Product Funnels Without Fooling Yourself With Pretty Percentages

The problem

Product funnels are one of those tools that look too obvious to get wrong.

You think:

• step 1
• step 2
• step 3

and done.

But a lot of bad product reading starts exactly there.

Because funnels create a dangerous feeling of clarity.

There is a number.

There is drop-off by step.

There is a percentage.

So it looks scientific.

But without care, it becomes only an elegant way to fool yourself.

Mental model

Think about it like this:

a good funnel does not exist to tell a pretty story. It exists to show where the journey loses traction in a way you can act on.

If it does not help answer:

• where is the friction?
• for whom does that friction happen?
• what is worth investigating first?

then it is probably just a pipeline drawing with numbers on top.

Breaking the problem down

A funnel step needs to represent real behavior

This is the first filter.

A good step is something observable and semantically clear.

Example:

• signup started
• form submitted
• email confirmed
• onboarding completed

A weak step is usually vague or mixes intent with technical detail:

• entered the experience
• engaged with the screen
• interacted with the journey

If nobody can explain precisely what the step means, the funnel already started weak.

Percentage without volume invites illusion

Classic example:

• 50% conversion

That sounds great.

But 50% of what?

• 2 people?
• 20?
• 20,000?

Absolute volume changes the whole interpretation.

That is why a serious funnel almost always needs to show:

• rate
• volume by step

Without both, you risk optimizing noise.

Mixing different audiences destroys the diagnosis

This is another very common mistake.

You put into the same funnel:

• new users
• returning users
• paid traffic
• organic traffic
• small companies
• enterprise accounts

and then try to draw one average conclusion.

That average often hides the actual problem.

Maybe the funnel is healthy for one group and bad for another.

But when everything is aggregated, the team only sees one lukewarm number that explains very little.

Not every drop between steps is a problem

Some drop-off is part of the nature of the flow.

If 100% of the people who visit the homepage do not continue to purchase, that alone does not say much.

The better question is:

• is the drop happening where we expected?
• did it change after the release?
• should this point in the journey convert better for this audience?

A funnel without a mental benchmark or historical comparison turns into a hunt for any number that looks ugly.

Too many steps also hurt

Some funnels become a trail of breadcrumbs.

Every micro-click becomes a step.

The result:

• slow reading
• too many drop-offs
• little useful signal

You need enough granularity to locate friction.

Not infinite granularity to simulate precision.

Simple example

Imagine an account creation flow.

Bad funnel:

• visited landing page
• clicked main button
• opened modal
• typed name
• typed email
• typed password
• clicked continue
• opened confirmation
• confirmed email

That looks detailed.

But it is bad for making quick decisions.

Better funnel:

• signup started
• signup submitted
• email confirmed
• onboarding completed

Now it is easier to ask:

• is the biggest loss before submission?
• is it in the confirmation step?
• is it in onboarding?

If you need more depth, you investigate later.

But the main funnel stays readable.

What usually goes wrong

• Treating any sequence of events like a useful funnel.
• Showing rate without volume.
• Mixing incompatible segments.
• Creating too many steps and losing readability.
• Reading every drop-off as a problem without context.
• Building a pretty funnel that nobody can turn into action.

How someone more senior thinks

A more mature person usually looks at the funnel with a few simple questions:

• does this step represent behavior or only interface detail?
• does this need segmentation before any conclusion?
• which drop-off would really change the team’s priority?
• if I see this number getting worse tomorrow, do I know what to investigate first?

That way of looking avoids two traps:

1. the ornamental funnel
2. the hyper-granular, useless funnel

Interview angle

This shows up in questions like:

• “how would you measure this journey?”
• “how would you find where the user drops off?”
• “how would you know which part of onboarding to focus on?”

The interviewer wants to see whether you:

• think in terms of the real journey
• do not get dazzled by percentages without context
• know how to segment and prioritize the reading

Weak answer:

I would build a funnel with all the steps and see where conversion drops the most.

Strong answer:

I would choose a few semantically strong steps in the journey, show rate together with volume, and separate segments that may behave very differently. The point is not to draw the whole flow in funnel form. It is to locate where there is meaningful friction worth investigating.

Closing

A good funnel does not try to impress.

It tries to simplify without lying.

When that works, it becomes a great tool for locating problems.

When it does not, it becomes one more pretty percentage the team looks at, comments on, and forgets.