Research Methodology

Every figure the research section publishes is an aggregate of completed Personalvalu.es assessments. There is no survey panel, no third-party data, no modelling and no weighting. Nothing is estimated and no gap is filled in: each number is a stored count or straightforward arithmetic on stored counts.

What counts as a completed assessment

An assessment is completed when a participant has put ten values in order from first to tenth and that ranking has been saved. The ordered ten is the whole of what the assessment contributes. An abandoned attempt produces no record at all: partial rankings are never excluded after the fact because they never enter. Across all 70 months, every completed assessment contributes exactly ten ranked positions, one first-place position and three top-three positions, with no exceptions.

What is stored

The aggregates keep one row per value per period: the period's assessment count, how many assessments ranked the value, the sum of the positions it was given, and how often it was ranked first and in the top three. They contain no individual rankings and no participant identifiers, and no individual result can be recovered from them.

Period covered and gaps in the record

The record runs from July 2020 to September 2026. It begins with July 2020 because that is the first full month; assessments completed before it are not included. The research hub reads the whole record: its rankings treat every assessment from 2020 to 2026 as one period, and its year-by-year figures run to the latest month on file, 2026 included. The yearly statistics pages publish complete calendar years only.

April 2022 to August 2022, 5 months in all, are absent from the record. Those months are missing from the underlying results as well as from the aggregates, so this is a gap in the record rather than a period in which nobody took part. Why they are absent is not something this dataset establishes, and no figure interpolates them.

2020 begins in July 2020 and 2026 is still under way, so neither is a full year. The research hub includes both, and a rate for either is a share of the months present rather than of a full year.

Everyone who takes the assessment sees the same 62 values, so a value's share of assessments is not affected by which options happened to be offered. All 70 months on record carry the same 62 values.

Each value is shown with a short first-person description to help participants interpret the term consistently. The full set is published as the core values list.

The framework contains additional classifications of the values, but no published research metric aggregates values by those classifications. The aggregates key on the individual value, so every published breakdown is by value rather than by category.

Step one, selecting

A participant reviews the full set of 62 values and marks the ones that matter to them. At least ten must be selected before the ranking can begin, and there is no upper limit.

Step two, ranking by pairwise comparison

The selected values are then ordered through repeated head-to-head choices: the participant is shown two values at a time and picks the one that matters more, until ten values stand in order from first to tenth. No value is ever rated on a scale or given a score. Every judgement recorded is a preference between two specific values.

What the final ranking represents

• An order, not a score. Position one is the value the participant preferred over every other value it was compared against, and position ten the lowest-ranked of the ten selected values.
• Exactly ten values, always. Where more than ten were selected, only ten reach the final order and the rest leave no trace in the aggregates.
• One strict order per assessment, so within a single assessment no two values share a position and ties are not possible.
• A ranking of what a participant selected, not of all the values they hold. A value they care about but did not select cannot appear.

Participants can enter the assessment through different product flows, differing only in how the selection is made. Each uses the same pairwise ranking of ten values and contributes the same kind of research record, so the aggregates do not distinguish between them.

Participants may also arrive through practitioner invitations, which can affect who takes part but do not change the values shown, the ranking method or how the result is counted.

Appearance rate

The share of a period's completed assessments whose ranked ten includes the value.

assessments whose ranked ten includes the value

all completed assessments in the period

× 100

Health was in the ranked ten of 977,603 of the 2,099,004 assessments since 2020, which is 46.57%.

• Because every assessment ranks exactly ten values, the appearance rates of all 62 values add up to 1,000% rather than 100%. A rate of 16% therefore means the value was ranked by roughly one assessment in six, not that it took a sixth of some fixed budget.
• A participant may consider a value important without it appearing here, because only the ten they ranked are recorded.
• Published on: Every research and statistics page.

First place, or Top 1

The share of a period's completed assessments that placed the value in position one.

assessments that placed the value first

all completed assessments in the period

× 100

Health was ranked first by 226,118 of the 2,099,004 assessments since 2020, or 10.77%.

• Every assessment contributes exactly one first place, so these 62 figures add up to 100%.
• Ranking first is a separate measure from appearing often, and the two order the values differently. A value can be ranked by comparatively few participants and still lead on first place among them.
• Published on: The research hub and the yearly and monthly statistics pages.

Top three

The share of a period's completed assessments that placed the value in positions one, two or three.

assessments that placed the value in positions 1 to 3

all completed assessments in the period

× 100

• Every assessment contributes exactly three top-three placements, so these figures add up to 300%.
• Published on: The yearly and monthly statistics pages.

Podium share

Of the participants who ranked a value at all, the share who placed it in positions one, two or three.

assessments that placed the value in positions 1 to 3

assessments that ranked the value at all

× 100

Since 2020, family has been ranked by 871,890 participants, and 518,940 of them put it in their top three, which is 59.52%.

• The denominator is the value's own rankers, not the period's assessments, so this measure says nothing about how many people ranked it.
• It is the top-three counterpart of average position, and like it, a value ranked by very few participants can lead on it.
• Published on: The research hub.

Average position

The mean position a value was given, averaged over only those assessments that ranked it.

sum of the positions the value was given

assessments that ranked the value at all

The positions health has been given since 2020 add up to 3,744,395. Divided by the 977,603 assessments that ranked it, its average position is 3.83.

• Lower is higher priority. Positions run from 1 to 10, and 1 is the value a participant put first.
• The denominator is not the period's assessments but the assessments that ranked this value, so the measure says nothing about how many people ranked it. A value ranked by very few participants, but placed near the top by all of them, has a strong average position.
• Ties are not possible: each assessment produces one strict order.
• Published on: The research hub and the yearly and monthly statistics pages.

Relative change

The change in a value's appearance rate between two periods, expressed as a share of the earlier rate.

later rate − earlier rate

earlier rate

× 100

A value ranked by 6.0% of assessments in one year and 7.5% in the next has a relative change of +25%.

• This is the reading the statistics tables report as a trend.
• Relative change can make a small absolute movement in a rarely ranked value appear large: a rise from 0.4% to 0.5% is +25%, the same figure as a rise from 6.0% to 7.5%.
• Published on: The research hub and the yearly and monthly statistics pages.

Percentage-point change

The plain difference between two appearance rates, in percentage points.

later rate − earlier rate

The same move from 6.0% to 7.5% is a change of 1.5 percentage points.

• Relative change and percentage-point change rank the values differently, so every figure states which of the two it is, and the two are never mixed inside one ranking.
• Published on: The research hub.

Country appearance rate

A value's appearance rate within one country, calculated over that country's own assessments.

assessments from the country whose ranked ten includes the value

all assessments from that country in the covered months

× 100

• The denominator is the country's own assessments, so the figure is a share of that country's participants rather than of the whole record.
• Published on: The country section of the research hub.

Difference from the shared rate

How far a country's appearance rate for a value sits above or below the rate across every assessment that carries a country, in percentage points.

country's appearance rate − rate across all assessments with a known country

• The comparison is against assessments that carry a country, not against the whole record, most of which carries no country at all.
• It is a difference between two groups of participants and not a ranking of countries.
• Published on: The country section of the research hub.

How periods are assigned

An assessment belongs to the calendar month in which it was completed, measured in UTC. Days, months and years all use UTC boundaries, with no adjustment for a participant's local time, so an assessment finished late in the evening in one time zone can fall into the following UTC day.

A month is aggregated only once it has fully passed, so the month in progress is never published, and a year is the sum of its months.

Comparing two periods

Every rate is a share of its own period's assessments. That is what makes periods of very different size comparable, and it also means the denominator changes from one period to the next: a comparison between two years compares two different groups of participants, of different sizes, who happened to take the assessment in those years.

Year-over-year change is reported in two ways that are never mixed inside one ranking, as relative change and as percentage-point change. Each figure states which of the two it is, because the two order the values differently.

Partial years are marked as such wherever they appear. Comparing one with a full year compares unequal spans, and any seasonal pattern in the missing months is simply absent from the shorter side.

Where country comes from

Country is derived from the network address of the device that opens the result page, through a lookup that returns a two-letter country code. It is not asked before or during the assessment, and it is not taken from a payment, an account or a profile.

Participants can correct it: the form that carries it shows the detected country in a selector they can change, alongside the optional age and gender questions. The stored data does not distinguish a detected country from a corrected one, so the share that was corrected by hand is unknown.

• Country indicates the apparent network location of a participant, not necessarily their place of residence. VPNs, proxies and travel can affect this location, and the dataset does not identify those cases.
• Granularity is the country, and nothing finer. No region, city or coordinate is stored.
• Country is known only where the result page was opened in a browser and the lookup succeeded. An assessment whose result was never opened carries no country.
• Anything that is not a valid two-letter code is discarded rather than grouped. There is no unknown or other bucket, so assessments without a country are absent from every country figure rather than pooled into one.

Coverage and the publishing floors

Country was not recorded for most of the record's life, and the country figures cover only the recent months where it is known for most assessments. Three floors decide what is published, and a period or a country that fails one simply has no figures rather than thin ones.

• A country needs at least 100 assessments within a month for that month to be recorded for it at all. A country that stays below this every month accumulates nothing, so it can be absent even over a long span.
• A month enters the country figures only if a country is known for at least 50% of its assessments. Below that threshold, the country breakdown would cover too small a share of the month's assessments to be published.
• A country is published only with at least 300 assessments behind it across the covered months. Smaller countries are dropped, because a percentage over a few dozen people moves several points on one more participant.

On that basis the country figures currently cover 10 months, December 2025 to September 2026. A country is known for 83.6% of the 326,540 assessments completed in those months, which is 272,982 assessments, and 25 countries clear the publishing floor.

The distribution is heavily concentrated: the largest single country accounts for 57% of all assessments with a known country, so the shared rate that each country is compared against is weighted towards it. This is the most important caveat on any comparison between countries here.

How country figures may be read

A country figure describes Personalvalu.es assessments associated with that country. It is not a sample of that country's population, and differences between countries include whatever differs about who arrives here from each one, which the data cannot separate from any difference in values. The supportable phrasing is “among Personalvalu.es participants in that country”.

Where age comes from

Age is a question on the result page, shown once the ranking is done, alongside the country selector and an optional gender question. A participant picks one of seven bands (under 18, 18–24, 25–34, 35–44, 45–54, 55–64, 65 and over) or leaves it blank. No date of birth is asked for or stored, and an answer can be given or changed after the result was first opened.

• Age is self-reported and optional. Participants who answer are a subset of those who finish, and plausibly a more engaged one, so every age figure describes the participants who answered rather than everyone in that band.
• A band is compared against everyone who gave an age, not against the whole record. The whole record is mostly participants who did not answer, and would be a different group.
• Anything that is not one of the seven bands is discarded rather than grouped.

Coverage and the publishing floors

• A month enters the age figures only if an age is given for at least 10% of its assessments. The floor is far lower than the one for country because age is optional: at the country floor no month would qualify. What it keeps out is the question's first weeks, when few results had been shown it.
• Every band is recorded for every covered month, however small. There are only seven, and a monthly floor would drop the oldest band from most months, so its total would cover different months from the others.
• A band is published only with at least 500 assessments behind it across the covered months.

On that basis the age figures currently cover 10 months, December 2025 to September 2026. An age is given for 20.4% of the 326,540 assessments completed in those months, which is 66,624 assessments, and 7 bands clear the publishing floor. The oldest band is the smallest by a wide margin, so its figures move the most from one update to the next.

How the trends are measured

A value is said to rise or fall with age by the slope of a straight line fitted through its rate in each published band, youngest to oldest, with each band weighted by its number of assessments. The slope is given as the average change in percentage points from one band to the next. Weighting keeps the smallest bands from deciding the order on their own, and fitting every band rather than comparing the two ends keeps a value that peaks in the middle from reading as one that climbs throughout. Only values with a rate in every published band are compared.

How age figures may be read

A difference between bands is a difference between the participants in them at the time they took part. It cannot separate what changes as people get older from what differs between generations, and it also includes whatever differs about who arrives here at each age. The supportable phrasing is “among Personalvalu.es participants aged 25–34”, not “people in their late twenties”.

Where gender comes from

Gender is a question on the result page, asked beside age and country once the ranking is done. A participant picks male, female, non-binary or “prefer not to say”, or leaves it blank, and can give or change the answer after the result was first opened.

• Gender is self-reported and optional, so every figure describes the participants who answered rather than everyone of that gender.
• “Prefer not to say” is an answer to the question but not a gender. It is counted in no group and not in the baseline either.
• Women are the larger share of those who answer. Each group is compared against everyone who gave a gender, and women and men are also compared with each other directly, since a baseline made up mostly of women would pin every difference on men.

Coverage and the publishing floors

• A month enters the gender figures only if a gender is given for at least 10% of its assessments, the same floor as for age and for the same reason.
• Every group is recorded for every covered month, however small, so each group's total covers the same months.
• A group is published only with at least 500 assessments behind it across the covered months. Without both women and men above that floor, no gender figures are published.

On that basis the gender figures currently cover 10 months, December 2025 to September 2026. A gender is given for 19.8% of the 326,540 assessments completed in those months, which is 64,656 assessments, and 3 groups clear the publishing floor.

How the gaps are measured

The gap for a value is its appearance rate among women minus its rate among men, in percentage points. Only values with a rate in both groups are compared. The shared-core figure counts how many of the 10 values women rank most often are also among the 10 men rank most often.

How gender figures may be read

A gap is a difference between the women and men who took part and chose to answer. It says nothing about why they differ, and it also includes whatever differs about who arrives here. The supportable phrasing is “among Personalvalu.es participants who gave a gender”, not “women value” or “men value”.

The dataset represents people who completed the Personalvalu.es assessment. They arrive through search, through links other people shared and through invitations, and they choose to take part. There is no sampling frame, no quota and no recruitment, so this is a large group rather than a representative sample, and no weighting is applied to make it stand in for one.

• Participants are self-selected. People who look for a values assessment, and finish one, may differ from the general population in ways this data cannot measure.
• The unit is an assessment, not a person. Someone who takes the assessment more than once contributes more than once, and the aggregates carry no way to tell those apart, so the assessment count is not a count of people.
• The assessment is presented in English only, which shapes who can take part.
• Country coverage is recent and concentrated, so the geographic mix of the record as a whole is not established.

None of this makes the figures less real. They accurately describe what this group of participants ranked, which is a large and growing body of evidence about how people prioritise values when made to choose between them. It is the step from that group to any wider population that the data does not support.

What the data can show

• How frequently each value is selected and ranked, over the whole record and within any published period.
• Ranking patterns, such as which values are ranked first out of all proportion to how often they are ranked at all.
• Change over time, as a shift in the share of assessments ranking a value.
• Differences between groups of participants where the sample sizes permit it, which at present means between countries in the recent months.

What the data does not establish

• Causation. A change between two periods is a change in what this group of participants ranked, not evidence of a cause, and no finding in the research section proposes one.
• Why a participant chose a value. Only the choices are recorded, never a reason, and the pairwise format gives no account of motivation.
• Psychological mechanisms or traits. The aggregates are counts of rankings, and no instrument here measures anything underlying them.
• Prevalence in a national population. Country figures describe participants from a country, not that country.
• Representative polling results. With no sampling frame and no weighting, these figures are not comparable with survey estimates that have both.

Dimensions that are not published

Age and gender are asked on the same optional form as country, but no figure is broken down by either. The published aggregates record both dimensions as “all”, so any comparison between age groups or between genders would be invented rather than measured.

Which values are ranked together is also not published. That would need counts of value pairs, which are not aggregated anywhere, so no statement about values co-occurring can be supported from this data.

The research figures are updated periodically rather than continuously. The aggregates are rebuilt from the underlying assessment data at no fixed interval, so a completed month appears in the figures some time after it ends rather than the moment it does. Country breakdowns are rebuilt separately from the totals, which is one reason they cover fewer months.

The research pages show the most recent published aggregates, so figures may change when the dataset is rebuilt. A figure taken from this section is a figure as of the period it covers, which is why every headline number states its period.

The aim is that any figure in the research section can be traced back to how it was produced. This page states the size of the dataset and the period it covers, the definition and formula behind every metric, the floors and exclusions that decide what appears at all, the gaps in the record, and the limits of what the data supports.

The aggregates behind every figure are published per month and per year in the statistics archive, and the full value framework in the core values list. Corrections and questions about the method are welcome on the contact page, and a correction to this page is worth more than a figure left standing.

Back to the research hub, or take the assessment and add one more ranking to the aggregate.