Rocketmap
Snapshot: October 23, 2026. The snapshot holds 97,756 active, nonduplicate roles at 4,475 companies. The map counts 94,289 open roles at 4,398 companies, and its insights, shared views and agent files count the same: the other 3,467 roles belong to 77 companies whose scraped job board belongs to a different company, posts client roles for others, or repeats one listing, and stay hidden until the source is corrected.
“Startup” on this map means a company in the Rocketlist catalog: venture-backed and bootstrapped companies from pre-seed to late-stage, including 370 acquired and 162 public tech companies in this snapshot.
Sources and placement
Jobs come from company career pages through public ATS APIs, including Greenhouse, Ashby, Lever, Personio and SmartRecruiters. /map places every role at its job location, using verified employer source locations or conservative posting-text matches. A multi-location role counts once in every eligible city and once globally, so city totals do not add up to the headline. Remote-only roles and Location unknown have explicit result lists and no invented city pins. Unknown job locations never fall back to HQ. Company HQ is an optional alternative placement with identical role inventory and filters: city pins and cards then count roles at companies headquartered there, which may be based elsewhere or remote. Unknown HQs are excluded from the globe in that mode. Open roles means active at snapshot time: roles already marked closed are not counted. The Remote bucket includes every explicitly remote-eligible role, including roles that also have verified city evidence; it uses the aligned location-evidence export, while the Remote insight uses the work-mode export. These are separate evidence fields and observations, so their totals can differ.
Rocketlist's curated startup catalog includes venture-backed and technology growth companies. There is no single age, funding or headcount cutoff; the catalog includes mature companies and is not a registry of all startups. Companies are grouped by stable catalog ID. Canonical role IDs and duplicate flags remove known repeated listings. The map retains its frozen role inventory; insights exclude roles marked inactive or duplicate at the data export.
Job-location coverage
Location evidence export: October 23, 2026. Of the 97,756-role inventory, 53,817 have one verified city, 6,798 have multiple locations, 14,298 are remote-only and 22,843 have unknown location. Source strings, unresolved locations and evidence tier remain available; an unavailable posting is not proof of a currently open role.
Posting dates
New listing counts use employer source publication timestamps, never scraper first-seen or updated dates. Greenhouse first_published, Ashby publishedAt and SmartRecruiters releasedDate are publication dates. Lever and Personio supply employer-created dates, labeled with their source provenance. Other ATS dates retain their exact field provenance. First seen records when Rocketlist discovered a listing and is not used in these insights. Unknown source dates: 13.62%. Older and unknown roles remain in the full map; unknown dates are excluded from new listing counts.
Field coverage
Role coverage uses all 97,756 active snapshot roles as its denominator, hidden companies included. Missing means unknown, never false.
- workMode: 97,647 / 97,756 (99.89%)
- seniority: 91,077 / 97,756 (93.17%)
- salaryPublished: 7,867 / 97,756 (8.05%)
- visaExplicit: 1,046 / 97,756 (1.07%)
- sourceDate: 84,441 / 97,756 (86.38%)
Company coverage uses the frozen mapped company inventory.
- investors: 3,949 (88.25%)
- founded_year: 4,376 (97.79%)
- employee_count: 3,972 (88.76%)
- total_raised_usd: 3,817 (85.3%)
- stage: 4,206 (93.99%)
- founders: 2,407 (53.79%)
- hq_country: 4,474 (99.98%)
Insight definitions
- Pay by city: median published annual salary range midpoint by HQ city, role family and currency. Minimum 30 published roles per cohort. Currency-denominated ranges are parsed from source descriptions; hourly, monthly and OTE ranges are excluded. No currency conversion or purchasing-power adjustment. Coverage denominator is all active roles in that city and family.
- Remote share: explicitly remote roles divided by roles with known work mode in the HQ city. Hybrid is separate. Coverage is known modes divided by all active roles. No minimum cohort; n is supplied.
- New listings last 30 days: currently active roles with employer source dates within the 30 calendar days ending on the snapshot day. Grouped by company and HQ city. Measures listings, not hires or net growth. No minimum cohort; n is source-dated roles.
- Junior vs senior: stored junior, entry, intern and graduate labels versus senior, lead, staff, principal, management and executive labels, grouped by HQ city and industry. Shares use the junior plus senior comparison cohort; other known levels stay separate. No minimum cohort; comparisonN is supplied.
- Backer portfolios hiring: companies and active roles with explicitly named YC, a16z, Sequoia, General Catalyst, Accel or Insight backers. Top hirers ranked by open roles. A company can appear in several portfolios. Coverage uses roles at companies with known backers. No minimum cohort.
- Hiring intensity: active open roles divided by reported employee count, minimum headcount 10. Headcount may lag current staffing. Coverage uses qualifying companies divided by active mapped companies.
Every aggregate includes n, denominator, coveragePct and a versioned method ID. Insight values are unfiltered precomputed cohorts; changing filters requires the exact role sidecar, not reusing an incompatible cohort.
Visa sponsorship
Method visa-explicit-v2: explicit English and German phrases in the source description. Sponsors, does-not-sponsor and not-mentioned remain distinct. Requirements to have work authorization do not imply refusal. Conflicting statements and mixed country or role-family policies abstain. A matched phrase and method ID are stored with each classified job. Existing enriched visa booleans are preserved separately.
Review: 200 active roles, stratified across predicted classes. Reviewer: Codex source-context review; not independent human review. Sponsors precision: 100%; does-not-sponsor precision: 100%. These are sample point estimates from agent review, not independent human validation or population guarantees. A separate 300-posting keyword sample measured sponsorship recall at 72.41% and refusal recall at 88.10%; it covers keyword-containing postings, not the whole catalog or every language.
Company history is a separate optional signal. Certified H-1B Labor Condition Applications come from the US Department of Labor disclosure data for FY2025 and FY2026. FY2025 combines all four quarterly files with case deduplication; FY2026 currently ends June 30, 2026. Counts include certified cases only, excluding certified-withdrawn cases, and measure applications rather than workers or hires. US matching requires an exact legal name, employer contact domain, and HQ city and state. Counts are conservative lower bounds for matched addresses in that state. Subsidiaries, parent companies and ambiguous names are excluded.
The UK Home Office worker sponsor register, dated October 8, 2026, is matched by unique exact legal name and city with the catalog website domain as corroboration. The register does not publish domains. Licence routes and ratings remain visible in the data. A 200-match source-record review found 200 correct matches; this is an agent-reviewed sample point estimate. Employer history or a licence does not guarantee sponsorship for an open role. Unmatched companies remain unknown.