For builders

The weekend build is the easy part. The upkeep is the product.

Feeds, parsers and a few prompts come together fast. What takes years is everything that breaks a little every month, and the checks that catch it. Fundz is that work, already done, one call away.

Or see your own leads first. No credit card. Show me my leads

1. What breaks every month

The maintenance is where the hours go

This is a real month of upkeep on our own data, from the production change log and database, last 30 days to 19 October 2026.

Change log
245

entries in 30 days, each a fix, a guard or a cleanup

Wrong websites
3,089

company rows with a wrong or dead website, found and repaired

Stale re-reports
935

old rounds reported again as if new, caught and held

Re-dated events
54,081

8-K events re-stamped from the filing cover page

Blocked sources
1 of 3

newswires we tried returned access denied to us as well

Running underneath
321

scheduled jobs and 27 kill-switched guards, about 20,000 rows written on a typical day

Upkeep grows with every rung you build Estimate

Hours per week to keep it right, for one engineer.

Estimated upkeep, hours per week, by how much you build: a weekend build 2 to 4, a month build 6 to 10, a quarter build 10 to 20.After a weekend build2-4 h(2-4)After a month of building6-10 h(6-10)After a quarter of building10-20 h(10-20)

These hours are our estimates, not measurements. The full ladder, with what each rung gets you, is in section 6.

2. Plug it in

Keep your stack. Swap the plumbing.

The same fused, checked, scored data is available four ways. Pick the one that fits what you are building.

REST API

api.fundz.net/v1. Funding rounds, SEC filings, 8-K events, executive moves, contacts and more, with the same quality gates as the app. Keyed endpoints need a Pro or Strategic seat; a few keyless teaser endpoints return the top 5 of each cohort.

API references

Agent API

Metered and built for agents and SDR tooling: evidence-backed company signals over REST and MCP at api.fundz.net/v1/agent/mcp.

Agent API

Claude connector

One connector in Claude: daily leads scored to your ICP, filing full-text search, contacts and who you know. Included with Pro and Strategic.

Fundz in Claude · the workflow version

Public accuracy endpoint

Check us before you trust us. Measured accuracy, sample sizes and limits, audited monthly, open to anyone and machine readable.

api.fundz.net/v1/data_quality · data.fundz.net

Also pushes to HubSpot, Affinity, Pipedrive and Slack, with CSV and file exports. Want the short version? Why Fundz, in charts.

3. The join

One company, every signal, scored for you, with the person to call

A feed gives you a list. Fundz gives you a lead: the same company seen through a dozen live signal lanes at once, ranked against your own ICP, matched to your own network, with a named person to reach. Each step below is a system you would have to build and keep running.

Signals fused on one company

Funding rounds, SEC filings, 8-K events, executive moves, launches, contracts, job postings, permits and more resolve to one company record, not a pile of headlines.

12 lanes. In the last 30 days: 5,560 Form Ds, 2,376 funding news items, 2,626 hiring and executive items, 3,501 acquisition items.

Resolved and guarded

Name, domain, state and city collisions are checked so the round lands on the right company, with the right website.

403,192 companies, 129,994 merges, 27 kill-switched guards.

Scored to your ICP

Each new event is scored against your industries, stage, geography, size and exclusions, and the best land in your daily leads.

Per user, every day. Delivered in the app, by email, by API, in claude.ai, and into your CRM.

Matched to your network

Your imported contacts and LinkedIn connections are matched against the people attached to each company and investor, so you see who you already know there.

1,093,153 people on file across 184,364 companies.

Stacked with the other signals

Raised, and hiring in the department you sell into, and a new executive who owns the budget. The stack is the lead, not any one alert.

397,953 job postings seen in 90 days.

The person to call

A named senior contact, with email or LinkedIn where we have them, and the reason to reach out now, with sources.

650,236 source links behind the rows.

Counts are from production on 2026-10-19. The 30-day counts are rows parsed from source feeds, before our guards remove non-rounds.

4. Breadth and accuracy, measured

More rounds, found sooner, with more of the row filled in

We tested this against a leading AI chat assistant and against feed builds over four weeks, then checked the data itself against samples read one row at a time. The headline numbers, each with its sample size.

More rounds
85%

of 52 press-announced rounds surfaced in a 4-week, 3-market test (95% interval 72-92). Newswire feed builds surfaced 27-31% of the same set.

Found faster
0 days

median lag between a round’s announcement and its row appearing in Fundz, in all three markets (109 served rounds).

Decision-maker contacts
40-55%

of served rounds carry a named senior contact in the same test (20, 35 and 54 rounds in the three markets). Contacts also reveal on demand.

Matched to your ICP, network and target accounts
Yours

Scoring and network matching are personal to each account, so there is no single benchmark. The leads preview linked at the end of this page runs it on your market.

More attributes per round

Website, company name, round type, date and amount, present on served rounds, out of five.

Fields present on served rounds out of five (website, name, round type, date, amount): US fintech 4.6, Southeast health 4.5, US B2B SaaS 4.9.US fintech4.6 of 5Southeast health4.5 of 5US B2B SaaS4.9 of 5

Fundz served rounds in the 4-week test, 109 rounds across three markets.

A named senior contact on the round

Share of served rounds with a senior named contact. 95% interval shown.

Share of served funding rounds with a named senior contact: US fintech 55 percent of 20, Southeast health 40 percent of 35, US B2B SaaS 44 percent of 54.US fintech (20 rounds)55%(34-74)Southeast health (35 rounds)40%(26-56)US B2B SaaS (54 rounds)44%(32-58)

Same 109 rounds. Contacts reveal on demand, and not every company has a public one.

Accuracy, lane by lane

Share of rows fully right, on seeded samples of 30, judged by AI against the source. 19 October 2026, 95% interval shown.

Share of rows fully right on seeded samples of 30, judged by AI against the source, 19 October 2026: 8-K events 100 percent, executive changes 93, product launches 87, funding rounds 83.8-K major events100%(88.6-100)Executive changes93%(79.6-97.3)Product launches87%(70.3-94.7)Funding rounds, as served83%(66-93)

The sample is of rows created in the last 12 months. Samples of 30 give wide intervals, so we publish them. Every lane, including the weaker ones, is on our public accuracy page, data.fundz.net.

5. What it really takes

7,936 raw funding items in. 2,399 rounds out.

The 5,560 SEC Form Ds and 2,376 funding news items parsed in the last 30 days are not 7,936 rounds. Most of the work is deciding what is not one, and doing it right every day.

Funding funnel, last 30 days

Rows at each stage. Share is of raw items.

Funding funnel over 30 days: 7,936 raw items, 7,000 rows in the database, 6,805 round rows created, 2,399 served, which is 30 percent of raw.Raw funding items parsed (5,560 SEC Form D + 2,376 news)7,936100%Funding rows that reached our database7,00088%Round rows created6,80586%Rounds served to customers2,39930%

2,399 is 35.2% of the 6,805 rows created, and 30% of raw. Source: production database, 2026-10-19.

Why created rows were held back

Of 6,805 round rows created. Reasons overlap a little, so do not add them to the served count.

Why created rows were not served, share of 6,805 created in 30 days: investment-fund filers 46 percent, held by other guards 17 percent, duplicates 3 percent.Filed by an investment fund, not a company3,16146% of createdHeld by other guards and checks1,13017%Duplicate of another source1813%

Most SEC Form Ds are filed by investment funds raising money to invest, not companies raising to operate. Fundz holds those by design and applies checks to the rest.

The scale behind it

77M

rows maintained across 386 tables and about 49 GB (about 76.95 million)

403,192

companies, with 129,994 merges of the same company seen twice

1.09M

people on file (1,093,153 contact rows) and 52,420 investors

24.3M

13F holdings rows, plus 14.6 million benefit plan filings and 1.9 million business formations

Row totals are database estimates. The backend alone is 209,962 lines of code and 5,597 commits since 2018, 755 of them in the last 30 days.

The accuracy program is most of the work

24+

seeded re-draws, about 700 rows judged one by one against their sources

~21,600

rows hidden or corrected since October 17 (an itemized total from our change log, overlaps possible)

245

entries in our production change log in the last 30 days

6. The honest DIY ladder

A weekend, a month, a quarter

You can build part of this, and the first rung is genuinely quick. Here is what each rung gets and what it still misses. Hours, upkeep and dollars are our estimates for one experienced engineer, at an assumed $100 an hour. They are not measured.

A weekend Estimate

16 h, about $1,600
You get
SEC Form D and 8-K feeds pulled by a script, a prompt to summarize, a filter on entity type.
Upkeep
2-4 hours a week for broken parsing and rate limits.
Closest measured result
A web-search-only assistant surfaced 13% (7-25%) of 52 press-announced rounds.

A month Estimate

120-160 h, about $12,000-$16,000
You get
Form D, 8-K and one or two newswire or RSS feeds, de-duplication, a company-to-website match, a simple ICP filter, an email digest.
Upkeep and services
6-10 hours a week, plus about $1,000 or more a month in scraping and model spend at our volume.
Closest measured result
Newswire feed builds surfaced 27-31% of the same 52 rounds (interval 17-44%).

A quarter Estimate

400-500 h, about $40,000-$50,000
You get
More feeds, entity resolution, an investor table, contacts through a vendor, a review queue and basic accuracy checks.
Upkeep
10-20 hours a week.
Still missing
Eight years of sourced history, merged company and investor records (129,994 and 1,669 merges), access to every wire (one of the three newswires we tried returned access denied to us as well), and an accuracy program.

Engineer hours to get there Estimate

Estimated engineer hours: weekend 16 hours, one month 120 to 160 hours, one quarter 400 to 500 hours.A weekend16 hA month120-160 hA quarter400-500 h

Fundz took about eight years of real use to reach this point. Even a quarter of focused work lands you at the start of the accuracy program, not the end of it.

7. Head to head

Where AI chat is strong, and where it swings

The AI chat with web search on is a genuinely good tool, and we say so. These are the results from our own tests, with the limits stated.

Press-announced rounds, 4 weeks

52 strict rounds, US fintech, Southeast health and US B2B SaaS, 21 Sep to 18 Oct 2026. 95% interval shown.

Press-announced funding rounds found, 4-week test, 52 rounds. Fundz 85 percent, AI chat 77, Newswire 1 feed build 31, Newswire 2 feed build 27, web-search-only assistant 13.Fundz85%(72-92)AI chat (web search on)77%(64-86)Newswire 1 feed build31%(20-44)Newswire 2 feed build27%(17-40)Web-search-only assistant13%(7-25)

Fundz 85% (72-92%) against AI chat 77% (64-86%). The intervals overlap, so call it close: both are strong here. A newswire-category crawl or a bare web search is far behind.

It depends on the market

The same test, split. Southeast health is left out because its answer key leaned on Fundz data.

By market, press-announced rounds. US fintech (n=12): AI chat 100 percent, Fundz 75. US B2B SaaS (n=36): Fundz 89, AI chat 72.US fintech, Series A-C (n=12): AI chat100%(76-100)US fintech (n=12): Fundz75%(47-91)US B2B SaaS, Series A-C (n=36): Fundz89%(75-96)US B2B SaaS (n=36): AI chat72%(56-84)

The AI chat found all 12 US fintech rounds. Fundz led in US B2B SaaS. Small samples, wide intervals.

SEC Form D filings, checked against EDGAR itself

US companies with an “Other Technology” Form D and money sold, a proxy for software. Every original Form D filed in the window, matched to EDGAR.

Fundz servedAI chat surfaced
SEC Form D filings by US Other Technology companies with money sold. Week of Oct 5 to 11, 74 filings: AI chat 29.7 percent, Fundz 94.6. Week of Oct 12 to 18, 63 filings: AI chat 68.3, Fundz 98.4. Both weeks, 137 filings: AI chat 47.4, Fundz 96.4.Oct 5-11 (74 filings)94.6%(86.9-97.9)29.7%(20.5-40.9)Oct 12-18 (63 filings)98.4%(91.5-99.7)68.3%(56-78.4)Both weeks (137 filings)96.4%(91.7-98.4)47.4%(39.3-55.8)

Fundz served 96.4% (132 of 137, interval 91.7-98.4). A leading AI chat assistant, given web access and Form D in the prompt, surfaced 47.4% (65 of 137, interval 39.3-55.8), and its two weeks differed by 38 points (29.7% and 68.3%). One account, one prompt per week.

Why this matters. A search of 40 randomly chosen US operating-company Form Ds found no press release or news article for 37 of them (92.5%, interval 80.1-97.4). Many raises are only in the filing, and a filing feed is a different build from a news feed.
Where AI chat did very well. Every company it named across four runs was a real EDGAR filing. For Southeast health Form Ds it named all 27 companies, twice, while Fundz served 24 of 27. For a small, specific list, it is a fast way to get a first pass.

Across all US operating-company Form Ds with money sold, Fundz serves 67.8% (660 of 973), or 87.3% (660 of 756) once property and investment vehicles, which our feed excludes by design, are set aside.

See your market before you decide

Tell us who you sell to. We show you the companies that fit this week, the signals behind each one, and who to reach, then recommend a plan. No credit card.

How we measured, and what these numbers are not

  • 4-week test. 21 Sep to 18 Oct 2026, three markets: US fintech and US B2B SaaS (Series A to C) and Southeast health (any round). Strict means the round fully meets the market definition. Recall is measured against a hand-judged union of everything every method found, so it is an upper bound on the true population, not a share of all rounds that exist.
  • Feed builds. The newswire builds are title-match crawls of two of three newswires. The third returned access denied and is not in the results. Not every item was opened at its primary source.
  • SEC Form D test. Ground truth is EDGAR: 3,460 original Form Ds filed 21 Sep to 18 Oct 2026, 1,384 from operating companies. “Other Technology” is a category filers choose, not a software label. The AI chat ran in one logged-in account, one prompt per week, on 19 October 2026, and could differ on a rerun.
  • Accuracy. Seeded samples of 28 to 39 rows per lane, committed before judging, judged by AI reading the source one row at a time. Intervals are 95% Wilson. Funding rows are shown as served by our API on 19 October 2026.
  • Estimates. Engineer hours, upkeep, the $100 hourly rate and dollar costs in the DIY ladder are estimates, not measurements. Row totals are database planner estimates.
  • Speed. The 0-day lag is Fundz row creation against the announcement date. We did not measure how long a chat assistant takes to answer.

Why this matters

Being first

Buyers settle on a shortlist before they talk to anyone, so the aim is to be known early and usefully, not to pitch first.

  • 94%ranked their shortlist before contacting any seller6sense 2025
  • 95%of winning vendors were on the Day One shortlist6sense 2025
  • 85%had worked with or evaluated the vendor they chose6sense 2025

Source and limits. About 4,000 B2B buyers, a self-reported survey by a vendor. It shows a pattern, not proof. Read the 6sense report.

Why this matters

Who you already know

Buyers often already know people at the vendors they consider, so who you know at a company is a real way in.

  • 75%personally knew sellers at the vendors they evaluated6sense 2025
  • 85%had worked with or evaluated the vendor they chose6sense 2025

Source and limits. About 4,000 B2B buyers, a self-reported survey by a vendor. It shows a pattern, not proof. Knowing someone is not the same as getting an introduction. Read the 6sense report.