NEA Portfolio Report

Build a filtered, normalized target book of New Enterprise Associates' portfolio, with a link to NEA's own page behind every single company.

Built for deal sourcing, competitive mapping and BD teams who need a workable list instead of a wall of logos.

Coverage is the headline, not the footnote

This actor sells a filtered book, so the worst thing it can do is hand you a short list that reads like a complete one. Every report states, in the record and on the first line of the brief:

universe_total       914     companies NEA publishes
companies_examined   150     companies this run actually read
coverage_complete    false
coverage_caveat      "maxCompanies=150 bounded the sweep to the first 150 of 914
                      companies in sitemap order"

A company missing from a partial book has not been ruled out — it was never read. Raise maxCompanies toward 914 for a complete book; budget about a minute per 120 companies.

Honouring your page budget perfectly still does not count as coverage. Coverage is measured against NEA's portfolio, not against the run's own limit.

Two things measured live that a naive read gets wrong

1. nea.com/portfolio is not the portfolio

The page the phrase "scrape the portfolio" points at is a 30-logo featured carousel. It carries no company name in text, no sector, no stage and no status — nothing to filter on.

The universe comes from the site's own declared sitemap: 914 /portfolio/ URLs out of 1,333, measured 2026-07-31. Each company's detail page carries the fields that make a target book: sectors, first-invested year, stage, company status and NEA's own one-line description.

2. Values must be sliced between labels, never by token offset

The detail page renders as a flat label/value stream:

Sectors | Technology | Enterprise | First Invested | 2014 | Early
| Company Status | Private | Associated Team | ... | Theme | ... | Website

Two traps, both measured on real pages and both closed:

• A current holding has no status detail. The token after the status is the next label. An offset parser reports every live company as status_detail = "Associated Team".
• An exited company's block is the last one on the page. /portfolio/polyserve has no team, theme or website section, so slicing "to the next label" ran the status detail into the site footer: "Acquired by Hewlett Packard © NEA, 2026 Follow us on social media…" — boilerplate glued onto a real field, on exactly the exited companies a buyer cares most about.

Both are asserted against verbatim excerpts of the live pages in the test suite.

Nothing is inferred

status and stage are carried verbatim from NEA's page. This actor never decides that a company exited, failed, or is still held, and it never scores or ranks anyone. It reports what NEA publishes and links to it.

NEA's published portfolio is the whole of this report. It is not a complete record of NEA's investments — a firm publishes what it chooses to. This is not investment advice.

Filters

filter	what it does
sectors	keep companies tagged with any of these, e.g. Healthcare, Technology
themes	keep companies tagged with any of these, e.g. AI, Infrastructure
stages	stage at first investment, e.g. Seed, Early, Growth
statuses	company status as NEA publishes it, e.g. Private, Acquired
firstInvestedFrom / firstInvestedTo	year range of NEA's first investment
nameContains	free text across company name and NEA's description
maxCompanies	how much of the 914 to read — default 150, a partial sweep

Filtering runs here, over parsed fields, because the source offers no server-side filter at all. Every parameter therefore provably changes the result set, proven control-vs-test in the suite — none of them can be quietly discarded by the site.

Each report also returns vocabularies_in_examined_set: the sectors, stages and statuses counted from the companies this run actually read, not from a hard-coded list.

Output

• one portfolio_company record per match — name, description, sectors, themes, first invested year, stage, status, status detail, and the NEA URL it came from
• one portfolio_report record — the filter, the coverage numbers, the vocabularies
• one source_receipt — terminal, HTTP statuses, request counts, pacing
• a rendered markdown brief in the key-value store

Reading policy

nea.com/robots.txt reads User-agent: * / Allow: / with no crawl-delay. This actor paces itself at 2 requests/second anyway, with a contact User-Agent. The rate is not an input — a buyer cannot dial it up.

Pricing

Event	FREE	BRONZE	SILVER	GOLD+
Actor Start (apify-actor-start)	$0.02	$0.02	$0.02	$0.02
Portfolio company record (portfolio-company-record)	$0.05	$0.045	$0.04	$0.0335

Prices are the filed pay-per-event amounts per plan tier (PLATINUM/DIAMOND match GOLD). Blocked and refused runs do not intentionally charge value events.