Most B2B intent data programs are built on topic taxonomy signals, not behavioral cluster intent. Topic taxonomy flags accounts when company IPs visit category-tagged publisher pages. It does not identify who was researching, at what seniority, or if the activity reflects vendor evaluation. Behavioral cluster intent tracks named contacts across multiple sessions at the same account. The difference explains why most intent programs produce signals but not pipeline.
The Intent Data Fraud: What You Are Buying Vs What You Think You Are Buying
ABM
Oct 14, 2026
The renewal came up in the Q3 budget review.
Twelve months, $84,000, and 1,400 accounts flagged as ‘high intent’. The SDR team worked through 600 of them, booked 11 meetings, and created 3 opportunities, with zero closed.
The CFO asked what ‘intent’ meant in this context.
The answer: accounts that appeared on topic taxonomy lists maintained by the intent vendor. A company shows up on a ‘marketing automation’ intent list when employees visit pages tagged to that category on partner publisher sites.
The CFO asked what the false positive rate was.
Nobody knew. The vendor did not provide that data.
The contract was not renewed.
During the renewal review: the methodology they had been paying for across 12 months was a page view counter with a topic label on it. An employee at a 3,000-person company reads a marketing blog post on their lunch break. The company appears on the intent list. The SDR calls. The company was never in evaluation.
87% of intent signals never reach qualified pipeline (DemandScience, 2026). This is not bad luck. It is the predictable outcome of buying topic taxonomy and calling it behavioral intent.
‘The Intent Data Fraud’ is not a single vendor problem. It is a category-level problem: the language of intent data marketing obscures a three-tier quality hierarchy that most buyers never learn to see.
The Three-tier Reality of Intent Data
Inside the intent data market are three fundamentally different things, all sold under the same label:
Tier 1 - Topic Taxonomy Matches: The most common type. The lowest quality. A data aggregator tracks page visits across a publisher network. When an account’s IP address appears on pages tagged to a topic, the account is flagged as “showing intent.” 50% of companies using third-party intent data report too many false positives (Forrester via Prospeo, 2026). At a company with 1,000 employees, someone reads about every topic every week. The signal is indistinguishable from background noise.
Tier 2 - Content Consumption Tracking: Better, but still limited. Vendors track content downloads and asset consumption across syndication networks. An account that downloads a whitepaper in a category is scored as in-market. One contact, one event, no pattern. A single-contact content download in a 20-person buying committee does not indicate the account is in evaluation. It indicates one person found the content relevant enough to download.
Tier 3 - Behavioral Cluster Signals: The rarest type. The only type that earns the word “intent.” Multiple contacts at the same account, consuming related content, across multiple sessions, within a bounded time window, with job titles correlated to the purchase decision. This is what buying committee research actually looks like. This is the signal worth activating on.
Most teams have never bought Tier 3. They have been told they are buying it. They have been buying Tier 1.
How Vendors Obscure the Difference
The language of intent data marketing is designed to prevent this distinction from being made.
Surge signals: An account is “surging” on a topic when its engagement rises above baseline. What this describes: an account that viewed 2 topic-related pages last month viewed 6 this month. A 3x surge. This sounds significant. It describes three additional employees browsing technology news.
Coverage claims: “100 million intent signals per month.” Volume obscures quality. 100 million topic taxonomy matches are worth less than 100,000 behavioral cluster signals. The number has no meaning without knowing what percentage of signals represent genuine multi-contact behavioral clustering.
Exclusivity language: “First-mover advantage,” “be first to the signal.” Most intent data is sold to multiple buyers. The account flagged as showing intent for marketing automation received outreach from four vendors the same week. None of them were first.
Confidence scores: A 92% confidence score on a topic taxonomy match means the algorithm is 92% confident the account viewed those pages. It says nothing about active evaluation. Confidence in signal measurement is not confidence in purchase intent.
Intent vendors should be able to provide transparent sourcing documentation when asked directly. The inability to answer the question “What exactly did we measure?” says it all.
The Cost of Too Many False Positives
Take an illustrative 65% false positive rate: 650 of every 1,000 intent-flagged accounts are not in active evaluation.
SDR time: An SDR working 40 intent-flagged accounts per week, at the same 65% rate and about an hour per account, spends 26 hours reaching out to accounts that are not buying.
Account damage: The 650 false positive accounts received outreach when they were not in evaluation. Cold outreach at the wrong moment with vendor-specific messaging is relationship-negative. Those accounts are now harder to reach when they do enter evaluation.
Attribution corruption: Programs using intent as a pipeline attribution signal attribute pipeline to accounts that would have converted regardless. Attribution models built on false positive intent produce misleading strategy.
Opportunity cost: The accounts actually in evaluation, not flagged by topic taxonomy because their employees happened not to visit the right publisher pages, received no outreach. They found another vendor.
What Genuine Intent Requires
The Behavioral Cluster Intent model is Machintel’s framework for identifying accounts in active evaluation from multi-signal, multi-contact behavioral data. See how it works.
Five criteria define a behavioral signal worth activating on:
Multiple contacts at the same account: A single page view or download is not an account signal. Two or more contacts at the same account showing related research behavior begins to be meaningful.
Correlated research across multiple sessions: Research spread across multiple days in a time window is evidence of sustained inquiry, not a one-time browse.
Job title relevance: Research activity from contacts in roles relevant to the purchase decision: demand gen, marketing ops, CMO, is more meaningful than activity from employees in unrelated functions.
Time-bounded recency: Signals from the last 7 days are materially more valuable than signals from 21 days ago. Active evaluation windows are short. Intent data without recency weighting treats a month-old signal the same as a signal from yesterday.
Account-level pattern coherence: The individual signals should cohere: the same account, researching the same category, across multiple contacts, in a compressed time window. That coherence is the behavioral cluster.
Knowing what real intent looks like, but having no clear way to find it, is a frustrating spot for any demand gen team. Machintel’s demand generation team works with B2B marketers on that problem, focusing on accounts where several contacts are researching the same category, not on single page views. Your SDRs then spend their time on accounts with real evidence behind them.
The Vendor Qualification Test
Before renewing or purchasing intent data, four questions reveal which tier you are buying:
- What exactly is the source of each signal?
- What is the false positive rate on your signals?
- What is the recency window on signals?
- Can you provide sourcing documentation for a sample of flagged accounts?
The Evaluation Window Problem
There is a timing problem that compounds the quality problem.
B2B purchase evaluations are not months-long processes spread evenly across a calendar. They are front-loaded: most shortlisting happens in the first 30-45 days, and most vendor decisions finalize within 90 days of the first identified signal.
Topic taxonomy intent signals are typically reported on a weekly or bi-weekly cycle. An account that enters active evaluation on a Monday appears on the Friday spreadsheet. SDR outreach happens the following week, ten or more days after the signal first appeared.
At that lag, 92% of buyers already have at least one vendor in mind (Forrester via Firebrand Marketing, 2026). The account is in conversation with vendors who were already there. The intent data that was supposed to surface the opportunity is surfacing it after the window to influence the shortlist has begun to close.
Behavioral cluster signals, when sourced from real-time or near-real-time data infrastructure, can surface within 24-48 hours of the pattern emerging. That difference in latency is not a minor technical detail. It is the difference between entering a conversation before the shortlist forms and entering it after.
The intent vendor question that rarely gets asked: what is your signal latency? How many days between behavioral activity and delivery to my CRM?
A vendor selling weekly batch delivery of topic taxonomy signals is delivering information that is 7-14 days old, about activity that may already be weeks in the past. By the time an SDR picks up the phone, the prospect may have already had three conversations with competitors.
Speed to signal matters as much as signal quality. Both have to be evaluated.
FAQs
What is the difference between topic taxonomy intent and behavioral cluster intent?
Topic taxonomy intent flags an account when company IPs visit pages tagged to a category. Behavioral cluster intent tracks multiple named contacts at the same account consuming related content across sessions within a time window. The first is a page view counter. The second is a buying signal.
How can a B2B marketer identify which type of intent data they are purchasing?
Ask the vendor three questions: what percentage of signals are contact-level versus IP-level, what is the multi-contact threshold required before an account is flagged, and what is the signal latency between behavioral activity and delivery. A vendor selling behavioral cluster intent can answer all three precisely.
Why do false positive rates matter in intent data programs?
False positives consume SDR capacity without producing meetings. At illustrative false positive rates of 60-70%, a team working 100 intent signals per week spends 60-70 hours, at an hour per signal, on accounts that were never in evaluation. That is the direct cost, and the indirect cost is the pipeline that was not built with those hours.
