What Changed in B2B Buying and Why the Old Model Broke
Written by Robin Caller, Leadscale CEO on March 6, 2026
What Actually Changed in B2B Buying?
Ninety-five percent of B2B buyers purchase from the vendor shortlist they built before making a single enquiry. That shortlist was assembled anonymously, through research you couldn’t see, in channels your CRM doesn’t touch. By the time a prospect fills in your contact form, the decision is already 80% made. The contact form is not the beginning of the buying journey. It is very nearly the end of it.
This is not a measurement problem. It is a structural problem. The demand generation model most B2B technology companies are running was designed for a buyer who entered the funnel through visible channels, engaged individually, and followed a roughly linear path from awareness to purchase. That buyer doesn’t exist any more — and hasn’t for several years. What replaces them is a buying process built around group decisions, anonymous research, and AI-mediated discovery. The transition this forces on demand generation is one Robin Caller, Leadscale’s CEO, describes as moving from “Campaign Thinking” to “Infrastructure Thinking.” Understanding why that shift is necessary starts with five changes in how B2B buyers actually behave.
Five structural shifts happened simultaneously. Each one, individually, would stress the traditional model. Together, they broke it.
Buying groups expanded beyond recognition
The average B2B buying group for deals averaging $250,000 now involves 10.1 members (6sense 2025 Buyer Experience Report). Forrester’s 2026 State of Business Buying puts the number higher still — 13 internal stakeholders and 9 external ones on a typical enterprise purchase. Procurement is involved from day one on 53% of deals, not waiting at contract stage as it once did.
The demand generation model built for an individual buyer — one form fill, one nurture sequence, one SQL — cannot account for this. A single enthusiastic researcher downloading three whitepapers generates MQL signals; the CFO who controls the budget and the procurement lead who will block the deal if the paperwork isn’t right generate none.
Buyers arrive with preferences already formed
Eighty percent of B2B buyers enter the vendor engagement phase with a pre-evaluation favourite (6sense 2025). Ninety-five percent ultimately buy from the shortlist they assembled before contacting anyone. The average buyer has completed 8.6 prior category purchase journeys — they are not starting from scratch, they are pattern-matching against previous experience.
This fundamentally reframes the purpose of demand generation. The question is not how to capture buyers who enter your funnel. It is how to get onto a shortlist that is being built in places you cannot see, by people who have no intention of talking to you until they are ready.
The buying timeline shifted
The average B2B buying cycle now runs 10.1 months (6sense 2025). First vendor contact happens at months seven or eight. Buyers control 61% of their journey before any vendor interaction occurs. The early-engagement window that traditional demand generation was built around — catch them early, nurture them through — has effectively closed.
This is not because buyers are more difficult. It is because they have better tools for doing their own research. They do not need you in the process until they want you.
Rep-free preference — with a paradox
Seventy-five percent of B2B buyers prefer a rep-free purchasing experience (Gartner, 2022 B2B Buyer Survey). That statistic is well known. The one that follows it usually isn’t: pure self-service produces 1.65 times more purchase regret than guided buying (Gartner, 2022 B2B Buyer Survey). Buyers want control. They do not want isolation.
The demand generation system has to be present without being intrusive — available when the buyer is ready, invisible when they are not. That is a significantly harder design problem than “send the right email at the right time.”
AI entered the research stack
Ninety-four percent of B2B buyers now use large language models during the purchasing process (Forrester 2025 B2B Buying Study, reported via BusinessWire, 10 October 2025). Sixty percent use ChatGPT or Gemini specifically to augment their vendor shortlists (Google / Ipsos, “How B2B Buyers Use Generative AI in the Purchase Process,” October 2025, reported via Think with Google). Every one of those queries generates zero CRM data. Zero attribution. Zero signal for your demand generation system. Article 2.3 examines how AI search and AI agents are accelerating this shift — and what it means for content strategy. Here, the point is simpler: the research channel that is growing fastest is the one your measurement infrastructure cannot see at all.
These five shifts are not independent problems. They are symptoms. The underlying condition is what Robin Caller describes as the structural transition from Campaign Thinking to Infrastructure Thinking. Campaign Thinking treats demand generation as a sequence of executional decisions — which channels, which messages, which offers. Infrastructure Thinking treats it as a governed system architecture — where signals, personas, content, and compliance operate as interconnected components rather than discrete campaigns. The five shifts make Campaign Thinking structurally incapable of generating pipeline. The question is what replaces it.
Where Did the Buying Journey Go? The Three Layers of the Dark Funnel
The dark funnel is not a gap in your measurement. It is the primary environment where B2B buying decisions are made.
To understand why, it helps to reframe the buying journey as two distinct phases. The Selection Phase is where buyers research, evaluate, build shortlists, and form preferences. It is largely anonymous and almost entirely invisible to vendor analytics. The Validation Phase is where buyers contact vendors, attend demos, and run procurement processes. This is what your CRM captures.
The critical insight — and the one most demand generation models fail to account for — is that 95% of the decision is effectively made during the Selection Phase. The Validation Phase is largely confirmatory. By the time a prospect appears in your pipeline, you have already won or lost.
The Selection Phase unfolds across three layers of the dark funnel, each with different characteristics and different implications.
Layer 1: Anonymous web activity. Ninety-seven percent of website visitors are anonymous (6sense aggregated platform data). They read your content, compare you against competitors, and shortlist or dismiss you without ever submitting a form. Your analytics show you traffic. They cannot show you intent, competitive context, or where in the decision process a given visitor sits.
Layer 2: Closed peer networks. Buying decisions are shaped by conversations in Slack communities, LinkedIn direct messages, analyst briefings, and board-level peer introductions. These interactions are often the highest-trust inputs in the entire buying process. A recommendation from a peer who has used your product is worth more than any amount of retargeting. None of it generates vendor-side data.
Layer 3: AI-mediated research. A buyer asks ChatGPT which demand generation platforms are worth evaluating. Gemini returns a shortlisted comparison. Perplexity summarises analyst commentary. The buyer forms a view. Your name is either on the list or it is not.
This is the fastest-growing layer of the dark funnel — and the one that has emerged most recently. It generates no CRM data, no attribution signal, and no opportunity for the traditional demand generation system to intervene.
The appropriate response to the dark funnel is not to find ways to “track the untrackable.” It is to be present, credible, and correctly positioned in all three layers before the buyer begins their active search. That is an infrastructure problem, not a campaign problem.
What Does the Old Model Actually Assume?
The traditional demand generation model was built on a set of assumptions about buyer behaviour. Each of those assumptions has been invalidated by the data.
| Legacy Assumption | Modern Reality |
|---|---|
| Buyers follow a linear funnel | The journey has bifurcated into a Selection Phase (60%) and a Validation Phase (40%). The winner is chosen before vendor contact (6sense 2025) |
| Form fills indicate buying intent | 97% of visitors never fill a form. An MQL generated at the 10% mark of the journey is a distraction when the buyer is already 61% through their evaluation |
| Individual leads represent opportunities | Committees of 10.1 members make collective decisions. Capturing one individual and calling it a lead misses 90% of the buying group |
| More outreach generates more pipeline | 73% of buyers actively avoid suppliers sending irrelevant outreach (Gartner 2025). Gartner predicts 2,000+ automated-outreach legal claims by end of 2026 |
| Website traffic equals visibility | Zero-click rates at 80–85% for AI-triggered results. Buyers get answers from AI without visiting your site |
| Sales reps guide the buying journey | 61% prefer rep-free experience. 69% report inconsistency between website and rep information, creating immediate mistrust (Gartner 2025) |
Each row represents a structural mismatch between how the model operates and how buyers actually behave. But the damage compounds in specific ways.
The first three rows — individual leads, more leads equals more pipeline, and form fills indicate interest — create the volume-over-quality incentive that drives most MQL-based operations. Marketing teams optimise for the metric they are measured on, which means optimising for form fills from individuals. The buying group, the anonymous research, and the preference formation that happened months earlier are invisible to that measurement system. The metric rewards the wrong behaviour.
The attribution row and the sales conversion row interact differently. Attribution models credit the last visible touchpoint — typically the form fill — while ignoring the months of anonymous research, peer conversations, and AI-mediated discovery that actually drove the decision. Sales then receives leads that look like they were generated by the campaign when they were generated by a process the campaign never touched. The result is a structural disconnect between what marketing reports and what sales experiences.
The final row — repeat leads treated as duplicates — compounds everything. Binary de-duplication suppresses one of the few signals that could indicate genuine escalating interest, removing it from the data before anyone can act on it. The economic model and the behavioural model are working against each other.
The cumulative effect is a demand generation system that is optimising for outcomes that bear almost no relationship to revenue.
Why the MQL Model Broke — and What the Data Says
The MQL is not a bad metric. It is an accurate metric for a buying process that no longer exists.
That is an important distinction. The problem with MQLs is not that they were badly designed — they were a reasonable proxy for intent in a world where individual buyers engaged visibly with vendor content and followed a roughly linear path. The problem is that the buying process changed, and the metric did not. Five structural failures compound each other.
The individual-lead problem. MQLs measure one person’s engagement in an era when 10 or more people make the decision. One researcher downloading three whitepapers triggers MQL signals; the CFO who holds budget authority and the procurement lead who will run the contract process generate none. The metric is measuring a single finger and calling it a hand.
The volume incentive. Marketing teams are rewarded for hitting MQL targets. This creates a structural incentive to optimise for form fills — gated content, aggressive retargeting, lead capture at the expense of brand presence. The tactics that generate MQLs are, in many cases, the same tactics that damage standing in the dark funnel. You trade long-term credibility for short-term numbers.
The attribution blind spot. MQL attribution tracks visible touchpoints. When 97% of website visitors are anonymous and first vendor contact happens at month seven or eight of a ten-month buying cycle, the vast majority of the journey is invisible to attribution. The models credit the last visible action — the form fill — while ignoring the months of anonymous research that actually drove the decision. The form fill looks like causation. It is usually correlation at best.
The sales-marketing disconnect. CMOs report hitting MQL targets. Sales reports that the leads are poor quality. Both observations are accurate. MQL targets are being met; the leads do not represent genuine buying intent. The metric has become structurally disconnected from revenue, and both functions are optimising against the wrong thing.
The economics are broken too. The industry’s standard commercial model compounds the problem in a way that rarely gets discussed. Binary de-duplication — the practice of treating any repeat lead submission as a non-billable duplicate — suppresses one of the most reliable buying signals available. When someone engages with your content for the second time eight weeks after the first, that is not duplication. It is escalating interest.
Leadscale’s Repeat Frequency Model (RFM) classifies repeat engagement into four categories: a Replica (an identical submission, genuinely worthless); a Repeat (a new enquiry from a known individual within a defined time window, signalling increasing intent); a Reactivation (engagement after extended dormancy, full value); and Net-New (a previously unknown individual). The traditional commercial model treats categories two and three as category one. The economic incentive structure of the lead generation industry is directly misaligned with the behavioural reality it is supposed to measure.
Three operational observations from Robin Caller’s client work make the consequences concrete.
The tenure paradox. “Look at your best customers. Many of them pre-date your current marketing operation. They found you despite your demand generation, not because of it.”
The nurturing investment gap. “It’s like BMW spending millions on brand awareness and then staffing the showroom with one person on a Saturday afternoon. Heavy investment in getting people interested. Thin provision for when they’re actually ready.”
The universal servicing failure. “Most B2B companies service every inbound lead identically, regardless of ICP fit. The result is that roughly 40% of servicing resource goes to accounts that will never close. That is not a sales execution problem. That is a system design problem.”
Where This Leaves You
The five structural shifts, three dark funnel layers, broken MQL model, and flawed commercial economics described in this article are not independent problems requiring individual fixes. They are symptoms of a single underlying condition: the demand generation model in widespread use was designed for a buying process that no longer exists.
The buyer who arrived through visible channels, engaged individually, followed a linear funnel, and could be captured with a well-timed email sequence is gone. The buyer who replaced them assembles shortlists anonymously, researches through AI tools, makes decisions through extended buying groups, and arrives at your door already knowing whether they want to talk to you — that buyer requires a different kind of demand generation system altogether.
The transition Robin Caller describes as moving from Campaign Thinking to Infrastructure Thinking is not a strategic preference. It is the structural reality of B2B buying in 2025 and 2026. If your demand generation reporting still centres on MQL volume, the gap between what you are measuring and what is actually happening is widening every quarter.
→ Article 2.3 examines how AI search and AI agents are accelerating these shifts — and what they mean for how your content needs to perform. Cluster 3 addresses the system design question directly: how you build a demand generation operating model that is calibrated for the way B2B buyers actually behave.
