Dark funnel statistics at a glance
| Category | Statistic | Source |
|---|---|---|
| Journey | Buyers first contact a vendor at 61% of the journey, down from 69% in 2024 | 6sense, 2025 |
| Journey | The average buying cycle runs 10.1 months, down from 11.3 months in 2024 | 6sense, 2025 |
| Journey | 187 days pass before a vendor is contacted | Computed from 6sense, 2025 |
| Touchpoints | The average journey carries 88 touchpoints, up from 76 | Dreamdata, 2026 |
| Touchpoints | 81% of the B2B customer journey happens outside the sales pipeline, up from 70% | Dreamdata, 2026 |
| Shortlist | The winner came from the day-one shortlist 95% of the time, up from 85% in 2024 | 6sense, 2025 |
| Shortlist | The first vendor contacted wins 77% of deals | 6sense, 2025 |
| Shortlist | About 90% of buyers buy from their day-one list, in a survey of 750 enterprise software buyers | Bain, June 2026 |
| Attribution | First-touch credited 28 of 189 leads that named an AI tool as their source | Omniscient Digital, 2026 |
| Dark social | 100% of test visits from TikTok, Slack, Discord, Mastodon and WhatsApp were logged as direct | SparkToro, 2023 |
| Buying group | Decisions involve 13 internal stakeholders and 9 external participants | Forrester, 2026 |
| Buying group | Dreamdata counts 10 stakeholders per deal, up from 6.8 | Dreamdata, 2026 |
| AI research | 51% of B2B software buyers start research in an AI chatbot, up from 29% in April 2025 | G2, March 2026 |
| AI research | Review sites (38%) passed AI chatbots (37%) as the top shortlist influence | G2, July 2026 |
| Familiarity | 85% of buyers had prior experience with the vendor they chose | 6sense, 2025 |
What is the dark funnel, and how much of a B2B purchase happens inside it?
The dark funnel is the part of a B2B purchase that a vendor's own systems never record, and Dreamdata's 2026 benchmarks put it at 81% of the customer journey. That figure comes from a vendor dataset covering more than 66 million sessions and 3.5 million customer journeys.
| Statistic | Source |
|---|---|
| 81% of the B2B customer journey happens outside the sales pipeline, up from 70% the year before. | Dreamdata, 2026 LinkedIn Ads Benchmarks Report |
| 69% of the purchase process is complete before B2B buyers engage a seller, from a survey of 2,509 recent buyers. | 6sense, 2024 |
| 81% of buyers have already chosen a preferred vendor before they speak to sales. | 6sense, 2024 |
| 85% of buyers set their purchase requirements before contacting sales. | 6sense, 2024 |
| 6sense splits the journey into a Selection Phase covering the first 70% and a Validation Phase covering the final 30%. | 6sense, 2024 |
| At any moment about 95% of a B2B market is out of market and only 5% is in market, the 95:5 rule LinkedIn's B2B Institute published with the Ehrenberg-Bass Institute and Dreamdata uses to plan LinkedIn budgets. | LinkedIn B2B Institute |
Two independent datasets, one from a survey and one from clickstream and CRM data, land within a few points of each other. That agreement is worth more than either number alone. It means the 80/20 split people quote from memory is roughly right, and it means most of what marketing does is judged on the 19% that happens to be measurable.
How long do B2B buyers research before a vendor knows they exist?
B2B buyers research for about 187 days before contacting a vendor, computed from 6sense's 2025 finding that first contact lands at 61% of a 10.1-month journey.
| Statistic | Source |
|---|---|
| The average buying cycle fell from 11.3 months in 2024 to 10.1 months in 2025, across roughly 4,000 buyers. | 6sense, 2025 |
| The point of first contact moved from 69% of the journey in 2024 to 61% in 2025, about six weeks earlier. | 6sense, 2025 |
| Dreamdata measures the average B2B buying journey at 272 days. | Dreamdata, 2026 |
| 281 days pass between the first ad impression and closed revenue. | Dreamdata, 2026 |
| LinkedIn's click attribution window closes at 180 days, roughly 100 days short of the average journey. | Dreamdata, 2026 |
| 62% of buyers said economic pressure pushed them to engage sellers earlier, and 49% said it shortened their cycle. | 6sense, 2025 supplemental survey, n=766 |
The 180-day attribution window against a 272-day journey is the single most useful pair of numbers here. Any ad platform report you read is structurally incapable of seeing the deals it helped create, because the deal closes after the platform has stopped counting. That is not a tracking bug you can configure away.
How many touchpoints in a B2B buying journey go untracked?
About 71 of the 88 touchpoints in Dreamdata's 2026 benchmark set, drawn from 3.5 million customer journeys, happen outside the sales pipeline. That works out at roughly seven untracked interactions for each of the ten stakeholders Dreamdata counts on a deal.
| Statistic | Source |
|---|---|
| The average journey carries 88 touchpoints, up from 76 the year before. | Dreamdata, 2026 |
| Stakeholders per deal rose from 6.8 to 10. | Dreamdata, 2026 |
| Buying decisions involve an average of 13 internal stakeholders plus 9 external participants such as consultants and partners. | Forrester, The State of Business Buying 2026 |
| 53% of buyers name procurement professionals as decision-makers, and procurement now joins earlier in the cycle. | Forrester, 2026 |
| More than 60% of business buyers run a trial before deciding, rising to 78% on purchases of $10 million or more. | Forrester, 2026 |
| 75% of buyers say they personally know sellers at the vendors they evaluated. | 6sense, 2025, n=~4,000 |
| Hidden buyers, the people a vendor never meets, hold half the influence over the decision, from a NewtonX survey of 750 enterprise software buyers in February 2026. | Bain, June 2026 |
Forrester counts 22 people and Dreamdata counts 10. The gap is the method: Forrester asks buyers who was involved, Dreamdata counts who left a digital trace. Half the buying group never touches anything you can log, which is a cleaner statement of the dark funnel problem than any percentage on this page.
How big is the gap between self-reported and software attribution?
First-touch attribution caught 15% of the leads that named an AI tool as their source, a 6.8x undercount. The data comes from Omniscient Digital, a B2B agency that published two years of its own lead records in August 2026, and 81% of the misses were filed under organic search or direct.
| Statistic | Source |
|---|---|
| Of 189 leads who named a generative engine in an optional "how did you hear about us" field, first-touch credited 28 to AI referrals. | Omniscient Digital, 2026 |
| The other 161 were filed as Organic Search (95), Direct (59) and Referral or Other (7). | Omniscient Digital, 2026 |
| Missed AI-sourced leads climbed from 6 per quarter to 50 per quarter over two years. | Omniscient Digital, 2026 |
| GA4 last-touch credited AI Search with 9.39% of conversions in the trailing 90 days to August 2026. | Omniscient Digital, 2026 |
This is one company's data, collected over two years from an optional form field, so treat the ratio as an order of magnitude rather than a benchmark. What makes it useful is the direction of the error. The misattributed leads did not land in an unknown bucket where a careful analyst would notice them. They were credited to organic search and direct, two channels that look fully measured on a dashboard. If you have read our marketing attribution statistics, this is the mechanism behind the model disagreements on that page.
I run a self-reported attribution field on every form for exactly this reason, and the answers rarely match the software. The interesting part is not that the software is wrong. It is that the software is confidently wrong in a specific direction, always crediting the last measurable channel, which happens to be the one closest to the sale and cheapest to defend in a budget meeting.
Why does dark social traffic show up as direct?
Dark social traffic appears as direct because the sending platform strips the referrer, and SparkToro's controlled test with Really Good Data found that 100% of visits from TikTok, Slack, Discord, Mastodon and WhatsApp reached analytics with no referral data at all.
| Statistic | Source |
|---|---|
| Across 1,113 deliberate test visits on 11 social networks, TikTok, Slack, Discord, Mastodon and WhatsApp passed no referrer on any visit. | SparkToro and Really Good Data, April 2023 |
| 75% of visits from Facebook Messenger carried no referral information. | SparkToro and Really Good Data, 2023 |
| Instagram DMs lost referral data on 30% of visits, LinkedIn public posts on 14% and Pinterest on 12%. | SparkToro and Really Good Data, 2023 |
| Google.com originated 63.41% of all US web traffic referrals from the top 170 referring sites, measured in January 2024 inside a 13-month clickstream panel. | SparkToro and Datos, March 2024 |
Both SparkToro studies are older than two years, which I flag because platforms change their referrer policy without telling anyone. I keep citing them because nobody has repeated the experiment since, and because the mechanism is a design choice rather than a trend: a private messaging app has no reason to tell your analytics where a link was shared. For B2B, this matters most in Slack and WhatsApp, which is where B2B shortlists get debated.
The practical consequence is small and annoying. Your direct traffic line is a mix of people typing your name, people clicking a link a colleague sent, and people who read about you somewhere and came back later. Those are three different marketing outcomes sharing one row in a report, and no amount of UTM discipline separates them, because you do not control the link your buyer forwards.
Who is in the buying group, and how much of the decision is already made?
Most of the decision is made before you are contacted: 6sense's 2025 survey of roughly 4,000 buyers found the winning vendor came from the buyer's day-one shortlist 95% of the time, up from 85% in 2024.
| Statistic | Source |
|---|---|
| Buyers fill 3.6 of their shortlist spots on day one and evaluate 5.1 vendors in total, up from 4.5. | 6sense, 2025 |
| The winning vendor came from the day-one shortlist 95% of the time, against 85% in 2024. | 6sense, 2025 |
| The first vendor a buyer contacts wins 77% of deals; with no pre-ranked shortlist, the first vendor wins only 57%. | 6sense, 2025, n=3,744 |
| 97% of buyers have prior experience with at least one vendor on their shortlist. | 6sense, 2025 |
| 85% of buyers report prior experience with the vendor they ultimately chose. | 6sense, 2025 |
| Around 90% of buyers purchase from their day-one list, from a NewtonX survey of 750 enterprise software buyers fielded in February 2026. | Bain, June 2026 |
| 83% of buyers shortlisted three or fewer products, from a January 2026 survey of 1,862 technology buyers. | TrustRadius, 2026 B2B Buying Disconnect |
Bain's 90% and 6sense's 95% come from different samples and different methods, and they land in the same place. A day-one shortlist slot is worth roughly eight times the win probability of a slot added later, and that is the strongest argument I know for treating brand and category presence as pipeline work rather than as a separate budget. It is also why account-based marketing programmes tend to measure their own reach rather than their own clicks.
The 95% number frightens people and it should not. Read it the other way round: the shortlist is written by memory, and memory is built by everything that happened months before the buying cycle opened. That is buyable. What is not buyable is a place on a list that was finished before your campaign started.
Are AI assistants making the dark funnel bigger?
Yes, measurably. G2's March 2026 survey of 1,076 software buyers found 51% now begin research inside an AI chatbot rather than a search engine, up from 29% in April 2025, and the assistant leaves no referrer behind.
| Statistic | Source |
|---|---|
| 51% of B2B software buyers start their research in an AI chatbot, up from 29% in April 2025, from a survey of 1,076 buyers. | G2 via PR Newswire, March 2026 |
| 69% of buyers chose a different vendor than they had planned based on AI chatbot guidance, and 33% bought from a vendor they had not previously known. | G2, March 2026 |
| 82% of buyers sourced software recommendations from an AI chatbot in the previous 24 months, from a G2 survey of more than 1,000 software buyers published in July 2026. | G2, July 2026 |
| Review sites (38%) overtook AI chatbots (37%) as the top source shaping which vendors make a shortlist. | G2, July 2026 |
| 94% of buyers used large language models during their purchase. | 6sense, 2025 |
| 63% of technology buyers used AI somewhere in their purchase journey, from a January 2026 survey of 1,862 buyers. | TrustRadius, 2026 B2B Buying Disconnect |
| 94% of the buyers who used AI fact-check its responses at least some of the time. | TrustRadius, January 2026 |
| 66% of B2B professionals regularly use AI to research vendors, and 92% say AI shaped their shortlist, from 622 valid responses collected March to April 2026. | Semrush, 2026 |
| After an AI assistant mentions a vendor, 71% visit the vendor site, 63% search Google for the company, 38% check G2 and 14% ask a colleague. | Semrush, 2026 |
Semrush's last row is the one to keep. When an assistant names you, most of the resulting behaviour is a branded search or a direct visit, so the credit lands on brand search and direct in your reports. The channel that changed the buyer's mind never appears. Our ChatGPT vs Google search statistics page covers the traffic side of the same shift.
Two G2 surveys four months apart disagree about whether AI or review sites lead the shortlist, 37% against 38%, and that is a tie, not a trend. The stable finding across both is that buyers now form opinions in two places you cannot instrument: a chat window and a reviews page you do not own.
How we calculated the original numbers
Four figures on this page are not published anywhere else. Here is the arithmetic, so you can check it or swap the inputs.
- 187 days of buying journey before a vendor is contacted. 6sense's 2025 report gives an average cycle of 10.1 months and a point of first contact at 61% of that journey. 10.1 months x 30.44 days = 307 days. 307 x 0.61 = 187 days. The 2024 equivalent was longer on both counts, 11.3 months at 69%, or 237 days, so the invisible stretch shortened by about 50 days in a year.
- 71 untracked touchpoints, about 7 per stakeholder. Dreamdata's 2026 benchmarks report 88 touchpoints per journey and 81% of the journey happening outside the sales pipeline. 88 x 0.81 = 71.3 touchpoints. Dreamdata counts 10 stakeholders per deal, so 71.3 / 10 = 7.1 untracked interactions per person. This assumes touchpoints are distributed evenly across the group, which they are not, so read it as an average and not as a per-person target.
- A 6.8x attribution undercount, with 81% of misses hidden in two channels. Omniscient Digital's two-year sample had 189 leads name a generative AI tool in a self-reported field, while first-touch attribution credited only 28 of them to AI referrals. 189 / 28 = 6.75. Of the 161 misses, 95 were filed as Organic Search and 59 as Direct: (95 + 59) / 189 = 81.5%.
- A day-one shortlist slot is worth about 8x a later one. 6sense reports 3.6 shortlist spots filled on day one out of 5.1 vendors evaluated, leaving roughly 1.5 slots filled later, and says the winner comes from the day-one list 95% of the time. Per slot, that is 0.95 / 3.6 = 26.4% for a day-one vendor and 0.05 / 1.5 = 3.3% for a late entrant, a ratio of 7.9.
What the 2026 dark funnel numbers mean
The dark funnel is the shape of the market rather than a measurement failure. Buyers do 187 days of work before they call you, involve up to 22 people, and settle their shortlist on day one. No tool fixes that, because nothing in it is happening on your property. The teams that handle it well stop trying to see the journey and start trying to be the memory that gets written into it.
Self-reported attribution is now the cheaper instrument. One form field caught 6.8 times more AI-sourced leads than a first-touch model, and it cost nothing to run. It is noisy, it is biased toward whatever the buyer remembers last, and it still beats a model that quietly files half your demand under direct.
The measurable share is shrinking, not growing. Dreamdata's untracked share went from 70% to 81% in a year, AI chat research nearly doubled in eleven months, and ad platform attribution windows still close 100 days before the average deal does. Plan your 2027 reporting on the assumption that the number you can defend will be smaller than this year's, and decide now what you will use in its place. See the full set of statistics pages for the neighbouring benchmarks.
FAQ
What is the dark funnel in B2B marketing?
The dark funnel is the 81% of the B2B customer journey that happens outside the sales pipeline, a share Dreamdata measured in 2026 across more than 66 million sessions and 3.5 million customer journeys. It covers private messages, community threads, peer conversations, review sites, AI assistants and word of mouth, none of which report back to a vendor's analytics.
How much of the B2B buying journey is untracked?
About 81%, according to Dreamdata's 2026 benchmarks, up from 70% the year before. 6sense's 2025 survey of around 4,000 buyers reaches a similar place from a different direction: buyers first contact a vendor at 61% of a 10.1-month journey, which is 187 days of research the vendor never sees.
How many touchpoints are in a B2B buying journey?
88 on average, according to Dreamdata's 2026 LinkedIn Ads Benchmarks Report, up from 76 the year before. Applying the same report's 81% untracked share, roughly 71 of those touchpoints happen outside the sales pipeline, spread across the 10 stakeholders Dreamdata counts per deal.
Is self-reported attribution better than software attribution?
It caught 6.8 times more AI-sourced leads in Omniscient Digital's 2026 analysis: 189 leads named a generative AI tool in a form field, while first-touch attribution credited 28. Self-reported data is biased toward what buyers remember, so most teams run both and treat a large gap as a signal rather than an error.
Why is so much of my traffic labelled direct?
Because the platforms that send it strip the referrer. SparkToro's 2023 test of 1,113 visits found 100% of visits from TikTok, Slack, Discord, Mastodon and WhatsApp arrived with no referral data, and 75% of Facebook Messenger visits did the same. That study is now dated, and no one has repeated it, so treat the exact percentages as indicative.
Do AI chatbots make attribution harder?
Yes. G2 found 51% of B2B software buyers start research in an AI chatbot in March 2026, up from 29% in April 2025. Semrush's 2026 survey of 622 professionals found that after an AI assistant names a vendor, 71% visit the vendor site and 63% search Google, so the credit lands on direct and branded search instead.
Sources
- 6sense, 2025 B2B Buyer Experience Report (October 2025, about 4,000 buyers)
- 6sense, B2B Buyer Experience Report (2024, n=2,509 buyers)
- Dreamdata, 2026 LinkedIn Ads Benchmarks Report (March 2026, 66M+ sessions, 3.5M+ journeys)
- Dreamdata, How to Win B2B LinkedIn Ads Across a 272-Day Sales Cycle (June 2026)
- Omniscient Digital, First-Touch Attribution Captures 15% of Our AI-Sourced Leads (August 2026)
- G2, 2026 Buyer Behavior Report (July 2026, more than 1,000 software buyers)
- G2 via PR Newswire, Half of B2B Software Buyers Now Start Their Research With AI Chatbots (March 2026, n=1,076)
- SparkToro and Really Good Data, Dark Social Falsely Attributes Web Traffic as Direct (April 2023, 1,113 test visits)
- SparkToro and Datos, Who Sends Traffic on the Web and How Much (March 2024, clickstream panel January 2023 to January 2024)
- Forrester, The State of Business Buying, 2026 (January 2026)
- Forrester via Digital Commerce 360, B2B buying groups expand as they question AI (January 2026)
- Semrush, How AI Tools Shape the B2B Buying Process (2026, n=622, fielded March to April 2026)
- Bain, Raising the Odds on a Deal: How Likelihood to Buy Rewires B2B Growth (June 2026, NewtonX survey of 750 enterprise software buyers, February 2026)
- TrustRadius, 2026 B2B Buying Disconnect Report (January 2026, 1,862 technology buyers and 444 vendors)
- LinkedIn B2B Institute, The 95-5 Rule (with the Ehrenberg-Bass Institute)
Methodology. Numbers were collected in September 2026 and checked against the original publisher's page, not against statistics roundups. Vendor claims without a stated sample or method were excluded. Gartner's B2B buying research is widely quoted on this topic and is not cited here because gartner.com blocks automated access, so none of its figures could be verified at source. The two SparkToro studies are older than 24 months and are flagged as dated in the text. Where a single company's first-party data is used, as with Omniscient Digital's attribution comparison, the sample and its limits are stated in the same paragraph. 6sense, Dreamdata, Omniscient Digital, G2, TrustRadius and Semrush publish data about their own market, so each figure carries its year and sample in the row or sentence that uses it. Last updated 20 September 2026.