Where we are today
Public API, methodology document, anonymisation pipeline, scheduled compute job. Live and verifiable. The first weekly aggregation is pending. We chose to publish the infrastructure first so you can audit it before any number is cited.
Until the first aggregation runs, every endpoint returns { "published": false }. We chose to publish the public infrastructure first and the data second. So you can audit the methodology before you cite a number.
Why a continuous signal
NHS England’s MHCYP is the gold-standard prevalence study. Since 2020 it has been published as annual follow-up waves. The most recent is now two and a half years old, with no successor scheduled. Pulse is a complementary cadence signal designed for the months between.
Pulse is a complement, not a replacement. MHCYP remains the gold-standard prevalence study; Pulse fills the years between waves, and the months after.
What Pulse publishes
Every weekly publication carries a single Wellbeing Score (0 to 100) plus the ten indicator categories that compose it. Each is broken down by the nine English regions plus the devolved nations, and by 19 institution types where the cohort meets the k=50 floor.
The composite score
How students are feeling right now. The traffic-light distribution across the whole cohort. The closest single read on the mood of a generation.
Where things are headed. A school where wellbeing is declining fast needs attention even if today’s number looks acceptable.
Are students actually using the support? High engagement is a proxy for trust. If they are not talking, you cannot measure anything, and the silence itself is data.
Self-efficacy and cognitive resilience. Weighted lowest because it moves slowly. But when it shifts, distress follows months later.
The full publication
Beyond the headline score, every weekly publication carries the indicator categories below. Each is broken down by region (9 English regions plus Scotland, Wales, Northern Ireland) and institution type (19 categories: secondary, PRU, special school, boarding, FE, youth service, children’s home, and so on).
Methodology
Every published number passes through five protective layers. The parameters are documented here in plain language and at /v1/pulse/methodology in machine-readable form. If they ever diverge, the API is the source of truth.
No published number represents fewer than 50 students.
If a cohort is too small, the data is suppressed entirely. The GSS administrative-data baseline is k=5 to k=10. ONS small-number suppression starts at cells under 3. We use 50 because this is mental-health data about children, not census tables.
Calibrated Gaussian noise injected before any number leaves the pipeline.
Smaller cohorts get more noise. Counts, percentages, and scores are all noised. UK Statistics Authority EAP215 (Dec 2024) treats ε ≤ 1 as strong protection; we sit at the conservative end of the published range. Quantisation is applied before noise so true values are not recoverable.
Every published score is a four-week rolling average.
A point-in-time snapshot would be both noisier and more re-identifying. The rolling average reflects how wellbeing actually moves and prevents differencing attacks where someone subtracts one week from the next to isolate a single contribution.
A cell that drops below k=50 stays suppressed for three weeks.
Without hysteresis, a cell flickering across the suppression threshold would itself become a side channel. The three-week hold prevents week-to-week presence/absence inferences. Suppressed cells return null, not a blank, so you cannot infer membership from gaps.
No more than three dimensions can be combined in any query.
Cross-tabs combinatorially shrink cell sizes. Capping at three dimensions (e.g. region × institution-type × week) keeps every published cell above the k=50 floor. Deeper splits would force suppression of nearly everything.
The live methodology JSON is publicly served by the API.
GET /v1/pulse/methodology returns a versioned, machine-readable description of every parameter. This page is the plain-language counterpart. If they ever diverge, the API is the source of truth.
See the full machine-readable document at /v1/pulse/methodology · current version 1.0.
Honest by design
We publish.
Methodology, anonymisation parameters, every change to either. Open API. No registration. No commercial gate. Use it in research, journalism, policy analysis, or your own dashboards.
We don’t paywall.
Pulse is a public good drawn from public-interest data. We will never sell tiered access, license cohort splits to insurers, or restrict the API behind a sales call.
We aggregate.
A minimum cohort of 50 students before any number is published. Differential-privacy noise injected before release. Suppressed cells stay suppressed for three weeks so flicker cannot be a side channel.
We don’t profile.
No individual student is ever identifiable from a Pulse figure. No re-identification by joining datasets. No targeting of a school, a postcode, or a child by anyone, ever.
We complement.
Pulse fills the years between MHCYP waves and the months after each publication. The methodology is documented for OSR-grade scrutiny.
We don’t replace MHCYP.
NHS England’s Mental Health of Children and Young People Survey remains the gold-standard prevalence study. Pulse is a complementary cadence signal, not a substitute, and we publish that limit on every page that carries our numbers.
This isn’t a transparency report.
It’s how the data product is built.
A note on what Pulse is not.
Pulse reflects only institutions using the Poyntr platform. It is not a representative sample of all UK young people. Self-selection bias applies: schools that adopt wellbeing technology may differ from those that do not. The data should be interpreted alongside MHCYP, APMS, and other national surveys, not as a replacement for them.
Until the first weekly aggregation runs, every API endpoint returns { "published": false }. We will not publish a Pulse value before its underlying cohort meets the k=50 floor and clears the differential-privacy noise step.
We publish these limitations because trust is built by saying what you cannot do, not just what you can.
Open API
No API key. No registration. No terms of service beyond a 60-requests-per-minute IP rate limit. Use it in research, journalism, policy analysis, or your own dashboards. Cite Poyntr Pulse alongside the methodology version returned by the API.
Public domain. If you cite Pulse in published work, please name “Poyntr Pulse” as the source and link to poyntr.ai/pulse alongside the methodology version returned by the API.
Stay posted
One email when the first weekly aggregation publishes, including the methodology version and a brief on how to interpret it. No commercial pitch attached.
Notify me