Intent data integration — where the signal is, and where it isn’t.

The premise — intent data is a complement, not a replacement

The operational question is never “should the list incorporate intent data.” The question is which intent signals are real signal versus marketed noise, which providers produce data quality that justifies the per-month cost, and at what point in the pipeline the intent layer is applied. An operator who buys intent data before defining their ICP is buying a categorically different product than the one being sold — a generated lead list, not an in-market signal — and the conversion rate reflects that.

The empirical baseline observed across multi-quarter outbound programs: a well-constructed ICP-only list converts at a certain rate; the same ICP filtered by a credible intent signal converts at 1.5 to 3x that rate; the same intent signal applied to accounts that do not match the ICP converts at roughly half the ICP-only rate. Intent data is a multiplier on a well-defined ICP. Applied to a poorly defined one, it is a directional distractor.

The seven intent categories

The category is sold as a single product. It is not. The third-party intent landscape consists of seven distinct data products, each with its own signal-to-noise profile, refresh cadence, per-vendor accuracy ceiling, and operational use case:

• Search intent.Aggregated signal from B2B research sites, content syndication networks, and bidstream data, attributed to a target account via IP-to-company resolution. Marketed as “account X is researching topic Y this week.”
• Content-consumption intent. Signal from gated B2B content networks — when accounts download whitepapers, watch webinars, or read syndicated articles tagged to a topic taxonomy.
• Technographic data. The detected stack at the target account — which CRM, which marketing automation, which security tooling, which cloud provider — derived from web scraping, job-listing analysis, and DNS fingerprinting.
• Hiring signal. Active job postings parsed to infer that the company is building a specific capability, expanding a specific function, or shifting a specific budget category.
• Funding signal. Disclosed funding events — venture rounds, debt facilities, secondary financings, public-market raises — sourced from public filings and press releases.
• Executive-move signal. Tracked changes in senior personnel — new VPs, new C-levels, departed champions — sourced from professional-network scraping and press release monitoring.
• Regulatory-filing signal. SEC filings, FDA submissions, FCC applications, customs records, patent filings — public-record streams that surface operational changes the company has not yet press-released.

Per-category signal quality and noise

The categories do not have remotely comparable signal-to-noise ratios. The marketing material from intent-data providers treats them as a single offering. The operator needs to evaluate them individually.

Search intent — high noise, topic-level false positives

Search intent is the most aggressively marketed category and the noisiest in practice. The attribution chain is long: a query from an anonymous IP gets resolved to a company via an IP-mapping database, gets attributed to a topic via keyword matching, and gets surfaced as “company X is researching topic Y.” Every step introduces error. The IP-to-company mapping is correct at roughly 60 to 75% at the SMB tier and 85 to 95% at the enterprise tier. The topic attribution is keyword-matched against a coarse taxonomy that conflates adjacent topics. The signal is frequently correct that something happened at this account and frequently wrong about what.

The category is most useful when filtered aggressively — top 5 to 10% of intent score, narrow topic basket, multi-week sustained signal rather than single-week spikes — and least useful as the sole driver of an account list.

Content-consumption intent — better signal, provider-coverage dependent

Content-consumption signal is direct: the company filled out a form, the company watched the webinar, the company downloaded the whitepaper. The attribution chain is one step long and the signal is unambiguous. The constraint is provider coverage — the signal only exists for the portion of research that happens inside the provider’s content network. A target ICP that researches outside the provider’s syndication footprint produces no signal, which the operator misreads as “this account is not in-market” rather than “this provider has no visibility into this account.”

Technographic — high accuracy, low velocity

Technographic data is the most accurate intent-adjacent category. Stack detection from scraped HTML, JavaScript snippets, DNS records, and job-listing text reaches 90% plus accuracy at the major-stack-component level (CRM, marketing automation, ecommerce platform, primary cloud). The signal is a state, not an event. The velocity is correspondingly low: companies change CRMs on multi-year cycles. Technographic is a powerful filter, a weak trigger.

Hiring signal — high signal, structurally lagging

A company posting a job for “Head of Revenue Operations” is a high-signal event for any vendor selling into the RevOps function. The role does not exist; the company has decided it should; the company is now hiring. The structural constraint is that the signal lags the decision by weeks — by the time the role is publicly posted, the budget cycle that authorized it concluded the prior month. Useful for “they are building this capability and will need tooling for it,” less useful for “they are evaluating tooling this week.”

Funding signal — high signal, commoditized

A company that closed a Series B in the prior 90 days has a step-function increase in budget availability, hiring velocity, and category-expansion willingness. The constraint is that every other vendor receives the same signal at the same time, which collapses the per-prospect reply rate on funding-triggered outbound. The operational lift comes from speed (first-week contact) and from message-specificity (referencing the specific use of funds in the announcement) rather than from the signal itself.

Executive moves — highest signal-to-noise for relationship-driven outbound

A new VP of Engineering arriving at a target account is, for the operator selling engineering tooling, the highest signal-to-noise event available from any third-party stream. The new executive has 90 to 120 days of accumulated political capital to evaluate and replace incumbent tooling, has not yet locked into the prior administration’s vendor relationships, and is operationally incentivized to make visible decisions. The closest the third-party intent space gets to a structurally underpriced asset.

Regulatory filings — domain-specific, asymmetric

Regulatory-filing signal is worth almost nothing for the generic vendor and almost everything for the operator whose product maps directly to a filing event — a compliance vendor watching SEC enforcement actions, a customs-broker vendor watching new import records, a clinical-research vendor watching FDA submissions. Binary in its applicability.

Per-provider pricing patterns

Intent-data pricing clusters into four tiers, with the per-tier-vendor count growing as the price compresses:

The pricing is not correlated with signal quality in any monotonic way. The $25,000/mo platform’s search-intent feed has the same underlying bidstream input as the $4,000/mo specialist’s; the premium is paid for the scoring layer, the CRM integration, and the account-executive relationship. The operational ROI question is not “is the data 5x better,” because it usually is not.

The per-cost-tier ROI calculation

A defensible ROI calculation requires three inputs: the baseline cold-outbound reply rate against an ICP-only list, the empirical lift on layered intent plus ICP for the specific category, and the per-month all-in cost. A 2,000-prospect list converting at a 4% reply rate produces 80 replies per month. The same list narrowed to the 600 prospects with credible intent overlay, converting at 9% (a 2.25x lift, within the 1.5-3x empirical band), produces 54 replies. The relevant comparison is not reply count but cost-per-meeting-booked. A $5,000/mo intent feed needs to produce roughly 50 incremental qualified meetings per quarter (versus the ICP-only baseline) to justify the spend at typical seat economics. Most do not. Some do.

The “intent data as ICP” failure

The most common operational failure in the category — and the failure that the intent-data sales motion is structurally aligned to produce — is the substitution of intent signal for ICP definition. The operator buys an intent platform, receives a weekly list of “in-market accounts,” and treats that list as the target universe. The ICP definition (Chapters 01-02) was never built; the intent signal is being asked to do work it cannot do. The empirical consequence is observable in reply data: the unfiltered intent list spans revenue ranges from $1M to $10B and converts at meaningfully below the rate of an intent-filtered ICP list, with meeting-to-opportunity conversion worse by a further factor. Intent signal absent ICP context is closer to a directory than to a sales-qualified-account stream.

Data-freshness decay

Every intent category has a half-life. The operational implication of data freshness is not “the data is wrong” — it is “the signal that triggered this account’s inclusion in the list happened N days ago, and the buying-window probability decays with N.” Per-category half-lives observed in practice:

• Search intent: 7-14 days. Provider refresh cadence is typically weekly.
• Content-consumption: 14-30 days. Download-to-evaluation window exceeds search-to-evaluation.
• Technographic: Multi-quarter. The signal is a state, not an event.
• Hiring: 30-90 days. The role gets filled, the budget gets allocated, the buying motion concludes.
• Funding: 60-120 days. Capital-deployment cycle takes a quarter to materialize.
• Executive moves: 90-180 days. Political-capital window is one to two quarters.
• Regulatory filings: Highly category-dependent.

The operational pattern is to tag every intent-derived prospect with the signal date and enforce a per-category maximum age before re-qualification. A 21-day-old search-intent prospect is a different prospect than yesterday’s, and the campaign architecture (Chapter 05) should reflect that.

The multi-source integration pattern

Single-provider intent dependency is the second-most-common operational failure in the category. Every provider has coverage gaps — segments where they have no panel data, geographies where their content network is absent, industries where their taxonomy is shallow. The operator who routes 100% of intent input through one vendor is routing 100% of their account targeting through that vendor’s blind spots.

The defensible architecture combines intent inputs from at minimum three sources: a paid intent provider for the primary commercial signal, a public-derivative source for funding and executive moves (near-zero marginal cost), and the operator’s own first-party signal stream (Chapter 02). The first-party stream is the highest-quality signal in the integrated graph and usually receives the least weight in the operator’s mental model of where intent comes from. The closed-won-derived signal (Chapter 01) is the calibration layer that tells the operator which sources are predictive of closed revenue versus merely predictive of meetings booked. An intent signal that produces 3x the reply rate but 0.5x the close rate is one the operator should be aggressively weighting down; invisible without the closed-won feedback loop.

The “intent signal but no ICP fit” segment

The segment with the highest marketed value in the intent-data sales motion is also the segment with the highest false-positive rate in practice: accounts showing strong intent signal that do not match the ICP. The provider surfaces them as “in-market opportunities you would miss.” The operational experience is that they reply at meeting-booking rates roughly equivalent to the ICP-fit segment and convert to closed revenue at meaningfully lower rates, frequently churn within the first year, and consume sales-cycle hours disproportionate to their contracted ACV. The defensible posture is to surface the segment as a distinct cohort, run it through a stricter qualification path, and explicitly reject the framing that intent signal alone qualifies an account.

Operational integration at the prospect-graph layer

Intent data integrates into the architecture defined in Chapter 04 as an additional signal layer on the prospect graph — not as a parallel list, not as a separate workflow, and not as a campaign trigger in isolation. Each intent input attaches to an account node as a typed signal with three required attributes: source (which provider), timestamp (when the signal was observed), and decay-class (which half-life applies). The downstream consumers — segmentation (Chapter 05) and campaign architecture — query against the signal layer rather than receiving pre-filtered lists from the intent provider. The architectural inversion is consequential: the provider stops being the source of the list and starts being a source of attributes on a list constructed from the operator’s own ICP and first-party signal.

Common operator failures observed in production

• Buying intent before defining the ICP. The provider is asked to do work the closed-won-deconstruction exercise (Chapter 01) was supposed to do. Downstream campaigns convert at the rate of a poorly defined ICP, not at the rate the intent signal would have produced against a well-defined one.
• Single-provider over-reliance.The vendor’s coverage gaps become the operator’s targeting blind spots. The vendor’s pricing leverage at renewal becomes maximal.
• No signal-decay enforcement. A 90-day-old search-intent spike is treated as the same input as a 7-day-old one. Reply data degrades and the cause is invisible.
• Treating intent as ICP. The intent-derived list is the campaign target universe, full stop. The downstream ACV and close-rate dilution is misattributed to copy and cadence rather than to the targeting layer.
• Paying for high-cost intent when budget should fund first-party capture. A $15,000/mo suite is deployed on a product with no event-tracking instrumentation, no inbound-form scoring, and no closed-won attribution. The same spend, applied to engineering hours on first-party signal capture (Chapter 02), produces structurally better data within one quarter.
• No closed-won feedback loop. Intent ROI is measured on reply and meeting rates, never on closed revenue. Signals that produce meetings but not deals receive equal weight to signals that produce both.

Pre-purchase checklist

• A documented ICP definition derived from closed-won deconstruction (Chapter 01), with sample size and confidence thresholds documented
• First-party signal capture (Chapter 02) operational — the primary signal layer the intent provider will be evaluated against
• The specific intent category has a defensible mapping to a known buying-cycle event in the operator’s sales motion
• Provider coverage verified in the target geographies, industries, and account-size band against a known list, not the provider’s marketing material
• The integration path to the prospect graph (Chapter 04) engineered, with signal-typing and decay-class attributes specified before contract signature
• The closed-won feedback loop from CRM back to the signal layer operational, so per-source revenue attribution can be calculated within one quarter
• A documented exit path — export format, termination terms, dependency-removal plan — before contract signature
• First 90 days include an explicit A/B against an ICP-only baseline, with a documented kill criterion if lift falls below 1.5x

Where intent data fits in the list-construction stack

Intent data is the third signal layer in a well-constructed prospect graph, after ICP definition (Chapters 01-03) and first-party signal capture (Chapter 02), and before enrichment (Chapter 07). The ordering is not incidental. An ICP layered with first-party signal already produces meaningfully better outbound conversion than a flat firmographic list; intent layered on top of both produces incremental lift that compounds. Intent applied to no foundation produces a list that looks expensive and converts cheaply. The next chapter (07) addresses enrichment — the data-quality layer that determines whether the constructed list is operationally usable at scale.