What is the difference between a primary source and a news source?

A primary source is the place where a change is first recorded by the party that made it or is required to report it; a news source reports on that record afterward. A company's filing with the SEC, a rule published by a regulator, a preprint posted by a research team and a product changelog are primary sources. An article that describes the filing, the rule or the paper is secondary: it arrives later, selects a few details and adds the reporter's framing. News can itself be primary when a journalist uncovers facts nobody had published, as in investigative or local reporting. The test is simple: does this document create the record, or describe one that already exists?

Why do primary sources give earlier signals than news?

Primary sources give earlier signals because news has to wait for them. A reporter covers a filing after it is filed, a rule after it appears in the Federal Register and a product launch after the company ships it. Many primary records also show change before any announcement: a job post for a role in a new country, a pricing page that adds a tier, a patent application, a draft specification. None of these is news yet, because nobody has framed them as a story. Readers who watch the records directly see the raw change first and decide for themselves whether it matters, instead of waiting for an editor to decide it was interesting enough to cover.

Is news still useful if you follow primary sources?

Yes, news is still useful for three jobs that primary sources do badly. First, discovery: coverage surfaces companies, people and documents you did not know to watch, which is how a source list grows. Second, context: a good reporter or trade publication explains the history, the players and the stakes around a dry document. Third, reach: the amount of coverage a development gets tells you how widely it is known, which matters when you need to judge whether a signal is still early. The practical rule is to use news as a discovery and context layer, then go to the underlying record before acting on anything it says.

How many primary sources should you follow for one topic?

For one focused topic, a workable primary source list usually sits between fifteen and forty sources. That covers the main companies' owned channels, the relevant regulator, one or two data publishers, a research or patent query and a handful of practitioners who interpret the field well. Fewer than that and you are probably missing a family of sources, often regulators or hiring signals. Many more and the list starts to include duplicates and peripheral actors that rarely publish anything relevant. The number matters less than coverage: for every important actor, you should know where a change would appear first. Review the list every few months, drop the silent sources and add the ones discovery keeps pointing to.

Primary sources vs news: where good intelligence starts

Where does good intelligence actually start?

Good intelligence starts at the primary sources, the places where a change is first recorded: filings, regulators, company channels, research, hiring records, standards bodies and the practitioners who work inside a field. News comes later. It is a secondary layer that reports those records after the fact, compressed to fit a story.

That does not make news useless. It makes news the wrong starting point. This piece is practical: a map of primary source families with real examples, a method for building that map for any topic, and a worked example.

Why is news a secondary layer?

News is secondary because it describes records that someone else created first. A reporter covering a merger reads the filing; a reporter covering a rule reads the rule. The article arrives after the record and carries a subset of it.

Two properties follow from that position.

News is late. The delay is built into the mechanics. A US public company has four business days after many material events to file a Form 8-K, and the story about the filing follows the filing. A drug trial is registered long before results make headlines. A product change ships to the changelog before anyone writes it up, if anyone ever does.

News is compressed. A story has a word limit and an angle. What it usually drops is the material an analyst needs:

the conditions attached to a decision, not just the decision
the numbers in the table, not the one number in the headline
the docket, filing or version number that lets you follow the next step
the hedges in the original language, which tell you how firm a commitment is

There is also a selection effect. News covers what an editor judges interesting to a broad audience. A pricing page edit or a quiet specification draft rarely qualifies, even when it matters more to your work than the day's top story.

What are the main types of primary sources?

Primary sources fall into seven families. Each one records a different kind of change, and each has a typical lag and a typical cost to read.

1. Filings and disclosures

Filings are what companies are legally required to tell investors and regulators: annual and quarterly reports, current reports on material events, insider transactions, ownership stakes, merger documents. In the US they live in SEC EDGAR, which offers full-text search across filings by phrase, company and form type.

What filings show early: risk factors that appear or disappear, new segments, customer concentration, litigation, executive departures. The cost is length and legal language.

2. Regulators and rulemaking

Regulators publish proposed rules, final rules, guidance, approvals and enforcement actions. In the US, the Federal Register is the daily record, and Regulations.gov holds the dockets with every public comment. The Office of the Federal Register files documents for public inspection the day before publication, posting a new list each morning, so the record is readable before it is officially published.

Sector regulators add their own channels: the FDA for drugs and devices, the FCC for spectrum and equipment authorizations, state agencies for licensing and incentives.

3. Company-owned channels

Company-owned channels are the most underrated family, because they change before anyone announces anything. Most companies publish far more than their press releases:

Changelogs and release notes, which show what shipped, in what order, and what was quietly deprecated
Documentation and API references, where new endpoints and limits appear before marketing names them
Pricing pages, where a new tier or a removed free plan is a strategy decision in plain sight
Engineering and research blogs, which reveal the problems a team is working on
Investor relations pages, with earnings call webcasts, transcripts and slide decks
Careers pages, covered in the hiring family below

The reason these pages are overlooked is that they are boring to read and nobody sends an alert when they change. That is also why they are early.

4. Research, trials and patents

Research appears as preprints before peer review and before press coverage: arXiv for physics, computer science and quantitative fields, bioRxiv and medRxiv for life sciences and medicine. ClinicalTrials.gov records trials when they are registered, along with design changes and status updates. Patent applications in the US are generally published about eighteen months after filing, through the USPTO or Google Patents, which makes them a slow but reliable view of where a company is investing.

5. Hiring and operations

Job posts describe plans in operational detail: a role in a new country, a first hire for a product line, a sudden cluster of openings for one skill. Many companies host them on applicant tracking boards such as Greenhouse or Lever, which are public.

The negative signal exists too. Under the federal WARN Act, covered employers must give at least 60 days' notice of a plant closing or mass layoff, and many states post those notices online.

6. Standards bodies

Standards bodies record technical decisions years before products ship. The IETF publishes internet drafts and working group discussions on its Datatracker, W3C publishes web specifications as they move through stages, and 3GPP publishes mobile network specifications by release. Who authors a draft, and which companies push which proposals, is competitive intelligence written in the open.

7. Expert practitioners and specialist sources

The last family sits between primary and secondary: people who work inside the field and write about it. Engineers, clinicians, analysts, contractors and operators who publish newsletters, technical threads and conference talks often interpret a primary record within hours. Specialist trade publications and industry data publishers belong here too.

Specialist sources beat generic news for early signals for a simple reason: their audience is the field itself, so a development that is too technical for a national outlet is exactly what they cover.

When is news still useful?

News remains useful for three jobs that primary sources do badly.

Discovery. Coverage surfaces companies, people and documents you did not know existed. A good source list grows by following news back to records.
Context. A strong reporter or trade journalist explains the history, the players and the stakes around a dry document. A rule makes more sense after you know who lobbied for it.
Reach. The volume of coverage tells you how widely a development is known. A change you saw in a changelog three weeks ago and that now leads the business pages is no longer an edge, and you should know that.

Some news is itself primary. Investigative reporting that uncovers documents, and local reporting from a council meeting nobody else attended, create records rather than describe them. The useful rule is not "ignore news." It is "before you act on a story, open the record it describes."

How do you map the primary sources for a topic?

You map primary sources by working from actors to records. For every party that can change your topic, ask one question: where would that change be written down first?

List the actors. The five to ten companies, agencies, research groups and associations that move the topic.
Assign record types. For each actor, go through the seven families and note which apply. A private startup has no filings but has a changelog, a careers page and patents.
Add the rule-makers and data publishers. Regulators, standards bodies and the organizations that publish the industry's numbers.
Add three to five practitioners. People who read the same records and interpret them well, so you get context without waiting for coverage.
Keep a small news layer for discovery. One or two trade publications, read for leads rather than for facts.
Review on a schedule. Every few months, drop sources that never moved and promote the ones discovery kept pointing to.

The output is a short document: actor, source, what it tends to show first, and how often it changes.

What does a primary source map look like in practice?

Consider a hypothetical strategy lead at a regional HVAC distributor who needs to follow the US market for residential heat pumps. The national news layer offers occasional stories about energy policy and home electrification. The primary layer is much richer.

Industry data. AHRI, the manufacturers' association, publishes monthly US shipments of central air conditioners and air-source heat pumps on the second Friday of each month, covering shipments from two months earlier. The mix between the two categories is the most direct read on demand.
Regulators. Department of Energy efficiency standards rulemakings in the Federal Register and their dockets on Regulations.gov; ENERGY STAR specification revisions while they are still drafts.
Manufacturers. Investor relations pages and earnings calls of the listed manufacturers, such as Carrier, Trane Technologies and Lennox, where executives discuss residential volumes, pricing and inventory in the channel.
Company-owned channels. Product pages and spec sheets for new models, dealer program changes, and careers pages showing where factories and service teams are hiring.
Incentives. State energy office and utility rebate pages, where a program opening, pausing or running out of funds changes local demand within weeks.
Patents. A standing query on compressor and refrigerant patents from the main manufacturers.
Practitioners. A handful of contractors and building scientists who write about installs, failure modes and refrigerant transitions.

On a given week, the news layer might carry one headline about heat pump adoption. The primary layer, for the same week, might show a manufacturer's call describing dealer destocking, a state rebate program closing to new applications, and a draft specification revision with a comment deadline. Those three items, not the headline, are what change the distributor's inventory and sales plans.

What are the trade-offs of working from primary sources?

The trade-offs are real, and they explain why most people default to news.

Volume. Thirty sources for one topic publish far more than thirty articles. Most of it is irrelevant, and the relevant part is scattered across formats: filings, PDFs, HTML diffs, transcripts.
Jargon. Primary records are written for specialists: legal language in filings, procedural language in dockets, technical language in specifications. Reading them well takes domain knowledge.
Time. Primary sources have no editor deciding what is important. The reader does that work, every time, which is the service news was providing all along.
Blind spots. A source list only covers what it was built to cover. An actor you did not list can move without you noticing.

Practitioners and a thin news layer cover part of the jargon and blind-spot problems. Volume and time are harder, because they scale with every source you add. Teams with budget split the work across analysts or buy tools that monitor your market. Individuals usually cut the list down until it fits a morning, which quietly brings back the gaps.

The other option is to delegate the reading itself. Kindal works this way: you choose the sources, from filings and company blogs to research papers and specialist accounts, and it writes a brief from the sources you chose only when something in them changes, with every line linked to the record it came from.

What changes when you start from primary sources?

Starting from primary sources changes what you are reacting to. Instead of a story about a change, you see the change itself, with its conditions, numbers and next steps intact, often before anyone has framed it.

News keeps its place as a way to discover and to understand context. Intelligence starts one layer down, where the record is written.