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 deterministic scoring before any model call. A vague ICP produces noisy results regardless of tooling; sharpening it is usually the highest-leverage improvement available, and it is 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 can be inspected by the person relying on them. In a lead-intelligence context, explainability means every shortlist, discard, or escalation carries the inputs and reasoning behind it. A score without an explanation is just a number you have to trust. LeadAtomic treats explainability as a precondition for shipping AI into the pipeline.
Rule-based scoring that runs before any model call. Industry and segment fit, exclusions, keywords, and structural junk checks handle the obvious cases cheaply. Deterministic-first is a discipline: the model is then escalation for ambiguity rather than the default engine. Reversing the order is expensive, opaque, and tends to amplify noise — in LeadAtomic’s development running, 91% of candidates were resolved by the cheap checks alone.
A long-lived signal attached to a sending domain and mailbox identity, used by inbox providers to decide where messages land. Bad sends compound — high bounce rates, spam complaints, and low engagement degrade reputation across all future messages, not just the ones that caused the problem. LeadAtomic treats sender reputation as an asset to protect by default, not a metric to recover after damage.
Whether a sent message actually reaches the recipient's inbox rather than spam or junk. Deliverability is the joint output of sender reputation, authentication (SPF, DKIM, DMARC), content patterns, and recipient engagement. The fastest way to ruin deliverability is to send low-signal messages to people who do not want them; the slowest way to recover it is to do that repeatedly.
The principle that a company has one objective identity (the canonical Company record) and many client-specific interpretations of that identity (ClientCompany records). LeadAtomic strictly separates the two: canonical data is shared and amortised across clients, while client decisions stay isolated. Blurring the boundary is treated as a data-integrity hazard.
Scoring a company specifically on how well it matches a client's defined ICP, as distinct from generic firmographic scoring. ICP fit scoring is client-scoped (it produces different scores for the same company across different clients) and is the first gate LeadAtomic applies before enrichment runs.
A company or person who has shown some signal of interest — filled a form, downloaded something, replied, asked a question. The word implies a hand raised. 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 does not produce leads and no longer uses the word for its own output. This entry exists because the distinction is the product, and because a reader checking whether we know what we are talking about 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 and its trigger documented. 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 is a fact with a source. What a trigger does not 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 the things nobody can observe are marked as unobservable rather than 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, which is the older industry term for the same work. LeadAtomic researches only after a company has passed the cheap gates, so the research budget is reserved for accounts worth the spend — and weak or missing evidence stays visible rather than 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. Cheap deterministic checks run first; a model resolves only the ambiguous middle.
Behavioural signals — research activity, content consumption, technology adoption — sold as evidence that a company may be actively considering a purchase. Usually attributed to a company by IP or cookie, which is a probabilistic step presented as a certain one. LeadAtomic does not buy or produce intent data. The product is built on fit and documented triggers, and what nobody can see from outside a company is labelled as unobservable rather than estimated.