Website Conversion Analyzer

What This Analyser Does, And Who It Is For

The analyser walks your funnel one stage at a time. Sessions to page engagement, engagement to form start, form start to submission, submission to a qualified enquiry, qualified enquiry to a booked call. At each step it reports two things: the percentage that moved forward, and the raw count. The second number is the one that matters, and it is the one most conversion reports bury.

It is built for people selling something with a human step in it: consultancies, agencies, trades, clinics, professional services, anyone whose site exists to produce a conversation rather than a checkout. If you run an ecommerce store, the mechanics are similar but the diagnosis is different and a cart-specific tool will serve you better.

There is one thing this page will not do, and it is worth saying at the top. It will not give you a benchmark table of good conversion rates to measure yourself against. The previous version of this page had one, and removing it is the single biggest improvement made here. The next section explains why, with the arithmetic.

Worked Example One: Why The Benchmark Table Had To Go

A conversion rate is a fraction. Outcomes over sessions. You control the denominator, which means you can improve the fraction without improving anything real.

Here are three scenarios. Same business, same offer, same close rate.

Read the first two rows against any published benchmark table and Site B is a well-optimised site while Site A is broken. Site B converts five times better on form fills and more than eight times better on booked calls. Both sites book twenty calls a month. Identical revenue.

The third row is the part that should end the argument. Site A has changed nothing about its funnel. It has switched off a traffic source that was sending 8,800 sessions a month and no enquiries. Its form conversion rate goes from 2.00% to 16.67%, which is comfortably "excellent" on every benchmark table published anywhere, and it books exactly the same twenty calls. Nobody's diary changed. The number went up by a factor of eight.

The reverse is just as easy to arrange by accident. Rank for one broad informational query, get 30,000 new sessions, and your conversion rate collapses while your booked calls rise. If your reporting is a percentage, you have just been punished for a good month.

This is why the analyser puts the count first and the ratio second, and why it asks you for a stage-by-stage denominator rather than a single site-wide one. The rate is only meaningful against traffic you actually intended to attract.

Worked Example Two: When Fixing The Form Makes Things Worse

"Reduce your form fields" is the most repeated conversion advice there is, and it is true in the narrow sense that shorter forms get submitted more often. Here is what happens when you follow it and keep measuring past the submission.

Take 1,000 visitors reaching a contact page, and assume the same 40% of qualified enquiries end up booking a call in both cases.

Submissions rose by 125%. Booked calls fell by 10%. The seven fields were not friction for the sake of friction; they were doing qualification work, and removing them moved that work onto whoever now has to read fifty extra enquiries a month and reply to the ones worth replying to.

Whether this trade is good depends entirely on what happens after the submission. If you have an automated qualification step, a shorter form is clearly right: let more people in, sort them with a system rather than a form field. That is what a lead qualification bot is actually for, and it is the reason the "shorter is better" advice keeps getting repeated by people who have one. If your process is you, reading an inbox between jobs, the long form was doing you a favour.

The analyser cannot make that call for you. What it can do is stop you declaring victory at the submission step, which is where almost every conversion report ends.

The Method: What Is Actually Measured

Five stages, each with an explicit denominator:

1. Sessions reaching the page. Segmented by source, because a site-wide figure hides the problem shown above.
2. Engagement. Did anything happen beyond arrival: a scroll past the first screen, a click, a video start.
3. Form or booking start. First field focused, or the booking widget opened.
4. Submission. Completed and accepted, not merely attempted. Validation failures count as a loss here, and they are more common than people expect.
5. Qualified and booked. Reached whatever your own definition of a real prospect is, and put something in a diary.

The gap between stages 3 and 4 is worth isolating rather than folding into a single form conversion number. Someone who focused a field and left has a different problem from someone who never touched the form. The first is a form problem. The second is a page problem, or a traffic problem, and no amount of field-trimming will touch it.

Where The "Three Second" Rule Actually Comes From

Page speed belongs in a conversion audit. The number everybody quotes for it does not survive being looked up, and since the earlier version of this page quoted it, it gets corrected here.

The claim is normally written as "53% of visitors abandon a page that takes over 3 seconds to load." The source is a Google Ad Manager post from 8 September 2016, "The need for mobile speed" by Alex Shellhammer and Juliette Neel. Its actual wording is "53% of visits are likely to be abandoned if pages take longer than 3 seconds to load," and its own footnote attributes it to aggregated Google Analytics data from mobile web sites that opted into benchmark sharing, n=3,700, global, March 2016.

Three things change once you read the footnote. It is visits, not visitors. It is mobile web only. And it is a decade old, measured against a web where, as the same post says, "the average load time for mobile sites is 19 seconds" over 3G. A threshold calibrated against a 19-second baseline is not a threshold you should be designing against now.

The number to manage instead is documented and current. Google's Largest Contentful Paint guidance, last updated 4 September 2025, states that "sites should strive to have Largest Contentful Paint of 2.5 seconds or less," measured at the 75th percentile of page loads and segmented across mobile and desktop, with anything over 4.0 seconds classed as poor. That is a render metric with a defined measurement point and a defined percentile, which makes it something you can hold a supplier to. "Three seconds" is not.

Everything else in the old version of this section has been deleted rather than rewritten: the claim that removing a form field lifts submissions by a fixed percentage, that testimonials near a call to action lift conversions by a stated range, that exit-intent popups recover a stated share of leaving visitors, and that eight in ten conversions happen after five or more touches. Each was stated as fact with no source attached, and none of them survived an attempt to find one. The follow-up claim in particular circulates everywhere without a study behind it. Follow-up genuinely does matter, and if you want a number for your own business the follow-up revenue calculator will produce one from your figures rather than from someone else's anecdote.

Common Mistakes

Reporting the rate without the count. Covered above, and it is worth restating because it is the mistake that produces confidently wrong decisions rather than merely useless reports. Always put the outcome count next to the percentage. If the two disagree about the direction of travel, trust the count.

Optimising the step you can see. Form pages are easy to test and easy to change, so they absorb attention out of proportion to their share of the loss. If nine in ten visitors never reach the form, the form is not your problem, and a year of button-colour tests will confirm that expensively.

Treating validation failures as user error. A submission that was attempted and rejected is a loss you caused. Postcode fields that reject valid formats, phone fields that reject a leading zero, required fields that only announce themselves after submission: these show up as low form conversion and get diagnosed as low intent.

A/B testing at volumes that cannot support it. With twenty booked calls a month, a variant that produces twenty-four is not a win, it is a normal month. Testing needs enough events to distinguish signal from noise, and most service businesses do not have them at the booking step. Test further up the funnel where the counts are larger, and change the booking step on judgement and qualitative evidence instead.

Adding a chatbot before writing the answers. A bot is a delivery mechanism for answers that already exist. If they do not exist, it will improvise them, and you have automated the production of wrong information. The comparison in chatbot versus contact form sets out when each is the right choice, and it is not always the bot.

Ignoring what happens in the hour after submission. This is the largest unmeasured loss on most sites and it is invisible to any on-page analysis, because it happens in an inbox. A perfect funnel feeding a reply that arrives on Tuesday is a broken funnel.

Frequently Asked Questions

So what is a good conversion rate?

The only defensible answer is: higher than yours was last quarter, on the same traffic mix, with the outcome count rising too. Any published figure compares your denominator to a stranger's denominator, and the arithmetic above shows what that comparison is worth. If you need a target, set it in booked calls per month and let the ratio be whatever it turns out to be.

Why does the analyser want traffic split by source?

Because a mixed denominator averages away the thing you were looking for. Paid traffic to a landing page and organic traffic to a blog post have no business being in the same fraction. Splitting them is usually the point at which the actual problem becomes obvious, and it often turns out to be one source doing all the damage.

Does page speed really affect conversions?

Yes, though the honest framing is that it is a hygiene factor rather than a lever. A slow page loses people before they see the offer, so speed sets a ceiling on everything else. Getting Largest Contentful Paint under 2.5 seconds at the 75th percentile is a well-defined job with a clear finish line. Chasing further improvements past that point rarely pays as well as fixing the next stage of the funnel.

Should I shorten my forms or not?

It depends on whether anything qualifies enquiries after the form. With automated qualification, shorten the form. Without it, the fields are your qualification, and removing them converts a page problem into a labour problem. Work out which situation you are in before taking the advice.

Why does this tool ask for my email?

Because the output is a staged report rather than a single number, and it is sent rather than rendered on the page. That costs visitors, and I would rather say so than pretend the gate is a feature. The calculators and developer utilities on this site are ungated for exactly that reason. If you only want the method, the five stages above are the whole method.

Can I run this on a competitor's site?

Only in part. The external signals are readable, so page speed, the number of steps to a form and the presence of a booking path are all observable. The counts are not, and the counts are where the diagnosis lives. Competitor conversion figures quoted in public should be read the same way as the benchmark table: as a fraction whose denominator you cannot see.

What This Tool Cannot See, And Who Fixes That

The analyser reads the funnel up to the point where a person takes over. It can tell you that submissions rose by 125% and booked calls fell by 10%. It cannot tell you which of those fifty extra enquiries was worth a reply, because that information only exists after someone has replied to them, and on most sites it is never written down anywhere a report can reach.

Closing that gap is a CRM job rather than a website job. Every enquiry needs to arrive with its source and its landing page attached, get a status that a human actually updates, and end in a recorded outcome, so that next quarter you can ask which page produces work rather than which page produces submissions. That is the piece I build most often: enquiry capture wired into GoHighLevel or a similar system, routing and follow-up that runs without anyone remembering, and reporting that closes the loop back to the page. Where a qualification step in front of the human makes sense, that becomes a chatbot or a form flow with logic in it, built against your own criteria rather than a template.

Two other tools here pair with this one. The follow-up revenue calculator prices the gap between submission and reply, which is usually the largest single loss in the whole funnel. The website automation audit looks at the structural gaps around the funnel rather than in it.

If your report comes back and the ordering is not obvious, that conversation is free, and I would rather look at your actual counts than guess from a benchmark.