Which landing page makes the sale? Page by page tracking
Short answer: To find which landing page makes the sale, tag each ad link so the page and the ad are known, count leads in Google Analytics 4 (Google's free website statistics tool) by landing page, and copy the landing page into each lead's record so you can follow it to a booked job or payment. Compare pages by what happens after the form, not by clicks, and do not pick a winner until each page has enough visitors. For many small sites, a spreadsheet and a hidden form field do the job.
Two landing pages both get clicks. One page gets form fills that never turn into jobs, the other fewer fills that do. If you judge only by the first number, you back the wrong page. The fix is to follow each lead from page to sale.
I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.
The fastest answer: give every landing page its own address, tag every ad that points to it, record the landing page and the ad on each lead, and compare pages by booked jobs and revenue per visitor once you have enough traffic. Clicks and form fills are useful early signals. Sales are the final test. This page shows the steps, with a worked example, a rule for how much traffic you need, and where tools fit.
It is written for owners who run paid ads to more than one page, or who have an old page and a new one and want to know which earns more. I also cover where Ploy, an AI website builder with built in analytics, helps, and where a dedicated landing page tool is the better pick.
This page may contain affiliate links from which Blur Studio earns commissions.
Why clicks and form fills mislead
A landing page is a page a visitor first lands on, usually after clicking an ad. Pages differ in who they attract. A page with a very low bar (a free quote button above everything) may collect many weak leads. A page with a price and a clear next step may collect fewer, better ones. If you stop at the form fill, the first page looks like the winner.
The real comparison has three steps: visitors to leads, leads to booked jobs, and booked jobs to revenue. A page can win the first and lose the other two. That is why you need the page recorded on the lead, not just in your website statistics.
Your options, from least to most effort
| Option | What it does | Cost to check | Limit |
|---|---|---|---|
| Tagged links plus a lead sheet | Records the source and page on each lead by hand or hidden fields | Free (your time) | Manual; needs discipline |
| Google Analytics 4 and Search Console | Landing page report, key events, paths to a lead, free search data | Free | Does not see sales on its own; hides small groups |
| Unbounce | A landing page builder with A/B testing (showing two versions of a page to split visitors and comparing results) | Starter $22 yearly or $29 monthly; Build $74 or $99; Experiment $112 or $149; Optimize $187 or $249; Agency from $349 billed annually (USD per month, Unbounce pricing page, as of November 2026) | Another tool and another bill; best when testing pages is your main job |
| HubSpot | A CRM (customer relationship manager) that holds contacts, deals and the source of each | Check HubSpot's pricing page; it shows per seat pricing by plan | Setup and a habit of using it |
| Ploy | An AI website builder that hosts the site, tracks web events and forms, connects to ad accounts, and has an experiments feature | Free $0; Starter $50 a month; Pro $300 a month | Sees web events, not phone calls or offline sales |
| A custom report | Joins ad, site and CRM data in one view | Quote | Overkill for a small account |
The tagging scheme: three labels on every link
You need to know three things for each visit: which ad, which page, and which offer. Use Google's UTM parameters for the first and third, and the page address itself for the second. Google's documentation says to always use utm_source, utm_medium and utm_campaign, that values are case sensitive so lowercase is standard, and that missing parameters can show as "(not set)". Add utm_content to say which ad or which page version.
| Question | Where it lives | Example |
|---|---|---|
| Which platform? | utm_source | google, facebook |
| Which kind of traffic? | utm_medium | cpc, paid-social |
| Which offer? | utm_campaign | boiler-service-spring |
| Which ad or version? | utm_content | video-1, page-b |
| Which landing page? | The page address | /boiler-service and /boiler-service-2 |
Rules that save pain: use one spelling for each name and write them in a shared sheet; never reuse a campaign name for a different offer; and give a changed page a new address or a version label so old and new results do not mix. For more on tags and a lead log see which ads bring leads.
Carry the landing page into the lead record
This is the step most guides skip. Website statistics tell you how many people filled in a form on each page. They do not tell you which of those became customers. To know that, the page and ad must travel with the lead.
- Add hidden fields to your form: landing page address, utm_source, utm_campaign and utm_content. Hidden means the visitor does not see them, and the page fills them in from the web address.
- Make sure the form sends them to your inbox, your sheet or your CRM along with the name and phone number. If the form does not send at all, fix it first with this checklist.
- For calls, give each landing page its own tracked number, or ask the caller and note the page or ad. Calls are the usual gap.
- When the job is booked or paid, update the lead's status and value. That one habit makes the whole comparison possible.
Hidden fields cannot work if the visitor lands on a page without the tags, for example if they come back later by typing your address. Expect some leads with "unknown" source. Track how many, and do not treat the unknowns as a failure of any one page.
Google Analytics 4: a workflow by landing page
Google Analytics 4 is free and shows the first page of each visit in its Landing page report. Google's page on the Pages and screens report says that for the first page you should use the Landing page report instead. Its Key events metric counts the times users triggered a key event, which is an action you mark as important, such as a form submit or a click on your phone number. Mark only the two or three that equal a lead.
- Open the Landing page report, set the date range to match your ad run, and add the Key events column. Sort by key events.
- Filter to your paid traffic using the source and medium you set in the UTM tags.
- For a page you care about, open a path exploration. Google's help says a path exploration can start from a point or end at a point but not both, defaults to the top 5 nodes per step (you can expand to 20), and starts a new session after 30 minutes of inactivity. Use the backward path with your key event as the end to see which pages came before it.
- Link Google Search Console (Google's free tool for your organic search data) to see the search queries that bring visitors to each landing page. Google says you need the Editor role on the property and ownership in Search Console, a link cannot be edited (you delete and recreate it), and data appears 48 hours after Search Console collects it.
Two cautions from Google's documentation. First, small groups can be hidden: Google applies data thresholds to some reports, which cannot be changed, and widening the date range may help. Second, the key events you see in Analytics and the conversions in Google Ads can be different counts, since Google Ads conversions do not appear in the standard Analytics reports. Pick one as your monthly source and write down which.
How much traffic you need before you pick a winner
This is where small accounts go wrong. With a few hundred visits, results swing from week to week and a "winner" is often luck. A standard way to size a test is the two proportion formula, with 95 percent confidence (two sided, z of 1.96) and 80 percent power (z of 0.84): visitors per page = (1.96 x sqrt(2 x p x (1 - p)) + 0.84 x sqrt(p1 x (1 - p1) + p2 x (1 - p2))) squared, divided by (p2 - p1) squared, where p1 and p2 are the two rates and p is their average. I redid the arithmetic with it for the figures below. They are derived examples, not promises, and online calculators may differ by a few visitors.
| Baseline rate | Lift you want to detect | Visitors needed per page (derived) |
|---|---|---|
| 5 percent of visitors become leads | to 7.5 percent | about 1,470 |
| 1 percent of visitors become booked jobs | to 1.5 percent | about 7,750 |
Read the second row twice. Comparing pages on booked jobs needs far more traffic than comparing them on leads, because jobs are rarer. Most small advertisers cannot reach that. The practical answer is to compare on leads for the winner call, and then check the lead to job rate of each page by hand as a sanity check. If one page has a much weaker lead to job rate, that is a reason to change it even without a formal test.
If your traffic is low, a before and after test is weaker but still useful. Run the new page for a stated number of weeks, keep the budget and offer the same, and write down any other change. A tool that splits traffic at the same time is better, because both pages face the same week.
A worked example with made-up numbers
An imaginary roofing company runs one Google ad group to two pages for six weeks, with the same budget. All numbers are invented to show the method.
| Measure | Page A: short form | Page B: price and process first |
|---|---|---|
| Ad spend | $1,200 | $1,200 |
| Visitors from the ad | 1,600 | 1,500 |
| Leads (form and tracked calls) | 96 (6.0 percent) | 75 (5.0 percent) |
| Booked jobs | 9 | 15 |
| Lead to job rate | 9.4 percent | 20.0 percent |
| Cost per lead | $12.50 | $16.00 |
| Cost per booked job | $133 | $80 |
| Revenue (made up average job $2,400) | $21,600 | $36,000 |
Page A wins on leads and cost per lead. Page B wins where it matters: cost per booked job ($80 against $133) and revenue. The sample is small, with 9 and 15 jobs, so I would treat it as a strong hint, keep running B, and look for a way to get B's price clarity into A's short form rather than declare a final result. With 1,600 and 1,500 visitors, it is also below the traffic needed to be confident about small differences in lead rate, as the table above shows.
Where Ploy fits, as one option of several
Ploy builds and hosts the website, so its analytics run from the first publish. Its documents list automatic tracking of page views, sessions, traffic channels (including Paid Search and Paid Social), scroll depth, clicks, and form starts and submissions, plus a way to send your own named events. It forwards browser events to Google Analytics 4, the Meta Pixel (Meta's ad tracking code), Google Ads (Google's ad platform) and the LinkedIn Insight Tag (LinkedIn's ad tracking code), and the tags wait for consent. Ploy does not add a cookie banner, and visitors who decline are not followed across days. Ploy's integrations page lists Meta Ads on all plans, and Google Ads and Reddit Ads on Pro ($300 a month as of November 2026). Its Ploybooks library (saved workflows the Ploy Agent runs) lists an Ad Campaign ROI Report, which uses AI credits each run, at a cost Ploy does not publish.
Ploy's docs also describe an experiments feature: you test one section of one page, split traffic between the original and a challenger, and choose a goal event. The docs say it is available on request through Ploy's sales team, and do not publish the statistics method or a minimum sample size. A page by page comparison on Ploy is therefore possible for web actions, but you still need the sales step from your own records. On Pro, Ploy's connection to Stripe (a payments company) can read payments, which may help match sales to leads, but test that on your own data before relying on it.
Where another tool is better: Unbounce (a landing page builder with built in A/B testing) is the better pick if testing page variants is your main job. See Ploy vs Unbounce and best landing page builders for ads. HubSpot is the better pick if you want one CRM that records the source and deal value on every lead. PloyAds, the ad product with full funnel attribution, is a waitlist page, not a live tool.
A test plan with pass and fail rules
| Check | Pass | Fail, fix by |
|---|---|---|
| Click each tagged ad yourself | Landing page and utm values appear in the form's hidden fields | Fix the field names or the script that fills them |
| Submit a test lead per page | Page and source arrive in the sheet or CRM | Fix the form connection |
| Compare GA4 key events to the lead sheet for one week | Within a believable range (gaps from blockers and consent are normal) | A page with zero events: check the key event and the thank you page |
| Unknown source share | Under about a fifth of leads (my rule of thumb) | Over that: find the untagged links |
| Traffic before a winner call | You reached the figure from the table above, or you label the result a hint | Keep running it |
When a tool will not help
- Mostly offline or phone sales. A web tool, Ploy included, sees web events. Use call tracking or ask every caller.
- Visitors who decline cookies. They are not counted by many tools, so use the comparison between pages, not the absolute numbers.
- Very low traffic. Under a few hundred visits per page, do not run a formal split. Use judgment and read the leads themselves.
- No one to update lead status. Without the sale step, you only have a better looking click report.
Related reading: ad landing pages that match the ad, trying a new version of a weak landing page safely, your first month of ads and client reporting that ties leads to revenue.
How I'd approach this
Decision rule: judge pages on cost per booked job when you have the data, on cost per lead when you do not, and never on clicks. If two pages differ by only a little, keep both running until each has the visitors the table suggests. If one page has a much worse lead to job rate, fix that page even before the numbers are final.
First week plan. Day 1: give each landing page its own address and write the tag table. Day 2: tag all ads and add hidden fields to the forms. Day 3: test each ad, page and form with your own clicks. Day 4: set up the Landing page report with key events and add a status and value column to your lead sheet. Day 5: write down the date you will compare pages and the minimum visitors you need first. If you want a custom report that joins ad, site and CRM data so that each page shows revenue, Blur Studio builds that, and you can see what it includes. This is general information; check each platform's current rules.
A closely related page covers the same ground from another angle: B2B Attribution: Connecting Ad Spend to the Companies That Visit.
Questions people ask
How do I track which landing page leads to a sale?
Give each page its own address, tag the ads that point to it, and add hidden form fields that record the landing page and tags with each lead. When a job is booked or paid, update the lead's status and value in your sheet or CRM.
What is the Landing page report in Google Analytics 4?
It shows the first page of each visit. Google's Pages and screens help page says to use it for first pages. Add the Key events column and sort to see which landing pages produce your marked leads.
How many visitors do I need to compare two pages?
It depends on the rates. Using the two proportion formula at 95 percent confidence and 80 percent power, I derived about 1,470 visitors per page to tell 5 percent from 7.5 percent, and about 7,750 per page to tell 1 percent from 1.5 percent. Rarer outcomes need more traffic.
Should I judge a page by leads or by sales?
By sales when you have the data, since a page can win on leads and lose on booked jobs. When you do not have enough sales to compare, use leads for the winner call and check each page's lead to job rate by hand.
How much does Unbounce cost?
As of November 2026, Unbounce's pricing page shows Starter at $22 a month billed yearly or $29 monthly, Build $74 or $99, Experiment $112 or $149, Optimize $187 or $249, and Agency from $349 billed annually. Check the vendor page for the current numbers.
Can Ploy compare landing pages?
Ploy tracks web events and form submissions, connects to ad accounts, and has an experiments feature available on request. It does not see phone calls or offline sales. Check Ploy's plans for what each tier includes before you decide.
Why do some leads show an unknown source?
Visitors who return by typing your address, use a blocker, or decline cookies arrive without tags. Keep the share small by tagging every ad link, and compare pages against each other rather than trusting the absolute totals.
Do I need a CRM for this?
Not at the start. A sheet with a status and a value column works. A CRM such as HubSpot becomes useful when several people handle leads or you want automatic reminders. Check HubSpot's pricing page for current plans.
More in this topic: Websites: ad tracking and attribution (5 guides) →
Sources
- UTM parameters and (not set) (Google Analytics Help)
- Key events (Google Analytics Help)
- Pages and screens report (Google Analytics Help)
- Path exploration (Google Analytics Help)
- Data thresholds (Google Analytics Help)
- Link Search Console to Google Analytics (Google Analytics Help)
- Unbounce pricing (Unbounce)
- Analytics (Ploy documentation)
- Experiments (Ploy documentation)
- Ploy pricing (Ploy)