30-second answer: An intent signal is an observable behaviour or content interaction that suggests an account is researching a category, vendor, or problem. The vocabulary covers source classes, signal grain, scoring, decay, merge, and false-positive control. This glossary defines 22 terms revenue operators encounter in intent-data platforms and orchestration tools.
See merged first- and third-party signals scoring an account live inside Abmatic AI, book a demo.
First-party signals come from properties an account interacts with directly: website visits, content downloads, demo requests, ad engagement on owned channels. They are deterministic and high precision. See first-party intent data.
Third-party signals come from external publisher networks observing research behaviour and reporting it under cooperation agreements. They are probabilistic, broader-coverage, lower precision per signal. See third-party intent data.
Predictive signals are model outputs combining first- and third-party data with historical conversion outcomes to forecast future intent. See predictive intent data.
Co-op signals come from networks of B2B vendors sharing observed buyer behaviour under explicit terms; G2 buyer-intent and TrustRadius downloads are common examples.
Bidstream signals come from real-time bidding data; coverage is broad but signal grain is coarse and identification quality varies by inference vendor.
Account-level signals are aggregated to the company; the originating contact may be unknown. Most third-party intent ships at account grain.
Contact-level signals identify the individual taking the action. First-party signals frequently land at contact grain when forms or known-cookie identification fires.
Topic signals tag the research subject (kubernetes, identity governance, lead routing). Topic taxonomies vary across vendors and require mapping for cross-vendor comparison.
A surge signal flags an unusually elevated rate of activity above a baseline. Surge detection adds context that flat signal counts cannot.
Trigger signals are discrete events of high importance: tech-stack additions, leadership hires, funding events. They warrant manual review when they fire.
The numeric multiplier applied to a signal type when summing into an intent score. Weights are tuned by historical signal-to-conversion correlation, not vendor recommendation.
A recency multiplier that reduces signal value as time passes since observation. Most categories use 14-day to 60-day half-lives.
Frequency weighting boosts repeated signals from the same account, capturing pattern signal that single-observation events miss.
The aggregate output combining multiple signal types into a single rank-orderable number. Composite scores feed routing and tiering. See how to set up account scoring.
Decay reduces the contribution of a signal as it ages. Without decay, dormant accounts retain inflated scores forever.
The time for a signal contribution to fall to half its initial weight. The right half-life depends on category buying-cycle length.
A cool-down prevents the same trigger signal from firing repeatedly across the same account in a short window, controlling outreach overload.
A reset event clears accumulated signal when a major lifecycle change occurs (deal closed-won, opportunity lost, contact churned).
The reconciliation of multiple signals against a unified account graph and contact graph. See signal merge and account graph.
The process of mapping observed identifiers (cookies, email, IP, fingerprint) to a canonical contact and account. See identity resolution.
Cross-domain stitching unifies signals across multiple owned domains and product properties for a single account view.
Bot filters remove non-human traffic before signal scoring; without them, scores inflate and noise drowns true intent.
ISP filters exclude residential and consumer-ISP traffic that maps to no valid B2B account.
Internal-IP filters exclude employee traffic from a vendor's own properties, avoiding self-engagement signals.
A window during which the same signal type from the same account counts only once, controlling for refresh storms and accidental double-counts.
Worked example: a B2B vendor in the security-tooling category combines first-party signals (pricing-page visits, demo requests) at 40 percent of composite weight, third-party co-op signals (Bombora surge on five mapped topics) at 30 percent, review-platform signals (G2 alternative-page views) at 20 percent, and trigger events (CISO hire, breach disclosure, funding) at 10 percent. Recency uses a 30-day half-life. The composite score routes accounts above 75 to BDR-direct outreach and accounts between 50 and 75 to nurture-with-alert.
Counter-example: the same program runs every signal at equal weight, with no recency decay, no bot filtering, and no identity resolution. The composite score has near-zero conversion correlation, sales loses trust within a quarter, and the program reverts to opinion-based prioritisation.
Operating tip: review the score-to-conversion correlation monthly. If correlation drops materially over two consecutive months, the model has drifted and weights should be re-tuned against fresh data. Calibration is continuous, not one-time.
Intent signals interact with adjacent disciplines: fit scoring supplies structural context, attribution ties signals to closed pipeline, identity resolution binds signals to the right account and contact.
The strongest programs treat intent as one input to a composite that includes fit and engagement, rather than a single answer to who to talk to.
Signal merge across first-party, third-party, predictive, and trigger sources is the operating activity that turns raw intent into an actionable score.
Vendors increasingly ship merge logic out of the box, but the calibration burden (weights, recency, decay) remains on the program. Merging first and third party intent and signal merge are the operating habits that make intent compound rather than fragment.
The most reliable intent-signal stacks follow a layered pattern. Layer one is identity: ensure every signal is bound to an account and where possible a contact through a unified graph. Layer two is normalisation: convert raw signals to a common schema (account, topic, source, weight, timestamp) so cross-vendor comparison and scoring are tractable. Layer three is scoring: apply weights, recency, and frequency multipliers to produce a composite. Layer four is action: route the score change into outbound, ad targeting, or sales alerts. Skipping any layer breaks the layers above. Common anti-patterns are scoring without identity resolution (signal sprays across phantom accounts), trusting vendor topic taxonomies without mapping (cross-vendor comparison becomes meaningless), and acting on raw signals without decay (dormant accounts trigger outreach storms). Avoiding these three patterns alone tends to materially improve intent-program reliability.
See merged first- and third-party signals scoring an account live inside Abmatic AI, book a demo.
First-party signals are deterministic and observed on properties you own. Third-party signals are probabilistic, observed across cooperating publishers, and reported at account grain. See how to merge first and third party intent.
Yes for routing decisions, no for diagnostic review. A composite score makes prioritisation tractable; underlying signal breakdowns let revenue teams understand why a score moved. See ABM metrics glossary.
Categories with multi-quarter cycles (enterprise platforms, healthcare IT) typically use 45 to 90 day half-lives. Faster cycles (developer tools, marketing tech) use 14 to 30 days. The right number is whatever produces the highest correlation between score and conversion in held-out historical data.
It depends on coverage and category. In categories where third-party intent has high vendor-topic match (cybersecurity, dev tools), third-party signals can carry 40 to 60 percent of composite weight. In bespoke or emerging categories, first-party signals dominate.
Quarterly is the standard cadence. Major product or pricing changes, ICP redefinitions, and channel mix shifts all justify off-cadence retuning.
Aggressive bot filtering typically removes 20 to 40 percent of raw signals in B2B; the remaining signal is much higher precision. The trade-off is worth it because routing rules built on noisy signal degrade fast.
Intent signals are the time-of-research dimension of B2B revenue programs. Used well, they shorten cycles, focus capacity, and surface accounts ready to talk. Used poorly, they generate noise that erodes sales trust. Use this glossary alongside the intent data glossary when evaluating vendors and designing scoring rules.
Ready to put this glossary into practice? Book a demo of Abmatic AI.