9 Funnel Analysis Mistakes That Lead to the Wrong Fix

Short answer: funnels mislead when they include steps people can skip, use a time window that is too short, blend devices and audiences, or are read without looking at the sessions behind the numbers. The step with the largest percentage drop is not always the step to fix. Check the absolute numbers, split by device, and open the recordings before deciding.

Table of Contents

1. The Nine Mistakes
2. A Review Checklist
3. Reading the Result Properly
4. Key Takeaways

The Nine Mistakes

1. Including optional steps

A funnel of Product → Cart page → Checkout breaks if your site lets shoppers go straight to checkout from a mini-cart. Everyone who skipped the cart page counts as dropping out there.

Fix: include only steps every converting user must pass through. Check by walking the flow yourself, every route.

2. Too short a time window

A buyer who views a product today and returns to buy on Thursday is a failure in a one-session funnel.

Fix: match the window to how long decisions really take. Minutes for a food order. Days or weeks for furniture or B2B software.

3. Blending devices

A step converting at 70% on desktop and 25% on mobile averages to something unremarkable.

Fix: split by device every time.

4. Blending new and returning visitors

Returning visitors know where they are going. New visitors are still deciding whether to trust you.

Fix: view them separately, and by traffic source for paid campaigns.

5. Fixing the biggest percentage drop

The first step of most funnels loses the most people, because most visitors were never going to buy. Meanwhile a smaller drop at payment involves people who had decided to pay.

Fix: weigh each drop by intent and recoverability. Late-stage losses are usually cheaper to recover.

6. Trusting click events as success

A step defined as "clicked Pay" counts failed payments as progress.

Fix: define outcome steps on confirmed success, from the server response or confirmation page (setting up goals).

7. Reading a single day or week

Small samples and weekday effects produce swings that look like regressions.

Fix: use at least two full weeks and compare with a matching period (comparing periods).

8. Ignoring tracking faults

A step shows zero because the event stopped firing after a redesign. Or the count doubled because the event fires twice.

Fix: when a step changes abruptly, check the tracking before the product. Complete the flow yourself and confirm each step registers once.

9. Stopping at the number

A funnel says where people leave. It cannot say why. Teams then guess, and redesign the step based on the guess.

Fix: open the sessions. In DeepSync, clicking a step's drop-off number goes straight to the sessions that stopped there. The docs describe the funnel chart as an index into sessions, not the finding itself (funnels).

A Review Checklist

• [ ] Every step is required for conversion
• [ ] Outcome steps use confirmed success
• [ ] The time window fits the buying cycle
• [ ] At least two full weeks of data
• [ ] Split by device
• [ ] Split by new and returning
• [ ] Each event fires once, verified by hand
• [ ] Absolute numbers shown beside percentages
• [ ] Ten to fifteen drop-off sessions watched for the chosen step

Reading the Result Properly

Take an illustrative funnel:

The biggest drop is product to cart, 90%. It is tempting to start there.

Look again. Of 900 people who submitted payment, 180 did not get a confirmed order. Those people entered their details and pressed pay. Something failed. That is very likely a technical fault, and recovering even half adds 90 orders, more than a twelve percent increase, without persuading a single extra visitor.

So the order of investigation is often the reverse of the size of the drop:

1. Late steps with failures, where intent is certain.
2. Steps with a device gap.
3. Steps that changed recently.
4. Early steps, where improvement needs persuasion.

Then watch the sessions for whichever you chose, and check JavaScript errors on that page.

Key Takeaways

• Only required steps, defined on confirmed outcomes.
• Split by device and visitor type.
• Prioritise by intent, not by the size of the percentage.
• Open the sessions before choosing a fix.

Conclusion

A funnel is a map of where to look. Built carefully and read with the sessions beside it, it points at the right fix. Built carelessly, it sends you to redesign the wrong page.

Build a funnel in DeepSync.