The gap
NHS England’s MHCYP has been published as annual follow-up waves since 2020. The most recent wave is from November 2023. APMS runs roughly every seven years (2023/24 most recent, Part 2 published November 2025). Both are gold-standard prevalence studies. Neither carries weekly cadence, and neither places mainstream secondary, PRU, special, hospital, boarding, FE, and youth-service settings inside the same instrument.
Pulse is a continuous complement, not a replacement. It cannot answer prevalence questions MHCYP is built for. It can answer cadence and cross-setting questions MHCYP is not.
Getting access
Public API today. Beyond that, no researcher portal, no accreditation workflow, no automated approval pipeline. We will not pretend otherwise. Anything past Tier 1 is negotiated case by case.
National, regional, and institution-type wellbeing aggregates, updated weekly. No API key. No registration. 60 requests per minute per IP. The methodology JSON is served live alongside the data.
k=50 · ε=1.0 Gaussian · 4-week rolling · 19 institution types · 9 English regions plus the devolved nationsRead the API referenceThere is no /v1/research/* route today. Anything beyond the public Pulse aggregates is negotiated case by case. Send a one-page proposal: research question, ethics-committee status, data needed, intended outputs, named PI. We will reply within ten working days.
No portal · no accreditation workflow · no fee · email [email protected]Propose a partnershipFor matched cohorts, custom aggregation pipelines, or tracking specific interventions over time, we work alongside you as data engineers. You hold the research design and the publication. Your name on the paper. We do not retain editorial control.
Same proposal route · separate scoping conversation if it fits · honest about capacityOpen a conversationTier 1 is real code. Tiers 2 and 3 are honest pathways. We chose to label them as proposals because there is no portal, no automated approval pipeline, and no commercial gate. If we ship coded researcher access in future, this panel will be the first thing that changes.
Open empirical questions
Each is testable with the cadence and cross-setting breadth Pulse provides. We have no published findings on any of them. They are research questions for academic partners to lead, not claims we are making about results.
Everyone assumes January and exam season. No-one has weekly population-level data across a full academic year. A continuous signal could map the real risk calendar and let pastoral teams prepare instead of react.
A school adopts a Mental Health Support Team. A local authority changes its pastoral policy. Did wellbeing measurably shift in the affected cohort, against matched controls? Pulse cadence and 19-institution-type coverage make this question testable for the first time.
Pupil referral units, special schools, boarding schools, hospital schools. Same age groups, very different environments. A common measurement framework lets you compare across settings without different surveys, different teams, different definitions.
Students type one thing and say another. Text and voice carry different signals, including emotional tone. When the two disagree, which one predicts what happens next? That question has not had data behind it before.
Cognitive shifts often precede distress: catastrophising, withdrawal, disengagement. If those shifts are detectable early enough at population level, intervention timing could move forward. The dataset exists to test whether that is real or romantic.
A boarding school at 10pm is not the same as a secondary school at 2pm. Setting alert thresholds is a calibration problem with real costs in both directions: missed cases and pastoral-team alarm fatigue. Cohort-aware thresholds are an open empirical question.
We have no published findings on any of these. They are open empirical questions we would like academic partners to lead. If you propose a study on one of them, we will support it without claiming co-authorship or editorial control.
Five Safes posture
The Five Safes (Desai, Ritchie & Welpton, 2016) is the de facto framework for trusted research environments in the UK, used by ONS, UK Statistics Authority, HDR UK, and ADR UK. We map honestly against each pillar. Safe Data and Safe Outputs are enforced in the publish pipeline. Safe People, Projects, and Settings are partnership-negotiated until their workflows ship.
Two of five satisfied in code. Three handled by partnership today. We will not relabel a row until the corresponding workflow is shipped and audit-able.
Honest by design
We support replication.
Methodology, anonymisation parameters, and the live JSON at /v1/pulse/methodology are versioned and dated. Any change is announced. The same query at the same week should return the same value, forever.
We don’t selectively report.
No internal pre-screening of releases for tone. Negative findings, null findings, methodology limitations all get the same publication channel as positive ones.
We disclose limitations.
Self-selection bias of opted-in institutions, sample composition, k=50 suppression effects, ε=1.0 noise on small cells. Documented inline with every figure.
We don’t hide the gaps.
Suppressed cells return null, not blank, so a researcher can see where coverage is thin. Pre-publication state is admitted on every page that carries our numbers.
You publish what you find.
No editorial review. No right of first refusal. No embargo unless you ask for one. If your work shows that something we built is not working, we want that out in the literature.
We don’t hold the pen.
Authorship is yours. We are happy to be named in acknowledgements when we contributed materially to data engineering. We do not negotiate for co-author lines or for veto rights over wording.
This isn’t a partnerships brochure.
It’s how the research relationship is built.
Every young person who talks to Poyntr is trusting us with something private. That trust extends to anyone who touches the data. Four rules apply, and they are not flexible.
Your ethics committee signs off first. Before any access beyond the public Pulse API, your institution’s research ethics committee approves the work. We do not shortcut this. If your university says no, the answer is no.
Your methods section names the anonymisation. Any publication using Pulse data includes the methodology version returned by /v1/pulse/methodology, plus the limitations we publish. We supply boilerplate you can drop in. Hiding methodology is not allowed.
Re-identification ends the relationship. Attempting to work backwards from aggregate data to identify a student or a school terminates the agreement immediately. Not after a warning. Not after a conversation.
Publish whatever you find. No editorial review, no embargo unless you ask, no veto. If the data shows something we built is not working, we want that in the literature. Public benefit beats reputational comfort.
Start a conversation
One-page proposal: research question, ethics-committee status, data needed, intended outputs, named PI. Reply within ten working days. No commercial pitch attached.
Propose a study