Analytics and Tracking

Funnel Analysis

Also called Drop-off analysis, funnel exploration

Step-by-step measurement of how many people survive each stage of a journey towards a conversion.

Quick facts: Funnel Analysis

Category
Analytics and Tracking
Also called
Drop-off analysis, funnel exploration
Level
Intermediate
Affects
Conversion rate, where budget is spent, roadmap priorities
Where to see it
GA4 funnel exploration, Looker Studio, ecommerce platform reports
In this article4
  1. How funnel analysis works
  2. Why funnel analysis matters
  3. Common mistakes with funnel analysis
  4. What to do about it

How funnel analysis works

You define an ordered set of steps that lead to something valuable — viewed the service page, opened the form, started filling it, submitted it — and the tool counts how many people reached each one. The result is a staircase, and the height of each drop tells you what the journey costs at that point.

Two settings change the picture more than most people realise. The first is whether steps must happen in sequence or merely at some point in the session: a strict funnel is precise but discards anyone who wandered, while an open one is forgiving and can flatter you. The second is the window. A journey allowed a day looks very different from the same journey allowed a fortnight, and for considered purchases the short window will show a collapse that is really just people taking their time.

Everything downstream depends on the steps being tracked correctly. A step that fires on the wrong page, or a form event that never fires on mobile, produces a dramatic drop-off that exists only in the tracking. Confirm the events before you interpret the shape.

Why funnel analysis matters

A single conversion rate tells you something is wrong without telling you where. The funnel converts that into a location, and a location is something you can brief work against — the page, the form, the delivery step or the payment options.

It also stops money being spent in the wrong place. If most visitors never reach the enquiry form, buying more traffic multiplies the same loss, and improving the step where they leave is much cheaper than replacing them at the top. In markets like Nepal, where payment options and delivery expectations vary, the drop-off often sits at a step nobody in the office thinks about because it works fine on their own device and connection.

Common mistakes with funnel analysis

The first is fixing the biggest drop rather than the most valuable one. The largest fall is often the earliest, where most of the audience was never going to buy anyway; a smaller improvement close to the money can be worth far more.

The second is defining steps around your internal process instead of the customer’s. People do not move in a straight line — they leave, compare, come back on another device and return through a different channel — and a funnel that assumes an orderly path reports abandonment that never happened.

The third is trusting a funnel nobody has tested, and the fourth is reading it in aggregate only. A journey that works on desktop and fails on a phone shows up as a modest overall drop rather than the serious problem it is.

What to do about it

Build the funnel around what the customer is trying to do, keep it to a handful of meaningful steps, and verify each event fires on real devices before you draw conclusions. Then look at the drops in order of value rather than size, and pick one to work on.

Take that step apart with tools that explain behaviour rather than count it: a session recording, a scroll map, or form analytics if a form is involved. The funnel says where; those say why. Then split the same funnel by device and by traffic source, because segmentation of results regularly turns a mild average into an obvious fault in one group. Fix, measure the same steps again, and move to the next drop.

Do and do not

Do

  • Verify every step's event fires on real devices
  • Rank drops by value, not by size
  • Split the funnel by device and traffic source

Do not

  • Model the funnel on your internal process
  • Use a window shorter than the buying cycle
  • Buy more traffic before fixing the leak

Questions people ask about this

How many steps should a funnel have?

Few enough that each one represents a real decision by the customer. Four or five meaningful steps are usually more useful than a dozen technical ones, because every extra step adds a place for tracking to fail and makes the report harder to act on. If a step cannot be fixed independently, it probably does not need to be a step.

Why do my funnel numbers not match my ad platform's?

Because they count different things over different periods. An ad platform attributes conversions back to the click that caused them, sometimes days earlier, while a funnel in your analytics counts steps within a session or a chosen window. Consent choices, blocked scripts and cross-device journeys widen the gap further. Use each for its own job rather than reconciling them.

Should I fix the biggest drop-off first?

Not automatically. The biggest fall is usually near the top, where much of the audience was only browsing, so recovering a share of it may be worth little. Multiply each step's drop by what the people at that stage are worth, and work on whichever produces the largest gain. Late-stage steps often win on that measure.

Related terms

Found this useful?

Share it, or ask an AI to summarise it

Back to the glossary

Knowing the term is the easy part

Applying it to your own site and budget is the work. Book a call and I will tell you what actually applies to you.