Funnel Chart Maker
Two columns — a stage name and a value — become a funnel, sorted widest to narrowest. Funnels are a familiar way to show a sequential process with attrition at each step, and the shape carries a couple of assumptions worth being explicit about.
How to use it
- Paste two columns: Stage and Value, with a header row.
- Confirm the stages represent a genuine sequence rather than unrelated categories.
- Read the stage-to-stage drop rather than only the overall rate.
- Choose a theme and export a high-resolution PNG.
What the shape assumes
A funnel encodes a sequential process where every entity at one stage either advances to the next or drops out. Visitors become signups become trials become customers. Each stage is a strict subset of the one before it.
Stages are sorted by value descending, which matches that assumption. If your data does not sort naturally into descending order, that is a signal worth attending to rather than a display problem: either the stages are not sequential, or entities are entering partway through, or the counts are measured over inconsistent windows.
The last of those is the most common data error in funnel analysis. If the top stage counts visitors this month and the bottom counts customers this month, some of those customers came from last month visitors and some of this month visitors have not converted yet. For a process with a long cycle, this cohort mismatch can shift the apparent conversion rate substantially. Tracking a fixed cohort through the stages avoids it and is usually more work.
Stage-to-stage against overall conversion
Two different rates get called conversion and they answer different questions. Overall conversion is the bottom divided by the top. Stage-to-stage conversion is each stage divided by the one before it.
Overall conversion is the headline and stage-to-stage is where the action is. A funnel of 10,000 visitors, 2,000 signups, 800 trials, and 200 customers is 2 percent overall, which tells you nothing about where to work. The stage rates are 20 percent, 40 percent, and 25 percent, which points directly at the first step.
The compounding is worth internalising: overall conversion is the product of the stage rates, so improving one stage multiplies through the whole funnel. Taking that first step from 20 to 25 percent, a modest change, raises overall conversion from 2 to 2.5 percent — a 25 percent increase in customers with no change anywhere else. This is why funnel analysis is usually worth more than top-of-funnel spend.
Where the visual misleads
The width of each band is proportional to its value, which is the correct encoding, but the funnel shape adds a suggestion the data may not support.
The tapering silhouette implies a smooth, natural narrowing, when real funnels usually have one or two catastrophic steps and several harmless ones. A stage losing 80 percent and a stage losing 5 percent both appear as a taper, and the reader has to compute the difference rather than see it.
Vertical spacing implies equal effort or equal time between stages, which is rarely true. The gap between visiting a page and signing up may be seconds; the gap between a trial and a purchase may be weeks with several people involved. The chart gives them the same visual weight.
And a wide top band draws the eye to the largest number, which is usually the least actionable one. The interesting information is in the ratios between adjacent bands.
A plain bar chart of the stage-to-stage conversion rates makes the problem stage obvious at a glance and loses the intuitive shape. Both together — the funnel for the volumes, a bar chart for the rates — is the version that answers both questions without asking the reader to do arithmetic.
When the process is not a funnel
Two situations break the format. The first is a process where entities can re-enter, skip stages, or move backward. A trial user who lapses and returns is not a strict subset of the prior stage, and a funnel will either double-count or drop them.
The second is branching. If users can reach the outcome by several routes — self-serve, sales-assisted, partner — a single funnel averages paths with very different rates and hides the fact that one route works and another does not. Separate funnels per path, or a Sankey diagram, represent it honestly.
The Sankey case is worth flagging specifically: it handles branching, merging, and drop-out as explicit flows, so a process with real structure is better served there than by forcing it into a linear taper.
At a glance
| Input format | Two columns: Stage and Value |
|---|---|
| Sorting | Descending by value, widest at the top |
| Assumption | Each stage is a subset of the one before it |
| Minimum rows | Two |
| Export | PNG at 2× display resolution |
Frequently asked questions
What columns does it need?
Two: a stage name and a value, with at least two stages.
Why are stages sorted automatically?
Because a funnel assumes each stage is a subset of the previous one. If your data does not sort descending, the process may not be a funnel.
Which conversion rate should I look at?
Stage-to-stage. Overall conversion is the product of the stage rates and does not tell you which step to fix.
How does the funnel shape mislead?
The taper makes an 80 percent loss and a 5 percent loss look similar, and equal vertical spacing implies stages take equal time.
What if users can take different paths?
A single funnel averages paths with different rates. Use separate funnels per path, or a Sankey diagram to show the branching.
Read more
Comparison and composition — People judge length against a common baseline far better than they judge angle or area. Most charting advice follows from that one fact.