same process; they do not silently change existing scores.
Methodological references.
Our approach to documenting combined scores and uncertainty draws on the OECD handbook on constructing composite indicators. Our approach to balanced, traceable reporting also draws on GRI 1: Foundation 2021. Our dimensions, weights, score levels and thresholds are editorial choices. These references do not certify or validate our model.
Scoring standard v1.1 · Published September 15, 2026.
Explore brands →A score has a date, not a lifetime.
7 daysScheduled evidence checks
90 daysA full review is due
180 daysExpires without a new full review
Review deadlines run from the last full review. Evidence checks follow their own schedule and do not renew that date. Material new evidence or a correction can trigger an earlier reassessment. Unsupported scores may be withdrawn; expired and withdrawn editions stay in the history and leave current rankings.
Research status, last review, next review and evidence-check dates are together in the brand directory.
See research status & review dates →Review status explained
Evidence checks are scheduled every 7 days. A full review is due every 90 days. Regulator findings, safety incidents, corrections, significant outcomes or changes in company scope can trigger an earlier review.
| Status | What happens |
|---|---|
| Current | The assessment meets the requirements and is within its review period. |
| Review due | At 90 days, the score is marked due for review. A full review is needed to renew it. |
| Expired | At 180 days without a full review, the score leaves current rankings. The edition stays in its history. |
An evidence check records a change or no material change. It does not update the score or renew its review date. A full review reconsiders the evidence, records its findings and meets the publication requirements, even when the score stays the same.
How we choose companies for research
We initially allocate half of research time to everyday impact, a quarter to overlooked sectors and affected groups, and a quarter to search interest. Editors review that balance and unusual search spikes before starting work.
A company in our directory enters the search priority queue after at least 20 accepted signals across three days within a rolling 28-day window. We cap each company at 20 signals a day and give older signals less weight. These rules help organize research; they do not measure popularity or unique people.
We match searches to verified directory names. Unrecognized searches count only toward a general coverage gap; their text is discarded and never becomes a public company record. Read about search privacy →
Before publication, we check the company, market, period, affected groups, source independence, missing evidence and later corrections. A draft or source link alone is not a published assessment. Our scores follow published editorial rules; the model is not a validated measure of happiness.
Happiness isn’t ours to define alone.
Tell us which dimensions matter to you. Each verified account has one active contribution per quarter, with equal influence regardless of purchases. You can edit or withdraw it.
Reviewed community priorities can contribute up to 10% to proposed future weights. Changes need human review and a new published method. Aggregate results appear after a quarter ends only when at least 30 contributions pass review; individual answers stay private.
Loading participation options…
Question a finding. Bring better evidence.
Anyone can challenge a factual detail, source, scope or score interpretation. You do not need to buy a report or represent the company. Include the disputed point, supporting sources, the period and the people affected.
Submit evidence or a correction
Submit evidence or a correction →Ask us to reconsider a decision.
For a disputed correction or a community moderation appeal, contact hello@happyworth.digital with your original submission, the decision you question and any new evidence. Paying does not give an appeal more weight.
For billing or account help, use your account or contact us. Privacy choices and requests are on the Privacy page.
Turn the evidence into useful questions.
Explore 20 areas for improvement, what to measure and how to test a change. This general guide is free. Company-specific recommendations appear only when included in a purchased report.
Explore the improvement tools
Use this library to frame questions. These are research-informed suggestions mapped to our five dimensions, not findings about your company or source-endorsed point values.
20 areas to explore · choose what fits your company and report
Publish workable schedules earlier, protect time off and make flexibility usable without career penalties.
What could improve
Less conflict between work and life outside it; more opportunity to recover.
What to measure
Actual schedule notice, late changes, leave denials and repeated confidential reports of work–life conflict.
How to check the evidence
Compare scheduling records with independently collected worker experience; include shifts, part-time staff and caring commitments.
What could weaken the evidence
A policy, a perk or better review stars alone does not establish improvement. Scheduling studies do not support promising the same benefit for every outcome or workplace.
Primary dimension: Wellbeing. Other relevant dimensions: Vitality, Agency. Each dimension needs its own outcome evidence. One fact does not count as several independent findings.
HOW THE CALCULATION WORKSFive dimensions.
Five dimensions.
One overall score.
Multiply each dimension score by its weight, add the results and round to the nearest whole number.
This calculator shows the arithmetic. Published scores follow our evidence rules; they are editorial assessments, not a validated measure of happiness. Reviews, sentiment and policies do not add points automatically.
Wellbeing
Less distress. More positive experience.
60/100 · 20.0% weight12.00 points toward the overall scoreConnection
Belonging, relationships and respect.
60/100 · 20.0% weight12.00 points toward the overall scoreVitality
Health, energy and time to recover.
60/100 · 20.0% weight12.00 points toward the overall scoreSecurity
Safety, stability and predictability.
60/100 · 20.0% weight12.00 points toward the overall scoreAgency
Informed choices and meaningful control.
60/100 · 20.0% weight12.00 points toward the overall scoreTRY THE MATH
Illustrative equal-weight example. Try a hypothetical change. This does not predict or edit a published score.
60/10062/100
+2.00 before rounding, with Wellbeing capped at 100. Other dimension scores stay fixed.
A 10-point increase in a dimension with 20% weight adds 2 points overall. This does not mean an action will earn a 10-point increase.
HOW WE REVIEW EVIDENCEShow what changed.
Show what changed.
And for whom.
Use these principles to review an improvement claim. They do not calculate a score or set a scientifically validated threshold.
What you changed.
Dates and records show what changed, when and who it reached.
What people experienced.
Relevant measures and people’s accounts show whether their lives improved, worsened or stayed the same.
What holds up.
Independent sources and repeated observations strengthen a claim. Conflicting evidence needs an explanation.
Glassdoor and online reviews
Look for repeated themes, dates, roles, locations and conflicting accounts. Glassdoor cannot fully verify identity, employment or truthfulness. Reviews offer useful clues, but can be misleading and do not represent everyone.
A sudden rise in reviews calls for a closer look at where they came from; it does not prove fraud. People who volunteer reviews may differ from those asked for feedback. The cited study includes authors affiliated with Glassdoor.
Never require proof of a posted review, offer employer incentives for Glassdoor reviews, identify anonymous reviewers or pressure people to give positive answers. Use sources you may access and can trace. HAPPYWorth does not currently scrape these services.
Check the original source
A company report and ten articles repeating it count as one original source. Check relevant records against independently collected experiences, original research or inspections.
Record who funded and collected the evidence, who took part, the questions, dates, sample size and incentives. Explain who responded and who may be missing. Voluntary feedback cannot tell us what everyone thinks.
An official finding may establish a specific event without a second source. It cannot establish an overall happiness score. Missing evidence means uncertainty, not a zero or proof that all is well.
Before supporting an improvement claim
- Be clear about the scope. Name the company, people affected, location and period. Distinguish workers, customers and communities. A parent-company result does not apply to every brand.
- Plan the test first. Define the starting point, desired outcome, comparison and follow-up period. Account for people who drop out, other changes and unwanted effects. A before-and-after comparison alone does not prove cause.
- Count evidence once. One action may affect several dimensions, but each needs its own outcome evidence. Repeated copies of a claim are not independent support.
- Show evidence of harm. Check how experiences differ and who may be missing. Strong averages and new campaigns cannot erase unresolved harm. Address serious harm before trying to improve an overall score.
- Protect participation. Make feedback voluntary, confidential and accessible. Share grouped results only when they can protect people’s identities, including in small groups. Never require employee health answers or reviewer identities for an assessment.
- Explain the decision. Publish the findings, uncertainty and limitations in a new report version. Apply the same review and correction process to paying and nonpaying companies.
Original records and confidential feedback can be useful without expensive certifications. We consider relevant evidence regardless of a company’s marketing budget.
“We support work-life balance.”
What would make that credible?
A specific problem
Suppose hourly staff report frequent last-minute shift changes. First check the pattern and who it affects. A low score alone cannot identify this problem.
A practical change
Test earlier schedules, reliable cover and shift swaps controlled by employees. Decide who takes part and when to follow up before looking at results.
Measure each outcome
Measure work–life conflict for wellbeing, recovery for vitality, predictable earnings for security and control over schedules for agency. Test each separately; none is guaranteed.
Evidence others can check
Bring records from before and after the change, confidential worker feedback and independent support. Explain who was covered, the limits, other changes and who did not benefit.
The randomized Stable Scheduling study found improved self-rated sleep quality, with mixed results for several other outcomes. It supports testing scheduling changes, but does not show they improve everyone’s happiness.
Where PR and customer trust fit
Explaining a verified change can help customers understand your practices. To claim it benefits customers, measure that separately: better-informed choices, less uncertainty or another relevant outcome.
A higher favorability score shows a change in reputation. Workers’ wellbeing matters in its own right, even if nobody changes a buying decision.
The proposed links are better conditions → worker outcomes, and truthful communication → better-informed customer choices. Each link needs evidence. Good workplace culture does not automatically establish happier customers.
Progress earns a fresh look.
Bring new evidence through the company correction and review process. It can support a higher, lower or unchanged assessment. Paying, updating a profile, announcing a policy or completing an action never directly changes a rating.
Published scores follow our evidence rules and equal 20% weights. These are editorial standards, not a scientifically validated measure of happiness. This guide’s examples are fictional and explain the calculation.
Find your company →