Where do B2B buying signals actually come from?

Almost all usable signals are public. The work is not getting access to rare information — it is monitoring, at the right cadence, sources anyone can read: legal and financial filings, trade and regional press, company newsrooms, professional networks, job postings, and your own CRM. The last one is the most neglected: a CRM holds signals nobody watches, because it gets queried about the present and never about what changed.

At what point does manual monitoring stop working?

Three ceilings appear, always in the same order. Coverage: past roughly a hundred accounts, monitoring silently becomes partial — you check the accounts you happen to remember. Consistency: monitoring is the first task dropped when a deal heats up, and a signal seen three weeks late has lost what made it valuable. Preparation: detecting takes minutes, while qualifying, rebuilding the account context, finding the right person and drafting a credible approach takes far longer. That third ceiling is not solved by hiring an intern to do monitoring.

What should detection tooling actually cover?

Three layers, and most disappointment comes from tooling only the first. Monitoring — continuous coverage of sources with deduplication — is the easiest to automate and the least differentiating, because it produces raw volume. Qualification — discarding what is irrelevant, connecting the event to the account and its history, writing the hypothesis, naming the right contact — is what decides whether the chain is useful at all. Preparation of the action is where most of the human time is actually spent.

Where should you start?

On a narrow scope, by hand. Twenty or thirty accounts, an explicit list of sources, one non-negotiable weekly slot, one written hypothesis per signal kept, and a record of the action decided. That apprenticeship is not skippable: it is what teaches you which signals matter for your offer, and no tool can decide that for you. Tooling comes after, once you know what you are looking for.

All guides

Method guide

Detecting buying signals: sources, method, tooling

Almost every usable signal is public. The difficulty is not access — it is watching ordinary sources at the right cadence, and doing something with what turns up. Here are the sources that pay, the manual method worth starting with, and the three ceilings it will hit.

Where the signals actually live

Nearly all usable signals are public. The work is not reaching rare information; it is monitoring, at the right cadence, sources anyone can read.

Legal and financial filings

Company registry announcements, statutory filings, published accounts, public tender registers. These are the most reliable and least exploited sources: structured, dated, and they reveal subsidiary creations, director changes, acquisitions and distress. Their weakness is latency — they record a decision rather than announce it.

Press and company communications

National business press, trade press, regional press, and the newsrooms of the companies themselves. This is where funding rounds, launches, new sites and partnerships appear, usually before the corresponding legal filing. Regional press is particularly underexploited: it covers industrial investments and site openings that national outlets ignore.

Professional networks

Role changes, new appointments and public commentary appear there immediately. It is the fastest source for organizational signals — an appointment shows up weeks before any official record — and the noisiest, requiring real filtering so that meaningless title changes do not bury what matters.

Job postings

A job posting is a requirements document published early. It names the capability the company wants to acquire, the technology it has settled on, the program it is opening and sometimes the deadline. For a services firm it is often the most directly actionable signal available, because it describes exactly what the company has concluded it cannot do alone.

Your own data

The most neglected source is internal: accounts lost eighteen months ago whose decision-maker has since changed, finished engagements at a client now opening a second site, former contacts who moved to a competitor of your client. A CRM holds signals nobody watches, because it gets queried about the present and never about what changed.

The manual method, and where it plateaus

Manual monitoring works, and it is the right place to start. The protocol fits in five points: define an explicit list of accounts, list the sources to cover for each, fix a non-negotiable weekly slot, write one hypothesis per signal you keep, and record the action decided.

A few dozen accounts can be handled properly this way, and the apprenticeship is irreplaceable — it is what teaches you which signals matter for your particular offer.

Three ceilings then appear, always in the same order.

Coverage. Past roughly a hundred accounts, monitoring becomes partial without anyone deciding that it should: you check the accounts you happen to remember.
Consistency. Monitoring is the first task dropped when a deal heats up. A signal seen three weeks late has lost what made it valuable.
Preparation. Detecting takes minutes. Qualifying the signal, rebuilding the account context, identifying the right person and drafting a credible approach takes far longer. That is where most of the time goes, and that ceiling is not cleared by hiring an intern to do monitoring.

Tooling it: monitoring, qualification, action

A tooled detection chain has three layers, and most of the disappointment in this category comes from only tooling the first.

Monitoring. Continuous coverage of sources across a defined scope, with deduplication. Easiest to automate, least differentiating: it produces raw volume.
Qualification. Discarding what is irrelevant, connecting the event to the account and its history, writing the hypothesis of need, naming the right contact. This layer decides whether the chain is worth anything — an unqualified alert is a burden, not information.
Preparation of the action. Assembling the account picture, drafting an approach that quotes the fact, and putting it in front of the person who will decide whether it goes out.

A chain that stops at the first layer moves the problem: you now receive more alerts than you can qualify, which feels like progress and is not.

Choosing tooling: the criteria that matter

Four questions cut through most vendor conversations. Which of the three layers does it actually cover, and which does it leave to you? Who on your team opens it every morning, realistically? What happens during a week when nobody tends to it? And does it know how to ask for approval before anything leaves — without asking for approval on everything, which turns it into a form?

The five families of approach, and what each one costs in human time, are compared in manual monitoring, sales intelligence or an agentic system.

Where to start

Narrow scope, by hand, for a few weeks. Twenty or thirty accounts you genuinely care about, monitored on a fixed slot, with a written hypothesis for every signal you keep. You will discard most of them, and the discarding is the point: it is how you learn which signals your offer can actually convert.

Only then does tooling make sense, because only then do you know what you are asking it to find. What a signal is and what it does not license you to conclude is covered in B2B buying signals: definition, types and uses.

Frequently asked questions

Where do B2B buying signals actually come from?: Almost all usable signals are public. The work is not getting access to rare information — it is monitoring, at the right cadence, sources anyone can read: legal and financial filings, trade and regional press, company newsrooms, professional networks, job postings, and your own CRM. The last one is the most neglected: a CRM holds signals nobody watches, because it gets queried about the present and never about what changed.
Why are job postings such a strong signal?: A job posting is a requirements document published early. It names the capability the company is trying to acquire, often the technology it has settled on, the program it is opening and sometimes the deadline. For a services firm it is frequently the most directly actionable signal there is, because it describes precisely what the company has decided it cannot do alone.
At what point does manual monitoring stop working?: Three ceilings appear, always in the same order. Coverage: past roughly a hundred accounts, monitoring silently becomes partial — you check the accounts you happen to remember. Consistency: monitoring is the first task dropped when a deal heats up, and a signal seen three weeks late has lost what made it valuable. Preparation: detecting takes minutes, while qualifying, rebuilding the account context, finding the right person and drafting a credible approach takes far longer. That third ceiling is not solved by hiring an intern to do monitoring.
What should detection tooling actually cover?: Three layers, and most disappointment comes from tooling only the first. Monitoring — continuous coverage of sources with deduplication — is the easiest to automate and the least differentiating, because it produces raw volume. Qualification — discarding what is irrelevant, connecting the event to the account and its history, writing the hypothesis, naming the right contact — is what decides whether the chain is useful at all. Preparation of the action is where most of the human time is actually spent.
Where should you start?: On a narrow scope, by hand. Twenty or thirty accounts, an explicit list of sources, one non-negotiable weekly slot, one written hypothesis per signal kept, and a record of the action decided. That apprenticeship is not skippable: it is what teaches you which signals matter for your offer, and no tool can decide that for you. Tooling comes after, once you know what you are looking for.