How to Build a HockeyStack Funnel Report That Explains Lost Deals
A HockeyStack funnel report shows conversion between defined steps — MQL to SQL to opportunity to closed-won — and lets you break each step by channel, persona, campaign, or any property in your data model. Build it under Reports → Funnels → New Funnel.
The default version of this report is useless. A straight funnel with four steps tells you 100% of MQLs became 12% of customers, which you already knew. The value comes from the breakdowns and from how you define each step.
How do you define funnel steps in HockeyStack?
Steps are event-based or CRM-property-based. You choose, and the choice changes the story.
- Event-based:
form_submiton demo form →meeting_booked→opportunity_created. Captures the buyer’s actual path. - CRM stage-based:
MQL date is not empty→SQL date is not empty→Opportunity stage = Proposal. Captures your team’s labels.
Use CRM stages for the top of the funnel where sales owns qualification, and event-based steps for product or PLG motion. Mixing them in one report is fine, but be explicit about it or someone will argue about “MQLs” for an hour without agreeing on a definition.
What breakdowns actually reveal leakage?
Drop-off by channel is the start, not the end. The dimensions that produce decisions:
| Breakdown | Question it answers |
|---|---|
| Source (first touch) | Does content-sourced pipeline close worse than paid? |
| Persona / job title | Are we qualifying the wrong roles? |
| Deal size band | Do small deals stall at proposal? |
| Rep / owner | Is one rep’s MQL-to-SQL rate half the team’s? |
| Number of touchpoints before demo | Is 3 touches enough or do we need 7? |
That last one is underused. HockeyStack will show you average touchpoints for won deals vs lost, and the gap is often the most actionable number in the account.
How do you build the report step by step?
- Go to Reports → Funnels → Create Funnel.
- Add steps in order. Start with three; four-step funnels with tiny counts at the bottom produce no signal.
- Set the date range to a fixed window (last quarter), not a rolling 30 days, if your deal cycle exceeds a month — a rolling window pulls in deals that have not had time to close.
- Set the starting population filter: exclude internal domains, exclude competitor domains, exclude known bots.
- Add breakdowns one at a time. Two breakdowns at once on a small dataset creates 40 rows with two leads each.
- Save as a shared report and set a weekly schedule if anyone will look at it.
The minimum volume rule
Nothing below about 30 records per bucket is a story. If a channel has 8 MQLs and 1 closed deal, that is not a 12.5% close rate, that is noise. HockeyStack shows counts, so read them before drawing conclusions. This is where most funnel analysis goes wrong and where confident bad decisions get made.
What does a good funnel report actually show?
An example from a mid-market SaaS company, illustrative but typical:
- Content-sourced MQLs convert to SQL at 34%, paid at 41%.
- But content-sourced SQLs close at 28%, paid at 12%.
- Net: content produces more revenue per MQL despite the lower top-of-funnel rate.
A report that only showed the first number would have led to cutting content spend. The second number flips the decision. Build your funnel deep enough to see both.
What does HockeyStack not tell you?
It tells you where deals stop, not why the buyer decided that. For that you need win/loss calls, and HockeyStack’s closed-lost reason field only holds whatever your reps typed, which is often “no budget” for every loss regardless of truth.
Fix the CRM hygiene first. A funnel report on top of lazy closed-lost reasons gives you confident charts about nothing.
Second limitation: multi-touch attribution models differ and HockeyStack supports several. Pick one (position-based is a reasonable default for B2B) and stop switching. Switching models between meetings is how attribution gets a reputation for being untrustworthy.
Key Takeaways
- Build HockeyStack funnels under Reports → Funnels, choosing event-based or CRM-stage-based steps deliberately.
- Break down by source, persona, deal size, and rep — channel alone rarely produces a decision.
- Use fixed date windows for deal cycles longer than 30 days.
- Ignore buckets under about 30 records; the percentages are noise.
- Attribution is only as good as your closed-lost reason data — clean the CRM first.
