The buying signals glossary: 44 terms, defined plainly

Terms in this category are used loosely enough that two vendors can describe opposite products with identical words. This is the working vocabulary, defined as precisely as we can manage, with a note where common usage and useful usage part company.

Where a definition is contested, it says so rather than picking a side quietly.

Signals and evidence

Buying signal

Public evidence, attributable to a specific party and dated, that a problem you can solve is currently costing somebody enough that they are acting on it. The key property is that it does not require the person to know your company exists. Fully worked through in anatomy of a buying signal.

Signal

Used by vendors to mean almost anything observable. When you see it in marketing, ask which of these it means: a public artefact you can read, an inference from anonymised behaviour, or a first-party event in your own systems. Those three support completely different actions.

Trigger event

A discrete, dated occurrence that changes a company's likelihood of buying: a funding round, a leadership hire, an office opening, a product launch, an outage. Most trigger events are one inferential hop away from a buying signal rather than being one themselves.

Intent signal

Loosely, any evidence of purchase intent. In practice the phrase has been captured by the third-party intent vendors and usually means a topic surge rather than a readable artefact.

Evidence

The actual text (the post, the review, the job ad) as opposed to a score derived from it. A signal you cannot quote is an inference.

First-party intent

Behaviour in systems you own: website visits, pricing-page views, docs reads, trial signups, product usage. The highest-confidence intent there is, and the smallest in volume, because it requires that the person already found you.

Second-party intent

Intent data sourced from a specific publisher or marketplace about activity on their property, review-site category browsing being the common example. Higher resolution than third-party, narrower coverage.

Third-party intent

Account-level inference of topic research, assembled from anonymised content consumption across a network of publisher sites. Gives you an account and a score. Gives you no person, no quote and nothing to reply to.

Company Surge

Bombora's term, and the reference implementation of third-party intent. It reports when an account's research on a topic over roughly a three-week window rises against that account's own twelve-week baseline. The baseline being per-account is the clever part: it corrects for companies that read a lot about everything.

Data co-operative

The mechanism behind third-party intent: a network of publisher sites that share anonymised visitor behaviour in exchange for access to the pooled result. Bombora's runs to over 5,000 B2B sites.

Dark funnel

Buying research that happens where you cannot measure it: private Slack communities, podcasts, word of mouth, Reddit threads, LinkedIn comments. The term is usually deployed to sell you a way to measure part of it. Most of it remains genuinely unmeasurable.

Social listening

Monitoring public conversation. In vendor usage it almost always means monitoring mentions of your name, which is brand monitoring. See social listening for why the distinction decides which product you should buy.

Brand monitoring

Counting and characterising mentions of your brand, with sentiment and reach. A legitimate and well-served product category. Not lead generation, because every mention is by definition somebody who already knows you.

Mention

An occurrence of a tracked string. The unit most listening tools meter on, which tells you what they are built to maximise.

Review mining

Reading public reviews (G2, Trustpilot, Google) as evidence of an operational problem. Especially productive for selling to local businesses, where a one-star review is a customer naming the problem publicly and with a date attached. See local lead generation.

Fit and targeting

ICP

Ideal customer profile. The description of the companies you sell to best. Distinct from intent and frequently confused with it: fit says whether someone should buy, intent says when.

Fit versus intent

The two axes of a prioritised list. High fit and low intent is a nurture target. Low fit and high intent is a distraction that will waste a rep's week. Only high on both is worth an interruption.

Firmographics

Company attributes: size, industry, geography, revenue, structure. The oldest targeting dimension and still the backbone of most ICP definitions.

Technographics

What technology a company runs, inferred from job ads, public code, DNS records or page markup. Useful for products that replace or integrate with a named tool.

Psychographics

Attitudinal attributes: risk appetite, buying style, innovation posture. Popular in decks, hard to source reliably, rarely actionable at the account level.

TAM, SAM, SOM

Total addressable market, serviceable addressable market, serviceable obtainable market. Board vocabulary rather than prospecting vocabulary, but they bound how much of a signal feed can possibly be relevant to you.

Account-based marketing

Treating individual accounts as markets of one, coordinating marketing and sales against a named list. The main consumer of third-party intent data.

Territory

The slice of the market one rep owns, usually by geography, segment or vertical. Signals that cross territory boundaries create routing problems that are organisational rather than technical.

The classification vocabulary

Words a signal gets labelled with. These are the ones Openpulse uses; other tools use different sets, and a vendor that only offers a relevance percentage is not classifying at all.

High intent

Somebody explicitly asking for something you sell. The strongest label available when watching a market.

Evaluating

Comparing options, or describing the pain in the present tense. One step behind high intent and often a better conversation, because the decision is not made.

Switching

Leaving, replacing or cancelling a named competitor. The strongest label when watching rivals.

Frustrated

An active complaint about a rival with no stated intent to move. A lead in waiting rather than a lead.

Comparing

Weighing a rival against alternatives without being unhappy yet. The moment your name most needs to be in the conversation.

Praise

Somebody happy with your competitor. Not a lead, and worth keeping anyway: it is the objection list you will meet on every call.

Pain confirmed

A business's own reviews naming the problem your product solves. The strongest label when prospecting local businesses.

At risk

Reviews showing strain around a problem without naming it. Earlier and weaker than pain confirmed, and often the better time to call.

Noise

A candidate the classifier judged to be nothing at all. The largest category in any honest system.

Filtering and quality

Precision and recall

Precision is the share of what you kept that was correct. Recall is the share of what was correct that you kept. Every filter trades one against the other, and a vendor that mentions neither is not measuring either.

False positive

Something the filter kept that should have been discarded. The visible failure: it wastes a rep's time and they notice.

False negative

Something the filter discarded that should have been kept. The invisible failure, and the more expensive one, which is why rejections are worth storing. See the false positive taxonomy.

Gate

A cheap filter applied before an expensive one. Lexical overlap, date windows, exclusion lists and duplicate detection all cost effectively nothing and remove most candidates before any model is called.

Reranker

A model that reorders retrieved candidates by relevance to a query, run between cheap retrieval and expensive classification. It is what lets a system retrieve broadly and still classify cheaply.

Classifier

The model that assigns a label and a score to a surviving candidate. The expensive step, which is why everything above exists.

Rejection reason

The recorded explanation for why a candidate was discarded. Without it, a quiet listener is indistinguishable from a broken one.

Full text versus snippet

Whether the judgement was made on a complete post or on a search preview. A snippet contains the keyword by construction (that is why it was returned), so it always looks relevant, and the disqualifying context is in the part that was cut off. The single largest cause of wrong labels.

Data window

How far back a run looks. Conversations go stale in about 30 days; reviews of local businesses stay informative closer to 90, because the operational problem underneath moves more slowly than a thread does.

Motions and delivery

Signal-based selling

Prioritising outreach by observable evidence rather than by list order. The opposite of working an alphabetised export.

Allbound

Marketing coinage for using signals to blur inbound and outbound: outbound timing driven by inbound-quality evidence. Useful idea, overexposed word.

Outbound

Contacting people who have not contacted you. Signals change what you say and when, not whether it is outbound.

Warm outbound

Outbound where you have a specific, citable reason for the timing. What signal-based selling is trying to produce.

Enrichment

Appending data (contact details, firmographics, technographics) to a record you already have. Distinct from detection, and usually a separate vendor.

Identity resolution

Stitching fragments across systems into one person or account: a community handle, a CRM contact and a product account as one entity. A genuinely hard problem and a separate product category.

Webhook

An HTTP callback fired when something happens, so another system can act without polling. The delivery mechanism that turns a signal into a row in your CRM. Openpulse signs every one with HMAC-SHA256 and retries with backoff. See the integration guide.

MCP

The Model Context Protocol: a standard by which AI assistants connect to external tools and data. It is how an agent reads a signal workspace directly, rather than you copying results into a chat. Openpulse is an MCP server; see MCP for AI agents.

Roll-up

One row per entity (per competitor, per account, per location) aggregating the signals underneath it. Usually the thing a salesperson should act on, with the individual signals as the evidence beneath.