Intent Data Didn't Break. The Wrong Signals Were Sold as Intent.

The prevailing story about intent data right now is that the category is tired. Reply rates are down, procurement is sharper about renewals, and every quarter another RevOps leader posts a version of "we cancelled our intent tool and nothing broke." The story is almost right. What actually failed was not intent data as a concept. It was the practice of treating a keyword search, a content download, a title change or a page visit as evidence that an account is ready to buy. Those things are context. They were sold as evidence. That is the category error the market is now correcting.

This piece is not another "intent data is dead" argument. It is narrower and more useful. Once you separate context from evidence, the pattern reads differently. Teams do not need less signal. They need to stop treating weak proxies as buying intent, and start acting on the observable events that show a buying motion is under way.

The category error at the heart of intent data

The original promise of intent data was straightforward. Firmographic targeting and a seller's own engagement data missed accounts that were in-market but not yet engaged. Third-party intent - bidstream activity, content syndication, keyword surges - was meant to close the gap. Reasonable in principle. What broke was the definition of the word "intent".

Somewhere between the deck and the dashboard, "intent" stopped meaning "evidence of a buying motion" and started meaning "any behaviour the vendor could measure". A page visit is intent. A whitepaper download is intent. A keyword surge on the aggregator's own network is intent. A job change at a target account is intent. Once the category is that loose, the score is measuring interest in a topic, not a buying decision at an account. The two are related. They are not the same thing.

This is the practical version of the point Nick Bennett made on Episode 1 of Intent, Decoded: intent data on a cold account with no ICP validation and no prior engagement is a reason to send a bad email faster. The platform did not fail. The definition did.

Context is not evidence

The simplest way to run a sanity check on any signal is to ask what it proves. Context tells you a topic is on someone's mind. Evidence tells you an account is actually doing something that changes its buying position. The two are useful in different ways, and they belong in different parts of the workflow.

Take four common signals as examples.

• A keyword search or a category page visit. Context. Someone at the account is curious about a topic. It does not tell you whether a decision is under way, who is driving it, or when a purchase might land.
• A content download. Context. It shows the topic is being researched. It does not tell you whether the download was for a live project, a personal learning goal, or a comparison exercise that will not close for eighteen months.
• A title change at a target account. Context. A new CRO or a new Head of Support often triggers procurement reviews. It does not, on its own, mean this one will.
• A competitor trial starting inside a customer of yours, or a churned trial with a named competitor. Evidence. A buying motion has left a footprint that can be observed and dated.

The category error is treating the first three the same way as the fourth. That is what most intent products encouraged, and it is why reps stopped trusting the score. As Leslie Venetz put it on Episode 2 of Intent, Decoded, a signal creates curiosity, not permission - and clicking a webinar link seven times is not a licence to send a demo invite. Curiosity is context. Permission comes from something stronger.

What the false positives cost the pipeline

The practical failure mode is well documented inside sales organisations, even when it does not make it into vendor case studies. When most of what a score flags is context rather than evidence, three things happen in sequence.

Reps chase false positives. The first month of a new intent layer usually produces a small lift, then flattens. Reps run down the "high intent" list, most of it produces no reply, and the working assumption inside the team becomes that the list is only marginally better than a well-built ICP list. Forrester's guidance on evaluating intent data providers flags false positives as one of the main structural risks.

Trust in the score drops. Once a rep has worked a dozen "hot" accounts that turned out to be cold, the score stops driving behaviour. It stays in the CRM. Nobody sorts on it. We wrote about this pattern in more depth in why CRM data is inaccurate: the signal field survives longer than the trust behind it.

The team stops acting on the data. After a couple of quarters, the layer becomes shelfware in everything but name. Marketing still reports on it. Sales quietly reverts to territory lists, referrals, and hand-raisers. The lift the score was supposed to unlock never compounds. That is not intent data failing; it is a definition problem that shipped into a workflow.

The measurable consequence is a pipeline where the highest-scoring accounts have similar conversion rates to unscored ones. The score sorts on something the buyer's actual behaviour does not reward. This is downstream of a wider pattern in B2B: on average only about 5% of any category's buyers are in-market at any moment, per Professor John Dawes's Ehrenberg-Bass 95-5 rule research. Scoring on inference and treating the top decile as buying-ready inevitably surfaces a majority of accounts that are still 12 to 24 months from a decision.

What an observable signal actually looks like

An observable signal is one that both parties could, in principle, describe the same way. A competitor trial has a start date. A renewal window has an end date. A vendor churn has a switch date. A technology change - an app installed, a stack swapped out - has a footprint you can point at.

Five signals do the heavy lifting for most GTM teams once context and evidence are separated.

• Installs. An account has just deployed a specific tool. It marks the entry into a buying motion for adjacent products and the start of the clock for switching costs.
• Active trials. An account is currently trialling a named vendor. This is a live evaluation, dated and observable. It has a natural window (usually 14 to 90 days) and a clear intervention point.
• Churned trials. An account trialled a named vendor and dropped the trial. The problem is still there. The chosen solution is not. Winning back interest is materially easier than convincing a happy customer to switch.
• Renewal windows. A customer of a competitor is 30, 60, or 90 days out from contract renewal. The window in which switching costs are lowest is short and predictable.
• Technology movement. An adjacent stack change (a new CRM, a new CDP, a new CS platform) that reshapes the buying committee, the integrations required, and the timing of a follow-on decision.

None of these signals require a rep to infer motive. Each one has a date on it, a named vendor attached, and a natural next action. This is the layer Subscription Intelligence was built to produce, and it is the reason evidence-based signals sort differently from inferred ones. The same signal set is the temporal component of the Qualified Opportunity framework - ICP Fit × Purchase Intent × Win Likelihood - and the broader taxonomy is covered in the purchase intent platform directory.

From more intent to who, why, and when

The right question is not "which accounts are showing intent". It is three separate questions the CRM does not answer on its own.

Who to prioritise. Firmographic and demographic fit still comes first. A signal on an account that does not fit is a distraction. Fit filters the universe; the signal decides which of the fit-qualified accounts is worth acting on this week. On Episode 2, Leslie was explicit about the order of operations: ICP filters first, signals on top. Reversing that produces a bigger list, not a better one.

Why now. A dated event tells the rep what changed and what to open with. "You started a Zendesk trial nine days ago" is a specific opening line that a buyer recognises. "Our platform tells us you're showing intent in the customer support space" is not. The named event is what lets the outreach earn its place in the inbox.

What action is appropriate. The same signal justifies a different move depending on the buyer's awareness stage. A vendor-aware buyer visiting your pricing page is ready for a qualified conversation. A problem-unaware account triggering a keyword signal is ready for an insight, not a demo request. Signal-led prioritisation only works if the action matches what the buyer is actually doing.

Niall's framing on the show is worth borrowing directly: MarketSizer's job is to help teams work out who to focus on, why now, and what to say. Not to add another intent score to the dashboard.

The future is evidence-led prioritisation

The corrective is not more intent. It is fewer signals, sorted on evidence, matched to buyer stage. Two moves change the shape of the workflow.

Move intent above the funnel, evidence to the trigger point. Keep the intent layer if it is useful for audience segmentation, territory planning, or paid media targeting. Stop asking it to decide which account a rep dials on Monday morning. The trigger for outreach should be an observable event: trial started, renewal opened, vendor churned, install detected. The intent score becomes a filter above the trigger, not the trigger itself.

Standardise on one repeatable use case first. The teams that get value fastest do not turn on every signal at once. They pick one - usually customers due for renewal in the next 90 to 120 days who are also trialling a named competitor - and build the play around it before adding the next. This is the point Niall made on Episode 2: sophisticated GTM teams routinely skip the fundamentals. Fixing the fundamentals is usually where the quick wins live. The deep-dive on that specific play, for churned accounts on the winback side, is in the B2B winback pillar; the inbound equivalent - using the same account-level signals to sort form fills - is covered in how to prioritise inbound leads with purchase intent.

Once the workflow anchors on observable evidence, the argument about intent data quietly resolves itself. Context is still useful upstream. Evidence is what triggers action. The teams that separate the two get to keep the good parts of the category, without paying for the noise.

Frequently Asked Questions

Is intent data still worth paying for?
Yes, if it is doing an audience or segmentation job rather than a prioritisation job. Intent layers can be useful for narrowing a large universe of accounts down to a research priority, or for informing paid media audiences. They are not the right layer to decide which account a rep should reach out to this week. Evidence-based signals do that job better.

What is the difference between context and evidence in a buying signal?
Context is a data point that suggests a topic is on someone's mind - a page visit, a content download, a keyword surge, a title change. Evidence is a dated observation of a specific buying motion at a specific account - a competitor trial started, a renewal window opened, a vendor churned. Both are useful. They belong in different parts of the workflow.

Are keyword searches and page visits useless?
No. They are useful as inputs to segmentation and audience decisions, and as one input among several when a rep is deciding what to research. They should not, on their own, trigger outreach. A page visit tells you a topic is being researched. It does not tell you an account is ready to buy.

What counts as an observable buying signal?
Any signal that both parties could describe the same way and that has a date attached. Competitor trial starts, active trials, churned trials, renewal windows, vendor churns, technology installs, and stack changes are all observable. They can be timestamped, attributed to a named vendor, and connected to a natural next action.

How do we start using evidence-based signals without ripping out our current stack?
Pick one repeatable use case first. A common starting point is a play that identifies customers of a specific competitor whose contracts renew in the next 90 to 120 days and who are also trialling one of your named alternatives. Build the workflow, prove the lift, then add the next signal. Turning every signal on at once is what makes the layer feel noisy again.

How is this different from Episode 2 of the podcast?
Episode 2 makes the case that a signal is not permission and that sellers have to earn the right to act on it. This piece runs the same argument one level up: a lot of what got sold as "intent" was context rather than evidence in the first place, and the market is now sorting the two out. The Episode 2 recap covers the seller's side of the same problem.

What should replace an intent-only sales prioritisation motion?
An evidence-led motion where a dated buying event triggers outreach, an ICP filter decides whether the account is worth the effort, and the message matches the buyer's awareness stage. The evidence layer is what Subscription Intelligence produces. The mechanics of running the workflow on top of it are covered in how to build a signal-led GTM motion.

Recommended reading

• From Signal to Trust: How Sellers Earn the Right to Reach Out - the Episode 2 recap with Leslie Venetz on the seller's side of the same argument.
• The Winnable Account Is Warm, Not Hot - what evidence-led prioritisation looks like when you sort on winnability rather than heat.
• What Is a Qualified Opportunity? - the formal three-part definition (ICP Fit × Purchase Intent × Win Likelihood) that the evidence layer plugs into.
• B2B Winback: Recover Churned Customers on Competitor Renewal Timing - the deep dive on the renewal + competitor-trial play, applied to the churned base.
• How to Prioritise Inbound Leads with Purchase Intent - the same three account-level signals applied to inbound rather than outbound.
• Purchase Intent vs Intent Data - the definitional pillar behind the context-versus-evidence argument.
• Why Intent Data Made B2B Outreach Worse - the effect of the category error on outreach quality, and how to unwind it.
• What Is Subscription Intelligence? - the data category that produces the evidence-based signal layer.
• Intent, Decoded - the podcast hub - both episodes and the wider conversation about how to use signals well.