Access weekly Series A predictions for Product Hunt launches.
External API requests require an API key passed via the x-api-key header. To request a key, email benchmark@vela.partners.
Returns scored featured Product Hunt launches, ranked by predicted Series A probability. By default only launches verified as not having raised Series A are included. Pass include_funded=true to receive every launch, each carrying afunding_stage and raised_series_a field so you can filter on your side.
| Parameter | Type | Description |
|---|---|---|
| days | int | Look back N days from now. Default: 7 |
| week | string | Monday date in YYYY-MM-DD format. Returns Mon-Sun of that week. |
| include_funded | bool | Include launches that have already raised Series A+. Default false (funded excluded). Each post always carries funding_stage + raised_series_a. |
| Field | Description |
|---|---|
| id | Product Hunt post ID |
| name | Product name |
| slug | PH URL slug |
| tagline | One-line product description |
| votes_count | Total upvotes on launch day |
| comments_count | Total comments |
| daily_rank | Rank on Product Hunt for that day (null if unranked) |
| weekly_rank | Rank for the week |
| topics | Pipe-separated category tags |
| score | Model probability of Series A within 18 months (0-1) |
| above_baseline | Whether score exceeds the 0.78% base rate |
| funding_stage | Latest financing stage: unknown | none | pre_seed | seed | series_a | series_b | series_c_plus | public |
| raised_series_a | True when the company has raised Series A or later (series_a and beyond) |
| funding_status | Legacy flag, derived: funded when raised_series_a, else not_funded |
| created_at | Launch timestamp (ISO 8601) |
| website_domain | Resolved company domain |
No hard rate limits currently enforced. Please be reasonable -- the data updates once per week. Usage is tracked per API key.
Predictions are generated by the PHBench champion model (Top-3 ensemble) trained on 67,292 featured Product Hunt launches from 2019-2025, with 528 verified Series A outcomes from Crunchbase. The model uses 68 engineered features from public launch signals. On held-out test data, it achieves 4.7x lift over random. See the paper for details.