A precise description of the kind of company a business sells to best: industry, size, what they handle and at what scale, plus exclusions. In LeadAtomic the ICP drives the rule-based scoring that runs before any AI model. A vague ICP produces noisy results whatever the tooling. Sharpening it is usually the biggest improvement available, and it's the part no software can do for you.
The older industry name for what LeadAtomic calls account enrichment. Kept here as a cross-reference; see “account enrichment” for the definition and for how the timing differs from a database lookup.
The older industry name for what LeadAtomic calls fit scoring. Kept here as a cross-reference; see “fit scoring” for the definition, and note that a fit score is not a prediction that anybody will buy.
An AI system whose decisions the person relying on them can inspect. In B2B prospecting, that means every shortlist, discard or escalation carries the inputs and reasoning behind it. A score without an explanation is a number you have to take on trust. In LeadAtomic, every score carries its reasons, including the discards.
Rule-based scoring that runs before any AI model. Industry and segment fit, exclusions, keywords and structural junk checks throw out the clear misfits cheaply, and an AI review checks the rest. Running rules first keeps the model's time, and its cost, for companies that could plausibly fit. Running the model first costs more, is harder to explain and tends to amplify noise.
A long-lived signal attached to a sending domain and mailbox, which inbox providers use to decide where messages land. Bad sends compound: high bounce rates, spam complaints and low engagement drag down every future message, not just the ones that caused the problem. LeadAtomic's sending limits exist to protect it before damage is done, and a recovery mode exists for when damage happens anyway.
Whether a sent message reaches the recipient's inbox rather than spam or junk. Deliverability comes from sender reputation, authentication (SPF, DKIM, DMARC), content patterns and recipient engagement together. The fastest way to ruin it is to send low-signal messages to people who don't want them; the slowest way to recover it is to keep doing that.
A workspace's record of a company: who it is and what research found. Each workspace keeps its own copy of every company and its research, even when two customers find the same business. One customer's research is never shared with another. Fit decisions and pipeline state are kept per workspace too. Only two records cross workspaces: a list of companies that asked us to delete their data, so they are never researched again, and an anonymous count of common email formats that names no one.
Scoring a company on how well it matches a client's defined ICP, as distinct from generic firmographic scoring. ICP fit scoring is client-scoped (the same company scores differently for different clients) and is the first gate LeadAtomic applies before any research runs.
A company or person who has shown some sign of interest: filled in a form, downloaded something, replied, asked a question. The word implies a raised hand. The industry uses it loosely for anything in a list, which is where the confusion starts, because a name in a database has raised nothing. LeadAtomic doesn't produce leads. This site calls what we deliver prospects or accounts, though the word still turns up in parts of the app. This entry exists because the distinction matters to the product, and because anyone checking whether we know our subject will look here first.
A company that fits what you sell and has a reason you might contact it, but has shown no interest and may not know you exist. This is what LeadAtomic produces: a researched account with its fit documented, and any trigger the research found. Calling a prospect a lead claims evidence of interest that nobody outside the company has. A prospect told us we were doing exactly that, in August 2026, and they were right.
Something documented and dated that changed at a company recently and makes contact more sensible now than six months ago: a funding round, a hiring pattern, a new product in a regulated area, a leadership change. A trigger is observable; it's a fact with a source. What a trigger doesn't tell you is whether anyone is shopping. Fit without a trigger is a company that might buy someday; a trigger without fit is noise.
A loose industry term covering anything from a documented trigger event to inferred browsing behaviour. The looseness is the problem: it lets a vendor present an inference with the confidence of a fact. LeadAtomic splits the two. An observed event is marked as observed, an inference is marked as an inference, and what nobody can observe is marked as unobservable instead of quietly dropped.
Attaching context to a discovered company: contactability, a summary of what it does, signals (hiring, funding, news, sales stack), risk flags, and the evidence behind each. Also written as lead enrichment, the older industry term for the same work. LeadAtomic researches only companies that passed the cheap gates and that you release for research. The budget goes on accounts worth it, and weak or missing evidence stays visible instead of being smoothed over.
Assigning a score for how well a company matches what a customer sells to, so the pipeline can shortlist, discard or escalate. Also called lead scoring. Scoring in LeadAtomic is workspace-specific (the same company scores differently for different customers) and carries its reasons, including for the discards. Rules throw out clear misfits before any AI runs; an AI review checks the rest.
Behavioural signals, such as research activity, content consumption and technology adoption, sold as evidence that a company may be considering a purchase. It's usually attributed to a company by IP address or cookie, a probabilistic step presented as a certain one. LeadAtomic doesn't use or produce intent data. The product is built on fit and documented triggers, and what nobody can see from outside a company is labelled unobservable rather than estimated.