Here's the facts, then the reasoning behind them: single-touch attribution models misattribute credit on more than 60% of multi-step B2B deals, because 61% to 83% of the buying journey now happens before a buyer ever contacts a vendor or an SDR or AE reaches out to someone on the buying team. Fixing this does not require an enterprise attribution platform. It requires clean CRM data, a shared sourced-and-influenced reporting standard, and, for most earlier-stage companies, nothing more expensive than the CRM they already pay for.
Here's what a real B2B deal looks like now, and where the story usually gets told wrong. A prospect signs up for a free trial with nothing but a work email, the lowest-friction conversion a company can ask for. That's the one real, trackable moment: no meeting booked, no sales conversation, just enough information to exist in the database at all. Then they go quiet. Over the next four months they read documentation, lurk in the product's community answering their own questions, work through an ungated video series at their own pace, and use the trial environment heavily enough to get their whole team comfortable with it, none of which pings anything a sales team is watching. As far as the CRM is concerned, that signup went cold the week it happened.
Then an SDR works through a list of stale trial signups and sends a routine check-in email. The reply comes back fast: we've actually been using this for months, the team's sold, are we still able to buy. The rep books the call. The deal closes a few weeks later.
CRM Deal Source: Sales Outbound. The quarterly review says sales-sourced deals are healthy, so why does marketing need more budget. Marketing has no good answer, not because the work didn't happen, but because the field that was supposed to tell the story only had room for the touch that happened to trigger the meeting, not the four months of product usage that made the buyer ready to take it.
That's not a hypothetical, and it's exactly why a single "Deal Source" field can't tell the truth about how a deal like this actually closes.
Quick answers
What percentage of the B2B buying journey happens before a buyer contacts sales?
Research converges on roughly 61% to 83%, depending on the study and how "contact" is defined. 6sense's Buyer Experience Report puts first vendor contact at 61% of the way through the journey, down from 69% the prior year, meaning buyers are waiting even longer to reach out.
How many touchpoints does it take to close a B2B deal, by company size?
It depends heavily on what's being counted. On meaningful sales touches, one 939-company benchmark finds SMB deals close in 5 to 7 touches, mid-market in 8 to 10, and enterprise in 12 to 15. Counted by ACV tier, a separate benchmark finds SMB deals average 14 touchpoints, mid-market 23, enterprise 38, and strategic deals above $500,000 average 51. Counting every digital impression instead of only meaningful engagements pushes the average as high as 266 touchpoints overall, and closer to 417 for deals at or above $100,000 in ACV.
What's the difference between marketing-sourced and marketing-influenced revenue?
Sourced revenue describes deals marketing originated from the start. Influenced revenue describes deals marketing touched at some point during the sales cycle without necessarily starting the relationship. Most companies report both but still treat sourced as the "real" metric, which misses most of what marketing actually contributes in a buying environment where most research happens anonymously.
Does sales compensation affect how deals get logged as sourced?
Often, yes. Comp plans that pay more for self-sourced deals give reps a real financial incentive to describe an ambiguous deal as self-sourced rather than digging into whether marketing touched it first. This isn't necessarily bad faith, but it means CRM attribution data is often skewed by pay structure, not just by tooling gaps.
Why does single-touch attribution get B2B marketing wrong?
Single-touch models are known to misattribute credit in more than 60% of multi-step buyer journeys. First-touch attribution overfunds top-of-funnel awareness work and starves what actually closes deals. Last-touch does the reverse, crediting the final interaction while erasing months of content, community, and self-serve product experience that built the trust behind it.
What is the dark funnel, and can it be measured?
The dark funnel is the portion of the B2B buying journey, anonymous browsing, peer conversations, AI-assisted research, review sites, that happens outside anything a tracking pixel or CRM field can capture. Even well-instrumented attribution programs report an average dark-funnel gap of around 38% of pipeline. AI-inferred signals and self-reported attribution can narrow it, but they're a supplementary read, not a full replacement.
What can actually be tracked with certainty, versus estimated?
The original source recorded the moment a contact enters the database, and a tightly time-bound trigger, like a demo request submitted minutes after clicking a specific email, are real facts, not models. Everything a multi-touch model weights across weeks or months of separate touches is a reasoned estimate, however accurate, and should be reported with that distinction intact.
What's the practical first step to fixing attribution instead of arguing about it?
Fix the data plumbing before switching models: clean campaign member associations, a formal UTM tagging policy, and a shared data layer marketing, sales, and finance all pull from. Roughly 30% to 40% of organizations cite dirty campaign data as their biggest attribution barrier, and no model survives dirty inputs regardless of sophistication.
What attribution software should a company with no dedicated ops person use?
Start with the native multi-touch reporting already built into HubSpot or Salesforce, paired with Google Analytics 4 and a self-reported "how did you hear about us" field. That combination is free beyond the CRM seat and closes most of the gap for companies under roughly $5 million in ARR. Dedicated platforms like Dreamdata or Factors.ai become worth their $150 to $750 monthly cost once volume justifies it, but tools built for RevOps teams, like HockeyStack or Heeet, are premature without someone accountable for running them.
The buyer really did do all of that before sales ever showed up
This isn't marketing being precious about credit. Multiple large-sample 2025 and 2026 studies converge on roughly 61% to 83% of the B2B buying journey happening before a buyer ever contacts a vendor directly, depending on how the study defines "contact." 6sense's Buyer Experience Report, drawing on nearly 4,000 buyers, puts first vendor contact at 61% of the way through the journey, down from 69% the year before, meaning buyers are reaching out even later than they used to. The same research found 83% of buyers fully define their purchase requirements before ever speaking with sales, 92% already have a vendor in mind before that first conversation, and 61% say they'd prefer a completely rep-free buying experience if they could get one. Forrester's most recent research on business buying puts the average deal at 13 stakeholders, each doing a meaningful share of that research independently. And across the whole cycle, buyers reportedly spend only around 17% of their total purchase time actually meeting with potential suppliers, split across every vendor they're considering, not just yours.
Every part of that anonymous research phase is marketing's work: the AI-search-visible answer, the community thread, the ungated video, the self-serve trial environment. None of it shows up as a "touch" in a CRM built around form fills and outbound emails. By the time sales makes contact, the deal is often functionally already decided, and the CRM records the one interaction that happened to be trackable as if it were the whole story.
How many touchpoints it actually takes, and why the number depends on who's counting
Ask how many touchpoints it takes to close a B2B deal and you'll get three genuinely different answers, and all three are correct for what they're actually measuring. A 939-company sales-ops benchmark from Optifai puts the average at 8 meaningful touches, with SMB deals closing in 5 to 7, mid-market in 8 to 10, and enterprise in 12 to 15. A separate 2026 benchmark segmented by deal size tells a longer story: SMB deals in the $1,000 to $25,000 ACV range average 14 touchpoints over an 84-day cycle, mid-market deals from $25,000 to $100,000 average 23 touchpoints over 121 days, enterprise deals from $100,000 to $500,000 average 38 touchpoints over 218 days, and strategic deals above $500,000 average 51 touchpoints over 312 days. HockeyStack's analysis of 150 B2B SaaS companies goes further still, counting every digital impression rather than only meaningful engagements, and lands on an average of 266 touchpoints and 2,879 impressions to close a deal overall, climbing to roughly 417 touchpoints for deals at or above $100,000 in ACV.
The gap between "8 touches" and "266 touchpoints" isn't a contradiction, it's a definitions problem, and it's the exact same problem this post is about. A touchpoint counted as a meaningful sales interaction, a call, a demo, a personalized email, is a completely different unit of measurement than a touchpoint counted as any recorded digital impression, an ad view, a retargeting hit, a passive content view. Both numbers are true. Neither is complete on its own. If your board asks how many touches your average deal needs and you answer with a single number, you're already repeating the same mistake this whole post is arguing against, just one level up: picking one methodology and treating it as the whole answer.
Why "Sales Outbound" as a deal source is technically true and completely misleading
A single-attribution field isn't lying, exactly. Sales genuinely did send that follow-up email, and the buyer genuinely did reply. The problem is that a single field can only hold one story, and it defaults to whichever touch happened to be the easiest one to log, almost always the most recent one, almost never the one that actually moved the buyer from curious to convinced.
This is precisely why attribution is described, accurately, as the fastest way to start a fight inside a B2B company. Marketing says they sourced the deal, pointing at months of content consumption. Sales says the relationship existed before any of that mattered, pointing at the reply that closed it. Both are describing something real. Neither is describing the whole thing. And the argument burns real time and real trust while the actual question, which activities across the whole cycle actually drove this revenue, goes unanswered.
The data backs up why the argument keeps happening: single-touch attribution models are known to misattribute credit in more than 60% of multi-step buyer journeys, and a typical enterprise B2B journey is nothing but multi-step. First-touch attribution systematically overfunds top-of-funnel awareness activity and starves the bottom-of-funnel work that actually closes deals. Last-touch does the reverse, crediting the final email while erasing the six months of content, community, and self-serve product experience that built the trust the email was riding on. Whichever single model a company picks, it's wrong for a meaningfully large share of its own pipeline, by design, not by accident.
The incentive nobody wants to name
There's a version of this fight that's rarely said out loud, and it isn't about dashboards. Plenty of sales comp plans still pay a real difference, a higher commission rate, an accelerator, a self-sourced bonus, for deals a rep can plausibly claim they brought in themselves, versus deals that came from a marketing-generated lead. That structure exists for a defensible reason: companies want reps who proactively build pipeline instead of waiting for leads to land in their queue. But it also means a rep has a direct financial incentive to describe a deal as self-sourced whenever the story is even a little bit ambiguous, and after everything in this post, the story is almost always at least a little bit ambiguous.
Nobody has to be acting in bad faith for this to distort the data. A rep who genuinely believes they closed a deal through their own outreach has no reason to dig into whether the prospect had been quietly reading the company's content for four months first, and every reason not to, if digging into it might reclassify the deal and shrink the commission. This isn't a call to redesign comp plans in this post, that's a separate and legitimate conversation. It's a reason to be honest about why "let's just ask the rep what happened" was never going to produce clean attribution data on its own, even before you get to the dark funnel problem. The incentive to log a deal one way rather than another is built into how sales gets paid, and any attribution fix that ignores that will keep finding the CRM data quietly skewed in the same direction, deal after deal, for reasons that have nothing to do with what buyers actually did.
Sourced versus influenced was never supposed to be either-or
The fix already exists in most attribution vocabulary, it's just usually implemented as a consolation prize instead of the actual answer. Marketing-sourced revenue describes deals marketing originated. Marketing-influenced revenue describes deals marketing touched at some point during the cycle without necessarily starting it. Most companies report both numbers, but still operate as if sourced is the "real" one and influenced is the participation trophy marketing gets instead of credit that counts.
That's backwards for exactly the buying pattern described above. In a world where the buyer does 60% to 80% of the work anonymously before ever filling out a form, "sourced" as a category is measuring an increasingly rare event, the moment a buyer happened to convert on a trackable form instead of just quietly deciding based on everything they'd already read. Influenced is where the actual explanatory power lives now, and treating it as secondary data instead of the primary story is exactly how a company ends up with a board saying "sales-sourced deals are doing fine, so why does marketing need more budget," when marketing built the entire case the sales-sourced deal closed on.
What to actually do about it
None of this means throwing out measurement, or pretending attribution is impossible and giving up. It means building the discipline multi-touch attribution was always supposed to be, and most companies never fully implemented, in the order that actually works.
Give people more than one honest way into the database. The free trial signup in the opening example works as an entry point, but it can't be the only one, and it shouldn't be. Gating every piece of content behind a form is falling out of favor for good reason: it trades a name and email for a worse experience at exactly the research stage where buyers are most likely to bounce to a competitor who didn't make them fill out a form. The fix isn't to gate more. It's to build several genuinely low-friction, recurring reasons for someone to hand over an email voluntarily: a live webinar signup, a recurring newsletter or content series worth actually subscribing to, a community account, a product waitlist, a benchmark tool or calculator that requires an email to save or share results. Each of these is a real, trackable conversion event, and having several of them instead of one single gated asset means a company isn't entirely dependent on trial signups to know a prospect exists at all.
Know the difference between what you can actually know and what you're estimating. Not every touchpoint carries the same evidentiary weight, and pretending otherwise is its own kind of dishonesty. Some things are genuinely, deterministically trackable: the original source recorded the moment a contact first enters the database, whatever campaign, referrer, or channel parameter was attached at that exact instant, is a fact, not a model. The same is true for a tightly time-bound trigger, if a contact clicks a specific email link and requests a demo in the same session minutes later, that's a real, direct causal chain you can point to, not an inference. Everything else, every touch that happened days or weeks before a conversion, every piece of content someone may have read without clicking through anything trackable, every influence a multi-touch model assigns partial credit to, is an estimate built on a methodology, however good that methodology is. Track the deterministic layer with real rigor, since it's the one part of this that doesn't require a model to trust. Be honest that the rest is a reasoned approximation, not a fact, and stop reporting modeled influence with the same false confidence as a hard conversion event.
Separate account-level signal from person-level contactability, and don't confuse the two. AI-driven intent tools can surface that a company is showing real buying signal, funding events, hiring patterns, technographic changes, before anyone at that company has ever given you their email. That's genuinely useful for prioritization. But it doesn't make that person reachable. Someone still has to convert, even through the lowest-friction path available, before a rep can follow up with them directly. The fix isn't pretending AI signal-surfacing replaces that conversion step. It's making sure everything a contact does after that initial low-friction signup, the product usage, the community activity, the content consumed, actually feeds back into their record instead of going dark the moment the original form fill stops being the most recent thing in the system.
Fix the plumbing before touching the model. 30% to 40% of organizations cite inaccurate campaign member associations as their single biggest attribution barrier, and 64% of B2B organizations still don't have a formal UTM tagging policy. No attribution model, however sophisticated, survives dirty inputs. A unified data layer that marketing, sales, and finance all pull from does more for attribution accuracy than switching models ever will, and this is free: it's a naming convention and a shared spreadsheet or CRM field, not a purchase.
Run sourced and influenced together, not as a hierarchy. Report both numbers for every deal, every quarter, and stop treating one as the real answer and the other as a footnote. The goal is a shared, board-legible language across marketing, sales, and finance, not a marketing-only metric marketing has to defend alone.
Add self-reported attribution as a cheap, high-signal layer. A simple "how did you hear about us" field, asked directly, starts producing genuinely useful directional data once you've collected roughly 30 or more responses. Buyers misremember the exact touchpoint, but they reliably remember which channel actually moved them, which is more useful for a B2B motion than pixel-perfect per-touch precision that dark-funnel behavior makes impossible to get anyway. This is a form field, not a tool purchase.
Choose a model that matches the actual buying motion, not the easiest one to set up. A nine-month enterprise sale with a twelve-person buying committee cannot be measured honestly by a model built for a single-session transactional purchase. Position-based, or full-path, models that weight the first touch, the last touch, and meaningful moments in between are a more honest fit for exactly the kind of long, multi-stakeholder cycle described earlier, and multi-touch adoption has grown from roughly 31% of B2B teams in 2023 to about 47% in 2026 for exactly this reason.
Treat the dark funnel as a real, permanent gap to plan around, not a problem to eventually solve. Even well-instrumented programs report an average dark-funnel gap of around 38% of pipeline that simply isn't visible to any attribution model, regardless of sophistication. AI-inferred signals, account-level engagement patterns, branded search lift, community and social mentions, can narrow that gap, but they're a supplementary read, not a replacement for admitting some of this will never show up as a clean, attributable line in the CRM.
Extend cohort thinking to match your actual sales cycle length. A model judged on 90 days of data in a business with a nine-month sales cycle will always look wrong, because it's being graded before the cycle it's measuring has even finished. Give the model as much runway as the buying motion actually needs before treating its output as settled.
The software stack, if you have next to no budget and no dedicated ops person
Most attribution advice is written for a company that already has a GTM Engineer or a RevOps team to run the tooling. Nearly all earlier-stage startups don't have that yet, and buying a sophisticated platform without someone to actually run it is how a real budget line turns into a dashboard nobody trusts.
Tier zero, cost nothing beyond what you already pay for. Your CRM's native attribution, HubSpot's built-in multi-touch reporting or Salesforce's Campaign Influence reports, does real multi-touch attribution today if the underlying data is clean. Pair it with Google Analytics 4, which is free and shows multi-channel conversion paths at the web-traffic level, a self-reported "how did you hear about us" field, and a written UTM naming convention everyone on the team actually follows. This tier alone closes most of the gap described in this post, and it requires discipline, not a purchase order.
Tier one, worth it once volume genuinely justifies a dedicated tool, roughly $150 to $750 a month. Dreamdata offers a genuine free tier for foundational B2B attribution before its paid plans start around $750 a month. Factors.ai combines account-level intent with multi-touch attribution starting near $399 a month on its self-serve tier. Ruler Analytics is the strongest option specifically for a business where phone calls close deals, with published pricing starting under $260 a month rather than a sales-call-only quote. All three connect directly to HubSpot or Salesforce and are built to be configured by a marketing generalist, not a specialist.
Tier two, needs a dedicated owner to actually pay off, $1,000 a month and up. HockeyStack, Heeet, and enterprise platforms like Adobe's Marketo Measure are built for complex, high-volume B2B motions with a RevOps or GTM Engineering function to configure and maintain them. These tools are genuinely more powerful, and they are also the tools most likely to become expensive, unused dashboards at a company that buys them before it has fixed its CRM data or before it has anyone whose job includes actually watching the output. If you don't yet have someone accountable for attribution as an ongoing responsibility, this tier is premature regardless of budget.
The honest sequencing: get tier zero genuinely right first. Most companies who think they need tier two actually have a tier zero data-hygiene problem that no amount of software spend fixes.
The real fix is cultural, not just technical
Every fix above is implementable with tools that already exist, several of them free. The harder fix is the one inside the room where the board says "marketing needs to drive more MQLs" right after looking at a chart that says sales-sourced deals are performing fine. That sentence is only coherent if sourced and influenced are being treated as competing categories instead of two views of the same revenue, and it gets worse, not better, once you factor in that the rep filling out that CRM field may have a commission reason to describe it one way rather than the other.
This is exactly the kind of gap I built the Win Report Builder to close from the deal-level side: every field on it, including how a deal actually got won, is meant to reflect the real, multi-touch story of a closed deal, not whichever single touch happened to be easiest to log at the time. And it's exactly the kind of misalignment my GTM Alignment Diagnostic is built to surface at the org level, before a board conversation turns a shared win into a manufactured fight between two functions that both did their job.
Kill the single-attribution model. Neither function gets a deal over the line alone, and the measurement system should say so out loud instead of quietly picking a winner every time a deal closes.
Sources
- 6sense, 2025 B2B Buyer Experience Report, and 6sense/Green Hat APAC research on anonymous buyer behavior (first-contact timing, requirement definition before sales contact, vendor-in-mind statistics, rep-free preference).
- Forrester, State of Business Buying research (average buying committee size, share of purchase time spent with vendors).
- Similarweb, "The Dark Funnel Paradox: Why Your Best B2B Leads Are Invisible" (AI referral traffic patterns, dark funnel measurement approach).
- Optifai, "B2B Lead Touches to Conversion, by Deal Size & Channel" (939-company sales-ops benchmark, touchpoints by segment).
- Digital Applied, "B2B Marketing Statistics 2026: 180+ Essential Data Points" (touchpoints and cycle length by ACV tier, buying committee size).
- HockeyStack Labs, "B2B Customer Journey Touchpoints: The Impact of Deal and Company Size" (impression and touchpoint volume by deal size, 150-company analysis).
- Pedowitz Group, "How to Choose Marketing Attribution Models in 2026" (sourced versus influenced definitions, attribution as an internal conflict driver).
- TapClicks, "Marketing Attribution Models Explained (2026 Guide)" and "Marketing Attribution in 2026: Why Multi-Touch and Marketing Mix Modeling Have to Work Together" (single-touch misattribution rate, multi-touch adoption growth, martech stack fragmentation).
- Digital Applied, "Marketing Attribution Statistics 2026: 140 Data Points" (multi-touch and MMM adoption figures, average dark-funnel pipeline gap, AI-attribution accuracy lift).
- Marketing Mary, "Marketing Attribution Models 2026: Multi-Touch vs Last Click" (campaign member association data-quality barriers, UTM policy adoption gap).
- ziellab.com, "B2B marketing attribution: why multi-touch lies in 2026" (self-reported attribution signal quality, cohort analysis timelines).
- DarwinApps, "9 Best Marketing Attribution Tools for Mid-Market SaaS Teams in 2026," Hey Sid, "Best Marketing Attribution Software for B2B 2026" and "Best Dreamdata Alternatives for B2B Attribution 2026," and AttributeIQ, "Top 5 Affordable Dreamdata Alternatives for B2B Marketing Attribution" (attribution software tiering and pricing).