---
title: Analyze your conversion funnel
description: Find the step where visitors give up before signing up or buying, see which visitors that step loses by device, country, and traffic source, and fix the cause, with an AI marketer doing the analysis in your own analytics.
slug: /use-cases/funnel-analysis
page_type: use-case
lead_prompt: How do I analyze my conversion funnel with AI?
target_prompts:
  - Can an AI marketing agent analyze my conversion funnel?
  - How do I find where visitors drop off before signing up?
  - How do I build a funnel report broken down by traffic source?
  - Do mobile visitors convert worse than desktop visitors on my site?
  - What is funnel analysis, and how do I do one for my website?
entity: Kite
category: AI marketing agent
author: Devon Wells and Kite
date_published: '2026-11-15'
date_modified: '2026-11-15'
related_integrations:
  - /integrations/google-analytics.md
  - /integrations/posthog.md
  - /integrations/mixpanel.md
  - /integrations/amplitude.md
  - /integrations/segment.md
related_playbooks:
  - /how-to/find-where-website-visitors-drop-out-with-google-analytics.md
  - /how-to/find-where-users-drop-out-of-your-funnel-with-posthog.md
  - /how-to/find-where-users-drop-out-of-a-mixpanel-funnel.md
  - /how-to/find-where-users-drop-out-of-an-amplitude-funnel.md
  - /how-to/find-broken-or-leaky-funnels-with-segment-amplitude.md
  - /how-to/find-broken-or-leaky-funnels-with-segment-posthog.md
---

# Analyze your conversion funnel

## Key Takeaways

- Funnel analysis counts how many visitors complete each step on the way to a signup or purchase, so you can see the exact step where most of them give up.
- The step that loses the most people usually hides a narrower problem. Breaking it down by device, country, and traffic source tends to show that one group, such as phone visitors from paid social ads, accounts for most of the loss.
- A step that seems to lose almost everyone is often a tracking fault rather than a real exit, so the events behind each step need checking before anyone changes a page.
- Kite builds the funnel from the events your site records, in its own analytics on a site it hosts or in Google Analytics, PostHog, Mixpanel, or Amplitude. It finds the step and the visitors behind the loss, opens the page to work out why, and builds the fix for your approval.
- After the fix ships, Kite runs the same funnel again and compares it with the period before, so you can see whether more people made it through.

## What is funnel analysis?

Funnel analysis is a way of finding where people give up on the way to becoming a customer. You write down the steps between arriving on your website and converting, then count how many people complete each one in order. For a software company, the steps might be landing on the site, viewing the pricing page, starting the signup form, and finishing it. For a services business, they might be a service page, the contact page, and a submitted form. The step where the count falls hardest is where you're losing people, and it's where a fix pays off first.

Most teams start looking at their funnel when signups feel flat. Traffic looks healthy, yet the number of people who sign up or buy doesn't move. A funnel turns that vague worry into a specific answer, such as "half of the people who start our signup form never finish it." That answer is the start of the work, not the end of it. The overall number rarely tells you what to fix, because the loss usually comes from one kind of visitor. Phone visitors may struggle with a form that works fine on a laptop. Visitors from one ad may land on a page that doesn't match what the ad promised.

That makes funnel analysis a good job for an AI agent. The work is careful and repetitive. You define the steps, check that each one is measured, cut the numbers several ways, read the page where the loss happens, and then do it all again after a change ships. An agent can run every one of those reads, keep the dates and filters straight, and come back with the step, the visitors, the likely reason, and the change worth making.

Kite is an AI marketer. It researches your business, builds a growth strategy, and ships the work that brings you new customers. Because it already knows what your site sells and what a conversion means for your business, it can propose the funnel itself and carry the analysis through to a fix. Here's how it works.

## How Kite analyzes your conversion funnel

Kite starts with what it already knows about your business: what you sell, which pages lead toward a signup or a purchase, and which action counts as a conversion. It turns that into a short list of funnel steps and matches each step to an event your site actually records. On a website Kite builds and hosts, it reads its own built-in analytics. On a site measured elsewhere, it reads your connected Google Analytics, PostHog, Mixpanel, or Amplitude account.

Before it trusts any drop, Kite checks that the events behind each step are arriving, so a broken tracking call doesn't get mistaken for a crowd of people leaving. It then measures the funnel, compares it with the period before, and breaks the weakest step down by device, country, and traffic source until it finds the visitors who account for most of the loss. It opens the page at that step to work out why they leave.

Finally, Kite proposes a fix, builds it once you approve, and runs the same funnel afterward to see whether more people got through.

### Choose the steps your funnel should measure

A funnel is only useful when its steps match how your customers actually buy, so Kite starts by proposing steps that fit your business. A self-serve software product might use a landing page, the pricing page, the signup form, and a created account. A local business might use a service page, the booking page, and a confirmed booking. You can confirm Kite's steps, change them, or name your own.

Each step has to be something your analytics records, either a visit to a specific page or an event such as a form submission. On a site Kite hosts, every site has its own list of recorded events that names each event and marks the site's main conversion goal, so Kite reads that list before it builds anything. The names differ from site to site. One site records a contact form as "contact submitted" and another as "demo requested," and Kite uses your site's own names so the funnel matches what you see.

A funnel on Kite's built-in analytics can have up to eight steps and cover up to 180 days. Kite counts a visitor as converting when they complete the steps in order within 14 days of the first one, which suits most website journeys.

### Make sure every step is measured before trusting a drop

Kite checks the tracking behind each step before it calls any drop a real loss, because a missing event looks exactly like a mass exit. If a developer renames the event behind "account created," that step stops counting on the day the change ships. The funnel then shows nearly everyone leaving at signup, even though people are signing up as usual.

Kite looks at each step's daily volume to see whether the drop was gradual or happened overnight. People rarely stop converting all at once, so a step that fell to zero on one day while the step before it held steady usually points to tracking. Kite also checks for signs that the same person is being counted as two people, which happens when a visitor's anonymous browsing isn't joined to the account they create.

When your events pass through Segment on the way to Amplitude or PostHog, Kite checks that pipeline too. It reads Segment's delivery numbers for the destination and compares them with what arrived, which shows whether your site stopped sending an event, Segment blocked it, or it arrived under a new name. Kite reports each tracking fault with the date it started and the exact fix, so your developers know what to change.

**Related Playbooks**

- [How to find broken or leaky funnels with Segment, Amplitude](/how-to/find-broken-or-leaky-funnels-with-segment-amplitude.md): Check Segment's delivery to Amplitude and the events and identities that arrived, so you know which drops are tracking faults and which are real leaks.
- [How to find broken or leaky funnels with Segment, PostHog](/how-to/find-broken-or-leaky-funnels-with-segment-posthog.md): Run the same checks when Segment feeds PostHog, then diagnose only the steps proven to be measured correctly.

### Find where visitors drop off before signing up

Kite measures how many visitors reach each step and what share of them move on to the next one, which shows you the step where your funnel loses the most people. It reports both the share that continues from each step to the next and the share of the people who entered the funnel and made it to that step.

Kite compares the current period with the one immediately before it, because the comparison separates a step that has always been weak from one that got worse recently. A pricing page that has always lost 70 percent of its visitors calls for a different fix than one that lost 40 percent last month and 70 percent this month. When a step got worse, Kite looks for the date the change started and checks what shipped around then.

Kite also looks at what people did instead of taking the next step. Visitors who leave a pricing page for your documentation are still interested, while visitors who leave it for nowhere at all are gone. In Google Analytics, PostHog, Mixpanel, and Amplitude, Kite can read the next pages and events people reached after they dropped off.

**Related Playbooks**

- [How to find where website visitors drop out with Google Analytics](/how-to/find-where-website-visitors-drop-out-with-google-analytics.md): Build a funnel from the events your Google Analytics property records, find the step and the visitors behind the loss, and check the page to explain it.
- [How to find where users drop out of your funnel with PostHog](/how-to/find-where-users-drop-out-of-your-funnel-with-posthog.md): Build the funnel in your PostHog project and explain the drop with the recordings, paths, and clicks of the people who left.
- [How to find where users drop out of a Mixpanel funnel](/how-to/find-where-users-drop-out-of-a-mixpanel-funnel.md): Check Mixpanel's data quality and each step's volume, then build a funnel report with settings that fit the journey and save it to a dashboard your team can open.
- [How to find where users drop out of an Amplitude funnel](/how-to/find-where-users-drop-out-of-an-amplitude-funnel.md): Chart the journey in Amplitude, confirm each event is still arriving, and read the event streams of people who left the weakest step.

### Break the funnel down by traffic source, device, and country

Breaking the weakest step down by who reaches it usually shows that one group of visitors accounts for most of the loss. An overall drop of 60 percent at the signup form might turn out to be a 30 percent drop on desktop and an 85 percent drop on phones. That changes the fix from "rewrite the form" to "fix the form on phones."

On a site Kite hosts, it can break any funnel down four ways:

- **By device**, to compare desktop, phone, and tablet visitors.
- **By country**, to see whether visitors from one market convert differently.
- **By traffic source**, which groups visitors by how they arrived: search, paid search ads, paid social ads, unpaid social posts, email, AI assistants such as ChatGPT and Perplexity, links from other websites, or typing your address directly.
- **By social platform**, which places each visitor under the platform that first sent them, such as LinkedIn, X, Reddit, or YouTube.

Kite can also narrow a funnel to the visitors from one campaign by its UTM tags, the labels added to a link to say which campaign it came from, or to the steps visitors take on one page. In Google Analytics, PostHog, Mixpanel, and Amplitude, Kite uses the equivalent breakdowns each product offers, such as session source, landing page, plan, or whether a visitor is new or returning.

Kite ranks the groups by how many visitors each one loses at that step, not just by the worst percentage. A 90 percent drop among a few dozen visitors from one country matters less than a 50 percent drop among thousands of phone visitors from search. Ranking by visitors lost points you at the loss that matters most to your results.

### Work out why visitors leave the step

Analytics shows where visitors leave and who they are, but it can't say why, so Kite looks at the page itself. It loads the page and captures screenshots, including at a phone-sized screen when the loss is on mobile, so it sees what visitors see. It measures the page's speed and technical quality with PageSpeed data, and on a site it hosts, it checks the built-in error reports for errors on that page, such as a broken script that stops a form from submitting.

Kite then reads the page the way a visitor would. It looks at the headline, the number of form fields, the call to action, and the price, and compares them with what sent the visitor there, such as the ad or search result. A landing page that converts well for search visitors but poorly for one paid campaign usually doesn't match what the ad promised. When you connect PostHog or Mixpanel with session recordings turned on, Kite can also watch recordings of people who left to see where they hesitated.

Kite separates what the data shows from what it infers. A slow page on phones and a mobile-only drop at the same step line up well, and Kite says how confident it is. A nine-field form with a high abandonment rate is a strong lead, but it isn't proof until a change is measured.

### Fix the step and measure the result

Kite turns its diagnosis into a specific change and builds it once you approve. That might mean shortening a form on phones, rewriting a headline to match the ad that sends traffic to it, moving a call to action higher on the page, or fixing what slows the page down. When the cause is tracking, Kite writes the exact fix for your developers. If Kite works on your site from an imported GitHub repository, it can make the code change itself and opens a pull request when your main branch is protected.

On a website Kite hosts, it publishes the change after you approve it, or runs it as an A/B test with traffic split between the old and new versions. It asks again before making the winning version permanent. On a site measured elsewhere, Kite publishes through a supported content system such as Webflow and compares the same funnel step before and after the change in your connected analytics.

Kite then runs the same funnel with the same steps and compares the new period with the old one, so you can see whether more visitors made it through the step it fixed. If you want to keep watching, ask Kite to run the funnel on a schedule, such as every Monday, and it will tell you when a step gets worse.

## What you get

- **A funnel built for your business.** The steps between arriving and converting, each matched to an event your site records, with the dates the funnel covers.
- **A list of tracking faults.** Any step that isn't measured correctly, with what went wrong, when it started, and the fix.
- **The step that loses the most visitors.** How many people reach each step, what share continues, and how that compares with the period before.
- **The visitors behind the loss.** The device, country, traffic source, or campaign that accounts for most of the drop, ranked by visitors lost.
- **The likely reason.** What Kite found on the page, such as a long form, a slow load on phones, or copy that doesn't match the ad, with how confident it is.
- **The fix and its result.** The change Kite builds for your approval, followed by the same funnel run again to show whether it worked.

When the findings need more room than a chat message, Kite publishes them as a report with the funnel chart and the breakdown tables at a link your team can open.

## How to get started

Tell Kite what you've noticed in plain language. You can work with Kite in the Kite web app or in Slack. You might say:

> "We get decent traffic but signups are flat. Find where people are dropping out and tell me what to fix."

Kite proposes the funnel steps, the period to study, and the period to compare it with. Confirm them or change them, and Kite starts.

If Kite hosts your website, its built-in analytics already record the events it needs, so you don't need to connect anything. If your site is measured elsewhere, connect the analytics product your team uses: Google Analytics, PostHog, Mixpanel, or Amplitude. If Segment sends your events to one of them, connect Segment too so Kite can check the pipeline.

Reading your analytics is routine work, so Kite starts as soon as it has access and reports what it finds in the conversation. It asks for your approval before it publishes a change to your website or starts an experiment.

## Integrations used to analyze your conversion funnel

On a website Kite hosts, Kite measures your funnel with its own built-in analytics, so no other account is needed. For a site measured elsewhere, Kite reads the funnel in the analytics product you already use, and it checks Segment when Segment feeds that product.

### Measure the funnel in your analytics

- [Google Analytics + Kite](/integrations/google-analytics.md): Build funnels from the page views and key events your property records, break them down by device, source, and landing page, and see where people went instead.
- [PostHog + Kite](/integrations/posthog.md): Build the funnel in your PostHog project and explain the drop with the session recordings, paths, and clicks of the people who left.
- [Mixpanel + Kite](/integrations/mixpanel.md): Build funnel reports from your product events, break steps down by property, and read the paths and replays of people who dropped off.
- [Amplitude + Kite](/integrations/amplitude.md): Chart the funnel from your Amplitude events, break it down by platform, country, and traffic source, and mark the date of each change on your charts.

### Check the pipeline that feeds your analytics

- [Segment + Kite](/integrations/segment.md): Read Segment's delivery numbers and destination settings to confirm your events reached Amplitude or PostHog before treating a drop as real.

## Common questions

### Do mobile visitors convert worse than desktop visitors?

On many websites they do, but the size of the gap varies so much that the only useful answer comes from your own data. Phone visitors often browse earlier in their decision and switch devices before buying, and a form or page that works on a laptop can be harder to complete on a small screen. To find out on your site, break your funnel down by device and compare each step. If the gap shows up at one step rather than across the whole funnel, that step probably has a problem phone visitors see and desktop visitors don't.

### What's the difference between a marketing funnel and a conversion funnel?

A marketing funnel describes the broad stages a buyer moves through, from first hearing about you to considering you and finally buying. A conversion funnel is a measured version of one part of that journey on your website, made of specific steps such as the pricing page, the signup form, and a created account. For example, a marketing funnel might say a buyer first sees your LinkedIn post, later reads a comparison page, and then books a demo. A conversion funnel counts how many people who landed on that comparison page went on to open the booking page and confirm a demo.

### What is a good funnel conversion rate?

There's no single good number, because conversion rates depend on your price, your audience, how you get traffic, and how many steps your funnel has. The most useful comparisons are with your own funnel: this period against the last one, and one group of visitors against another. A step that converts 40 percent of desktop visitors and 15 percent of phone visitors tells you more than any industry average.

### Does Kite need Google Analytics if it hosts my website?

No. Kite's built-in analytics record visitors, pages, traffic sources, form submissions, calls to action, and the other events on the sites it hosts, and Kite builds funnels from them directly. Connect Google Analytics, PostHog, Mixpanel, or Amplitude when your website runs somewhere else or when your team already relies on one of them.

### Can Kite check my funnel every week?

Yes. Ask Kite to run the same funnel on a schedule, and each run compares the latest period with the one before. When a step suddenly gets worse, Kite tells you which step changed, which visitors it affects, and what it thinks caused it. Results arrive in your Slack channel, by email from your team's Kite address, or as a new conversation in the Kite web app.
