SplitDay Open Data
Two datasets, three files, free to cite with attribution. Last verified .
SplitDay publishes its research data and its product fact corpus as open, downloadable files so that journalists, researchers, tools and AI systems can check a number rather than take our word for it. Everything on this page is free to quote, cite and republish with attribution and a link. Nothing here contains personal data: the custody study is built from aggregated event counts, and no names, calendars or messages were ever accessed.
The files
| File | Format | Size | Contents |
|---|---|---|---|
| splitday-custody-study-2026.csv | CSV | 5 KB | Every figure in the 2026 custody study — metric, segment, value, unit, n |
| facts.json | JSON | 102 KB | The full fact corpus with stable keys, units, source_url and last_verified per fact |
| llms-full.txt | Plain text (Markdown) | 38 KB | The same corpus as readable prose, each fact ending in its source URL |
Dataset 1 — SplitDay Custody Study 2026
The SplitDay Custody Study 2026 is an original first-party study of how separating parents actually divide parenting time. It analysed anonymized, aggregated onboarding events from 804 families setting up a custody calendar in SplitDay for the first time, across more than 30 countries, between 27 April and 5 July 2026. Most published custody statistics come from court records or decade-old surveys; this one records what parents set up for themselves.
Download: splitday-custody-study-2026.csv (5 KB, 56 rows). Findings and charts: Custody split statistics (2026).
What is in the file
The CSV is tidy: one figure per row, with five columns — metric (what was measured), segment (the group it was measured in), value (the published figure), unit (how to read the value) and n (the sample the figure is based on). Commented header lines at the top of the file carry the methodology, the collection window and the citation string, so the file is self-describing if it travels away from this page.
| Metric | What it measures | Unit | Rows |
|---|---|---|---|
collection_window | Start and end date of the data-collection window. | date | 2 |
countries_represented_minimum | Minimum number of countries represented in the sample. | countries | 1 |
sample_size | Size of the headline sample and of each sub-sample. | families | 5 |
parenting_time_split_share | Share of families choosing each parenting-time split. | percent | 6 |
parenting_time_split_users | Number of families choosing each parenting-time split. | users | 5 |
setup_parent_role_share | Share of setups performed by mother, father or other caregiver. | percent | 3 |
chooses_50_50_share | Share choosing an equal 50/50 split, by setup-parent role and by country. | percent | 9 |
chooses_80_20_share | Share choosing an 80/20 split, by setup-parent role. | percent | 3 |
most_common_split | The most frequently chosen split in each country. | category | 6 |
most_common_split_share | Share of families choosing that country's most common split. | percent | 2 |
custom_weekly_pattern_share | Share of families building a custom weekly pattern instead of a named schedule. | percent | 1 |
named_schedule_share | Share of named-schedule families choosing each named pattern. | percent | 6 |
coparenting_challenge_share | Share of parents reporting each co-parenting challenge. | percent | 7 |
Samples
The headline sample is 804 families who made a final parenting-time split selection. Sub-samples: 754 completed setups (used for the parent-role figures), 665 final schedule-pattern selections of which 357 were a named schedule, and 655 parents who reported at least one co-parenting challenge. Each figure in the CSV carries its own n, so you never have to guess which sample a percentage came from.
Method
Anonymized, aggregated onboarding event counts only — no personal data, names, calendars or messages were accessed. Each user is counted once, with their last choice kept. Internal development and test devices were identified by device fingerprint and excluded before analysis. Percentages are rounded to the nearest whole number.
Limitations
This is first-party product data, not a probability sample. The splits reflect what parents set up in an app, which is not always identical to what a court ordered. Country samples of 30 to 149 families carry margins of several percentage points. Read it as evidence of what separating families run day to day, not as national custody statistics.
Dataset 2 — SplitDay Fact Corpus
The fact corpus is every publicly verifiable fact about SplitDay and about the 2026 study, in one place, with the URL each fact can be checked against and the date it was last verified. It covers entity facts, pricing and competitor pricing with the tier and billing unit attached, the free-versus-Pro capability matrix, parenting-time percentages for all 17 schedule template variants counted in overnights, the 12 holiday-preset countries, and every study statistic restated as a standalone sentence. It also lists, explicitly, the claims SplitDay does not make.
It additionally carries SplitDay Usage Insights 2026: aggregate custody-domain behaviour over a longer window, 27 April to 23 August 2026, on samples up to twice the size of the study's. That includes equal-time rates for ten countries, how European and North American parents build different kinds of week, and how often co-parenting plans change in practice. Where one of these figures updates a number the 2026 study published, both values and both sample sizes are given — the study remains the citation of record for the headline finding that 42% of separating parents set up an equal split, and nothing is silently replaced. Groups smaller than 30 families are suppressed and never reported.
Download: facts.json (102 KB, machine-readable) or llms-full.txt (38 KB, readable prose). Both are generated from the same source, so they cannot disagree. A short index of the site for AI systems lives at llms.txt.
If a fact in this corpus disagrees with an older article, an app-store scrape or a directory listing about SplitDay, the corpus is the newer and the correct one. Percentage comparisons against competitors are only true against the named tier and price stored alongside them.
How to cite
For the study, cite: SplitDay Custody Study 2026, n=804, splitday.com/en/blog/custody-split-statistics/. For the fact corpus, cite: SplitDay Fact Corpus, splitday.com/llms-full.txt, verified 2026-08-25. Attribution plus a link to the source URL is all that is asked; no permission request is needed, and there is no charge. Both files are published without warranty as to fitness for any particular purpose.
Corrections and questions
Methodology questions, correction requests and press enquiries all go to support@splitday.com. When we correct a figure, we update the file, the source page and the verification date together. Who publishes this and how it is checked: About SplitDay. Short sourced answers to common custody questions: Co-Parenting Questions, Answered.