Alternative Data

Custom alternative data, built to your spec

When the dataset you want is not for sale, sieve builds it: sourced from websites, PDFs and public records, normalized into one schema, and updated on a schedule.

https://example-store.com/products
Data from API (not visible on page)
🌐 Intercepted
API Response Data:
{
"inventory": "147 units",
"supplier": "Acme Widgets Inc",
"lastUpdated": "2 hours ago",
"storeLocation": "New York, 5th Ave"
}
💡 This data is not visible on the page

Extracted Data

Capture URL

🔗

example-store.com/products

Extract Name

Premium Widget Pro

Extract Price

$299

Inspect Network

🌐 From API (not on page):

inventory:147 units
supplier:Acme Widgets Inc
last Updated:2 hours ago
store Location:New York, 5th Ave
Extraction Complete4/4 fields

Tell us the dataset you wish existed

Describe the question, the universe and the frequency. We will come back with sources and a sample.

Our clients

Trusted by leading hedge funds

Powering data-driven decisions for the world's most sophisticated investment firms

$100B+

in Assets Under Management

100K+

Data Points Extracted & Validated

The useful datasets are the ones nobody sells yet

Off-the-shelf alternative data is priced for everyone and seen by everyone. The signal a fund wants often sits in public sources that were never meant to be combined: port authority releases, company websites, government portals, industry reports. Building that in-house means scrapers, parsers, and an engineer on call when they break.

  • •Fragmented sources: there is no central database, so the data has to be found source by source.
  • •Mixed formats: PDFs, static pages and filtered web interfaces, often in several languages.
  • •Inconsistent reporting: one source splits a number three ways, the next reports only a total.
  • •Pipelines that break: a format change at 5:30 AM becomes the data engineer's emergency.

How it works

1

Scope the dataset

Tell us the question you want answered, the universe, the fields and the frequency. We come back with the sources and a sample.

One call to scope.

2

We source, extract and normalize

sieve collects from every source in whatever format it publishes, then maps it all into one schema with consistent units and names.

Web pages, PDFs, portals.

3

You get the dataset and the updates

The history arrives first, then new data on a schedule, through the API, as CSV or JSON, or in a spreadsheet.

Backfill, then recurring.

Proprietary by construction

A dataset built around your thesis, not a feed every other fund already licenses.

Every format, one schema

PDFs, websites and filtered portals in several languages end up as one clean table.

Traceable to the source

Each value links back to the table and page it came from, so an analyst can check any number in seconds.

Reviewed on enterprise plans

On enterprise plans, domain reviewers verify each data point before it reaches your models, because one bad value can move a signal.

Datasets we have built for investors

  • Global container port volumes (TEU) assembled from port-level records to track trade flows as tariffs changed
  • Commodity export data from national statistics releases, available before vendor feeds update
  • Production time series from industry publications, backfilled for years and updated on release
  • New filings and events for small-cap companies outside Bloomberg's coverage
  • Pricing, product and hiring changes on company websites, checked daily

Our investors

Backed by the best

backed by world-class investors who believe in our vision

Y Combinator logo

Questions people ask

Not a catalog vendor. We do not sell one fixed dataset to many funds. We build the dataset you specify from public web and document sources and deliver it to your team.

It depends on the number of sources. In the port records project, the first clean, normalized dataset arrived in days, where the fund had expected weeks of manual work.

Through the API as CSV or JSON, as a spreadsheet, or by email. Recurring updates run on a schedule, and you can get a webhook when new data lands.

Yes. Funds start with one dataset and a sample before committing, and the same agent is self-serve at scrape.usesieve.com for teams that want to run collections themselves.

Custom alternative data, built to your spec

2025 Sieve Data Inc. All Rights Reserved.