The webinar is the worst top-of-funnel in the matrix and the best middle — the model uses it accordingly

Growth has slowed and budgets have not grown to compensate. SaaS Capital's 2026 survey of more than a thousand private B2B SaaS companies puts median growth at 22%, down from 25% the year before, with marketing at 8% of ARR and sales at 15% — figures that have not moved to accommodate the slowdown. Inside an 8% marketing budget the question is not which channel produces the most leads, it is which channel produces the fewest leads that waste sales time. The published funnel matrix answers that in a way most teams have backwards.

A webinar converts visitors to leads worse than any other channel in the published matrix and converts those leads to qualified opportunities better than any other. It is a filter that has been sold as a funnel.

1. 01

   The webinar is the weakest top of funnel and the strongest middle

   In the published channel-by-stage matrix, webinar visitor-to-lead sits at 0.9%, the lowest of the five channels, while lead-to-MQL is 44% and opportunity-to-close is 40%, both the highest. Measuring a webinar on registrations therefore measures it on the only stage where it is bad.

2. 02

   Attention, not registration, is the asset the format delivers

   Attendees engage for 51 minutes against a 49-minute average runtime, and 60% of registrants show up. Nothing else in the demand mix holds a qualified buyer's attention for that long, which is why the session content has to be worth the hour on its own terms.

3. 03

   The outbound benchmark favours a scaleup selling to scaleups

   Targets at companies of 0 to 10 employees reply at 0.72% against 0.22% for companies of 10,000 or more, and founders and owners are the highest-responding seniority group at 0.57%. A small company selling to small companies is working with the grain of the published data; the same programme aimed at enterprise is not.

4. 04

   Cost per registrant is not published anywhere, so it has to be instrumented from day one

   No credible publisher reports a webinar cost per registrant — not the platform vendors, not the research houses. The model computes it from first-party send cost and registration count in the first cycle rather than benchmarking it, and treats any agency quoting a market figure as quoting nothing.

5. 05

   Inside an 8% marketing budget the sequencing matters more than the channel mix

   One session per quarter with a properly built list and a real follow-up beats a monthly cadence with a thin list, because the format's value sits in the 44% lead-to-MQL step and that step depends entirely on who was in the room.

**~57% of registrants attend** · average engagement **~49–51 minutes** (ON24 2025→2026) · ~half of all attendees watch on-demand. *(The 51- vs 49-minute figures are consecutive-edition engagement numbers, not "engagement vs runtime.")*

The model builds the list before it builds the session, weighting toward the company sizes and seniorities the published reply data favours. Invitations run as a two-step: a first email that asks whether the topic is relevant and carries no link, then a registration link only to the people who engaged. The session is designed to be useful to a non-buyer, because a room full of attendees who came for the content is the input to the 44% qualification step. Attendance and engagement time are recorded per attendee, and follow-up is prioritised on engagement time rather than on job title. Cost per registrant is computed from first-party numbers in cycle one and becomes the internal benchmark thereafter.

1. 1

   Weeks 1–2

   Demand audit

   Current channel mix costed against an 8%-of-ARR marketing envelope

   Owner: OmniFlow

2. 2

   Weeks 2–4

   List build

   Weighted to company sizes and seniorities the published reply data favours

   Owner: OmniFlow

3. 3

   Weeks 3–5

   Session design

   A session useful to a non-buyer; engagement tracking configured per attendee

   Owner: OmniFlow + client experts

4. 4

   Week 6

   Invitation sequence

   Interest check with no link, then registration link to engaged recipients only

   Owner: OmniFlow

5. 5

   Weeks 7–8

   Delivery and follow-up

   Live session, recording to non-attendees, follow-up ordered by engagement time

   Owner: Client + OmniFlow

6. 6

   Week 12

   Review

   First-party cost per registrant, cost per qualified opportunity, engagement-time distribution

   Owner: OmniFlow

At the published rates, 100 registrations produce roughly 60 attendees, and the webinar-channel benchmark would qualify roughly 44% of the resulting leads to MQL — a higher share than any other channel in the matrix, from a smaller starting number than any other channel would have produced. That trade is the entire argument for the format. The model reports cost per qualified opportunity rather than cost per registration, and replaces every published rate with first-party data after one cycle.

Modelled. Inputs: ON24, 2026 Webinar Benchmarks Report (2025 data); First Page Sage, B2B SaaS Funnel Conversion Benchmarks (June 2025); Belkins, B2B cold email response rates 2026 study (7,530,489 emails); 6sense, B2B Buyer Experience Report 2025 (~4,000 respondents). Modelled outputs are not a forecast or a guarantee of results. Cited: ON24 *2025 / 2026 Webinar Benchmarks*.

Every figure below is published research, not a client result. They are the inputs to the arithmetic above, listed so it can be checked rather than taken on trust. The bracketed number points to the full citation at the end of this page.

These are the surfaces this engagement is run and measured from, shown with representative figures built around the benchmarks cited on this page. Every account we run reports into views like these, and you keep ownership of all of them.

Each figure on this page, the system it is read from, and the definition and window it is measured over.

Not applicable — this is a modelled engagement. Its weakest input is the funnel matrix itself: it is agency client data gathered under stated conditions of competent execution rather than a survey, it discloses no sample size per cell, and it is the only channel-by-stage matrix published anywhere, which means there is nothing to cross-check it against. Every stage rate in this model should be replaced by the client's own within one cycle, and the 44% figure in particular should be treated as a hypothesis rather than as a rate.

A modelled capability, not a client account. Figures illustrate what the model produces and are labelled as modelled wherever they appear. See all our work

Full citations for everything cited on this page, with the sample and period each study covers, so you can go and read the original.