A funnel is about movement. Not how many records are in each stage, which a pivot answers, but how they travel between stages and how long it takes.
Two questions it answers
Velocity. How long records spend at each step. Where things sit.
Transitions. How many move from one step to the next, and how many do not.
They are different questions and they produce different conclusions from the same data. A stage with a high drop-off is a qualification problem. A stage where everything eventually passes but takes three weeks is a capacity problem. Both look like "a problem at stage four" until you separate them.
What you set
A record type and its pipeline, since the stages come from the pipeline rather than from a list you maintain here.
A window, the period being examined.
Which of the two questions you are asking.
The stages are read from the live pipeline, so a pipeline that gains a stage gains it here without anybody editing the funnel.
Windows
The window decides which records the funnel considers, and it is the setting most likely to produce a misleading result.
Shorter than your cycle and the funnel describes an arbitrary slice: records that entered before the window are invisible, so early stages look empty and late ones look efficient.
Much longer than your cycle and it averages together periods that were genuinely different.
The rule of thumb: two to three times your typical cycle length. If a deal takes a month, look at a quarter.
Reading velocity
Time at each stage, which is the more actionable of the two and the less used.
A long first stage usually means qualification is happening late or not at all.
A long final stage is frequently procurement or legal rather than anything your team controls, and is worth separating before anybody is asked to improve it.
One stage much longer than the rest is where to look, and it is often a handover between two people rather than the work itself.
Reading transitions
Where records stop.
A big drop early is either honest qualification or a lead source problem, and which one it is depends on whether the drop is getting worse.
A big drop late is expensive, because everything that reached that stage consumed real effort.
No drop anywhere is not good news. It usually means stages are being updated in bulk at the end rather than as work happens, so the funnel is describing the data entry rather than the business.
Which question first
If you can only build one, build velocity.
Transitions tell you where records stop, which is the question everybody asks. Velocity tells you where records wait, which is usually the question that has an answer somebody can act on this week.
A drop-off frequently reflects the market, the pricing, or the quality of the lead. Time sitting at a stage almost always reflects something the business controls, and that makes it the more useful place to start.
The records behind a step
Any step opens into the records that make it up.
This is the most useful thing on the surface and the least used. A drop-off is an abstraction; eleven named deals that stalled at the same stage is a conversation with the person who owns them.
Look before concluding. A frequent outcome is that the drop is three unusual records rather than a pattern.
The records drawer
Every step opens into the actual records at that step, with enough detail to act.
This is where a funnel stops being a chart and becomes work. A drop-off is a percentage; the eleven deals behind it have owners, values and dates, and one of them probably explains the other ten.
Saving
A funnel saves as a first-class widget and goes on a board like anything else, re-running against live data.
Worth doing for the one or two funnels that describe how your business actually works. Worth not doing for exploratory ones: a board with six funnels is a board nobody reads.
Where the stages come from
The steps are the pipeline's own stages, read live rather than configured here.
Two consequences. Adding a stage to the pipeline adds it to every funnel over that pipeline, with no editing. And renaming a stage renames it here, so a funnel never disagrees with the board people actually work.
That is the unified-database property doing real work: a standalone funnel tool would need the stage list re-entered, and it would be wrong the first time somebody changed the pipeline.
Comparing two funnels
The most useful thing you can do with this engine, and the least obvious.
Run the same funnel over two windows, or over two pipelines, and compare. A single funnel tells you where records stop; two tell you whether that is changing, which is the question anybody senior will actually ask.
Keep everything else identical when you do. A funnel over a different window and a different pipeline is two changes and explains nothing.
Honest limits
It describes, it does not explain. A funnel shows where records stop. Why they stop is in the conversations, and Meeting Intelligence is a better source for that than any chart.
It depends on stages being current. A funnel over a pipeline that people update weekly measures the update habit as much as the work.
Short windows mislead. A funnel over two weeks of a three-month sales cycle is describing an arbitrary slice, and it will move alarmingly for reasons that are not real.
What a healthy funnel looks like
Not a smooth taper. A real funnel has one or two places where a lot falls away, and those are usually the stages that are doing actual qualification.
The shape to worry about is the opposite: an even, gentle decline across every stage. That normally means the stages are not decision points, they are a checklist somebody advances records through, and the funnel is measuring administration.
When to use something else
A pivot for counts within stages rather than movement between them.
An analysis for one measure over time, which is a simpler question and a clearer chart.
Did this answer your question?
No, ask a person