Open the analytics tab in your webinar platform right now. You'll see registrations, attendance rate, peak concurrent viewers, average watch time, maybe a Q&A count and a poll response rate. Nice charts. Big numbers.
Now answer this: which one of those tells you who's going to buy?
None of them. They tell you who showed up. Pipeline is a different question. (If you're still pulling that data out of Zoom by hand, the Zoom webinar attendee report walkthrough covers the basics of where to find each column before you start scoring them.)
That's the gap most webinar programs never close. You wrote the page, ran the ads, sent reminders, hosted the session, made the offer. The platform tells you 142 of 380 registrants joined for an average of 38 minutes. Now what?
Now you need a different set of metrics. The five below are the ones that actually move when pipeline moves.
The Vanity 4: metrics that look good and mean little
Before the predictive metrics, name the ones to stop optimizing.
- Total registrations. Measures top-of-funnel demand. You can double it and not move a single deal. Useful as a leading indicator of ad spend efficiency, not as a quality signal.
- Attendance rate. Measures reminder cadence and timing. ON24's 2024 Webinar Benchmarks Report puts the average registrant-to-attendee conversion at 57% when you count live and on-demand together; live-only attendance for most B2B programs sits in the 35 to 40% range. If yours is below 25% live, you have a reminder problem, not a content problem (see our breakdown on how to increase webinar attendance). Above 45% usually means a very warm list. Neither tells you about intent.
- Peak concurrent viewers. Measures the strength of your opening 10 minutes. It tells you the hook worked. It does not tell you anything about who stayed for the offer.
- Average watch time. This is the most misleading metric webinar platforms surface. It averages the person who left at minute 5 with the person who stayed for the full 60 and called it engagement. Two very different signals collapsed into one useless number.
You can run a program that beats every benchmark on the Vanity 4 and source zero pipeline. The teams that win on webinars stopped reading these as primary metrics years ago.
The Pipeline-Predictive 5
These are the ones to track instead. Each one is a direct or near-direct signal of buyer intent.
1. Chat engagement density
Messages per attendee per 10-minute window.
Raw chat count is meaningless. A 200-person webinar with 80 chat messages looks active until you compute the per-attendee number: 0.4 messages per person. A 50-person webinar with 200 messages is 4 per person. That second room has a fundamentally different relationship to your content.
What it predicts: audience fit and depth of attention. People type while they watch when something landed enough to react to. Density across the session, not just the opening, is the strongest signal that you're in front of the right audience.
How to read it: track density by 10-min window. If it spikes in the first 15 minutes and dies, your opening worked but the body lost them. If it builds toward the offer, you're warming the room correctly.
2. Question-intent ratio
Percentage of attendee chat messages that are buying-shaped questions versus general or educational ones.
A buying-shaped question is someone mentally trying on your solution. Examples:
- "How does this integrate with our existing CRM?"
- "Is there a plan for teams under 10?"
- "What's the implementation timeline?"
- "Do you have case studies in healthcare?"
A general question is someone exploring the topic:
- "Can you say more about that?"
- "What was the chart at minute 20?"
- "Where can I read more?"
Both have value. Only one signals proximity to a purchase. The ratio of the first to the second is the strongest single-signal predictor of pipeline from any given event.
How to read it: high question-intent ratio plus low reply rate to your follow-up means your follow-up is broken. The buyers were in the room and you failed them. High ratio plus high reply rate means the system is working.
3. Watch-time depth past the offer
Percentage of attendees who stayed in the session beyond the moment you made the offer or revealed pricing.
This replaces average watch time. Time the offer reveal, export the attendance log, count who stayed past that minute.
Why it predicts: most people who aren't going to buy leave the moment friction appears. Pricing, the ask, the CTA, the "if you want to work with me" beat. The people who stay past that point have already decided you might be a fit. They're sticking around to learn how, not whether.
How to read it: this is your hot list. Past-offer stayers convert at multiples of average attendees. If past-offer stay rate is below 30%, your offer is either mis-positioned or the audience is wrong. Above 50% means you're in front of the right people.
4. In-session link clicks
Percentage of attendees who clicked any link you shared live during the session.
A live click is a costly action. The person had to leave the video to do it. They did it anyway. That's active intent.
How to read it: in-session clicks are not distributed evenly. They concentrate in your buyers. If you tag the link with UTMs you can also see which audience segment is clicking, which gives you a real-time ICP read on the room.
The catch: Zoom does not natively track clicks on links shared inside the session. Its webinar source-tracking links capture where registrants came from before the event, not what they clicked during it. You need UTM-tagged links and a click attribution tool, or a layer that observes the session.
5. Reply rate by intent segment
Reply rate to the post-event follow-up, broken out by intent segment.
This is the closing-the-loop metric. The first four tell you what happened in the room. This one tells you whether your follow-up is meeting the room. (If your segmented follow-up is still one broadcast to the whole list, the webinar follow-up email guide covers how to split it.)
How to read it: if Hot reply rate is 3 to 5x Cold reply rate, your segmentation is calibrated and your follow-up is differentiated. If reply rates are flat across segments, either your scoring is broken or your follow-up isn't actually different across segments. Both are fixable. Neither is visible if you only track the average reply rate.
This metric also validates the first four. If your question-intent ratio was high but Hot reply rate was low, the buyers were there and you sent the wrong email. That's a useful failure to surface.
Vanity vs. Pipeline-Predictive at a glance
| Metric | Type | What it actually predicts |
|---|---|---|
| Total registrations | Vanity | Top-of-funnel demand and ad spend efficiency |
| Attendance rate | Vanity | Reminder cadence quality |
| Peak concurrent viewers | Vanity | Strength of the opening 10 minutes |
| Average watch time | Vanity | Nothing reliable. Stop using it. |
| Chat engagement density | Predictive | Audience fit and depth of attention |
| Question-intent ratio | Predictive | Presence of buyers in the room |
| Watch-time depth past the offer | Predictive | Interest in the offer specifically |
| In-session link clicks | Predictive | Active, costly intent |
| Reply rate by intent segment | Predictive | Whether your follow-up is meeting buyers |
How to actually measure these
The honest part: several of these are painful to measure manually.
- Chat engagement density is straightforward. Export the chat log, divide messages by attendees by 10-min windows. A spreadsheet does it.
- Question-intent ratio requires tagging each chat message as buying-shaped or not. A 60-minute webinar with 200 attendees produces around 400 chat messages on average. Tagging each by hand takes hours. This is the metric that breaks manual workflows first.
- Watch-time past the offer requires timestamping the offer reveal, exporting attendance, and filtering. Doable manually but tedious.
- In-session link clicks requires UTM tagging and a click tool. Set it up once, runs forever.
- Reply rate by segment requires intent-scored segmentation upstream. Without segmentation, you're back to averaging.
Most programs stall on question-intent ratio. It's the most predictive metric and the hardest to compute by hand. Either you build a tagging workflow you actually maintain, or you use a layer that reads the chat and scores it automatically. Sponja does the second by default: it scores each attendee 0 to 100, surfaces the question-intent ratio per event, and shows you which specific messages drove the score.
What to do with the data
Each predictive metric points at a different fix:
- Low chat density across events points at wrong audience or wrong format. Revisit who you're inviting and what promise the registration page makes.
- High question-intent ratio, low reply rate points at a broken follow-up. The buyers were there and you failed them. Most fixable case.
- Low past-offer stay rate points at an offer mis-positioned or priced wrong for this audience. Try a different framing before changing the content.
- In-session clicks concentrated in one segment points at that segment being your ICP. Promote into more of them.
- Flat reply rate across intent segments points at segmentation that is not differentiated enough, or follow-up that isn't either. Diagnose the score before redoing the emails.
If you only track two
Most programs are not going to maintain a five-metric dashboard. If you have to pick two:
- Question-intent ratio per event. Tells you if buyers are in the room.
- Reply rate by intent segment. Tells you if your follow-up is meeting them.
These two diagnose roughly 80% of webinar-program problems. Everything else is supporting detail.
If your webinar program is measured on attendance and ranked on registrations, the answer to "did this event drive pipeline" will keep being a guess. Track the five above and the answer becomes specific. Sponja computes them automatically for every Zoom session. Try it free on your next event at app.sponja.ai.
