How to tell which leads came from your website, Google Ads and Google Business Profile

An enquiry can be traced to its source by recording where the visitor came from at the moment they ring or send the form. Google Ads, your website's organic traffic and Google Business Profile (GBP) each leave different clues, and some leave none. This page lists what each source shows, a 20 minute method to check your own numbers, and what we measured across five Australian service business sites we run. Figures are as at November 2026.

Where does each kind of lead show up?

Each source is visible in one place and invisible in others. The table below is the one to check against when your numbers do not add up.

How do I check where my leads really come from in 20 minutes?

Count the same period three ways and compare the totals. Use the last 28 days.

1. Count real enquiries from your own records. Form emails in your inbox, calls in your phone log, messages on your profile. This is the number everything else gets compared with.
2. Open Google Ads, go to Goals, then Summary, and list each conversion action with its source. Treat Google hosted actions as suspect. Our post on Google Ads conversions that were nearly ten times the real leads shows what to look for.
3. Open your website analytics and read the lead events by source. Check the phone event fires only on a real tap of a number, not on a click into a form field. See phone click tracking that counts the wrong clicks.
4. Open the GBP Performance screen and note calls, messages, website clicks and directions.
5. Put the totals side by side. If Ads plus analytics plus the profile add up to much more than your records, something is counted twice or counted wrongly. If they add up to less, some leads are invisible, and the next section lists the usual reasons.
6. For one week, ask every caller how they found you and write it on a sheet. Compare it with the tracker.

If you only do one thing, do step 1. A cost per lead built on a conversion column nobody has checked is not a cost per lead.

What we measured across five of our sites

We run a first party tracker on the sites we host. It records a phone tap or a server side form send, with the visit's source. From 14 October to 4 November 2026 (22 days) we pooled five service business sites: an appliance repairer, a carpenter, a pest controller, a builder and a workwear supplier. A lead session here means a visit with at least one phone tap or form send.

Visits from Google Ads are left out, because only two of the five sites run Ads and we do not publish a rate for fewer than three. The AI row is 39 visits, so treat the rate as an indication and not a benchmark.

The row that matters most for tracking is Direct. Of the 133 lead sessions in the table, 41 (31 per cent) arrived with no recorded source. If you rely on a report that stops at "direct", a third of your leads have no origin. Some of those are people who saw a van or typed the address from memory. Some are taps where the referrer was stripped.

Leads also arrive by different routes depending on the trade. The same 22 days, counting sessions with at least one tap or send:

Pooled, that is 124 phone taps and 94 form sends, so a form-only report would miss more than half the contact on the trades that get phoned. Two caveats apply. The workwear form count includes some spam and vendor pitches, and the builder phone count includes a few taps from our own testing. A small business with a handful of leads a week will see these ratios move from week to week.

Why does my tracking miss leads?

These are the faults we have actually found on live sites. Each one made a count wrong without any visible error.

• A security header blanked the booking form and stopped analytics. On one service site a content security policy was missing the booking widget's host and Google's newer data endpoints. The page returned 200 and looked fine, but the form area rendered empty and Google Analytics received nothing for about three months. Eight automated site audits ran in that period and none flagged it.
• Phone numbers typed as plain text. On a phone, the operating system turns a plain-text number into a link, but the site's tracker only listens for real number links. We found plain-text numbers on five sites and fixed them in October.
• Taps before the page finished loading. The tap listener only works once the page's scripts have started, which can take about three seconds on a mid-range 4G connection. A fast tap is missed.
• Removing one tool broke another. On one site the Google Ads phone conversion was only installed if the old analytics tool's key was set. We caught it before deploying the removal. Check what depends on a tracker before you delete it.
• The wrong thing counted as a phone call. A rule that treated any event with "phone" in its name as a call counted people clicking into the phone box on the contact form. See the section on phone click tracking below.
• Your own browser may block the tags. Ad blockers and privacy settings can stop the tags in an owner's browser while they run normally for customers. Test in a clean browser profile before concluding that tracking is broken.
• Spam and bots inflate counts. One site had about 3,240 US mobile visits in a single hour, from a bot, in both our analytics and Google's. On another, two of five form entries in a week were vendor pitches, not customers.

For the Ads side of this, why Google Ads tracking breaks without anyone noticing covers blocked tags in more detail.

Why can phone click tracking count the wrong clicks?

Because the rule behind the conversion may be broader than a tap on a phone number. On one local service account we manage, the old tracking bridge fired a "phone call" event for the proper phone click and also for any event whose label contained the word "phone". Over a September fortnight about one in six of the recorded "calls" were people clicking into the phone number box of the contact form. Nobody rang anyone. The rest were real taps on a phone number link, almost all on mobile.

A click into the phone box is not a call. If the visitor then sends the form, that is a lead worth counting once as a form submission. When we replaced the bridge with a single listener on tel: links, recorded phone clicks fell sharply on the day. We checked before trusting either number:

1. Do the buttons fire? A scripted browser clicked the first visible phone link on two page types and watched the network calls. Each tap fired the site event and the Ads conversion.
2. Is the old baseline fair? Recounting the old period without the form field events gave a lower baseline. Comparing the new period against the inflated one would have overstated the drop.
3. Did the calls move? Clicks on the ad's own call button and calls from ads did not rise to make up the difference, so the calls had not moved to another route.

To test your own account:

1. List every phone related conversion action in Goals, then Summary, and note which are website events, which are call reporting from the ad and which are Google hosted.
2. Ask whoever built the tracking what exactly fires the website one. A rule that matches a word in a label will count more than calls.
3. Tap each number on your own site on a real phone, and confirm one event arrives and that tapping into a form box produces none.
4. Keep taps and connected calls as separate actions, because a tap shows intent and a connected call is a stronger signal.
5. Compare with your phone log for the same dates. A gap of several times is a sign of broken counting.
6. Do not compare periods across a tracking change without recounting the old period under the new rule.

If automatic bidding uses these numbers, the check matters more, because bidding learns from what you tell it counts. The related problem of Google created actions is in why Google Ads reports about ten times your real leads.

How do I track Google Business Profile leads?

Calls, messages, bookings and directions show in the profile's Performance screen, and nowhere on your website. Website visits from the profile reach your site, so tag the profile's website link so they show up as their own source. A common way is to add campaign parameters to the link, for example ?utm_source=google&utm_medium=organic&utm_campaign=gbp, and to use the same address your site actually serves, with https.

That detail matters because we have seen the opposite. On an appliance repairer's site, the profile's website and appointment links pointed at the old http:// address. Search Console then listed clicks on those links as separate http:// pages, and that series fell from 35 to 19 clicks over two comparable 14 day periods while the https organic pages held steady. It looked like a ranking loss, and it was the profile traffic. Our post on the Google Business Profile map pack covers the profile itself.

How many enquiries did your website bring last month?

Most owners cannot give a number, because the inbox feels like it is already counting. It is not. An inbox mixes form submissions with supplier emails, spam and old threads. It has nothing on people who rang instead or left a form half finished, and nothing on where anyone came from. A quiet month can also be a form that stopped sending, since a broken form does not email you to say so.

The first step needs no tools. Keep a shared spreadsheet, an email label or a note on your phone with one row per enquiry: date, name, how they found you if you know, and what happened in the end. Do it for one month. For a business with around fifteen enquiries a month it captures phone calls and outcomes that most analytics setups miss.

Then let the site count form submissions itself. In Google Analytics 4, mark the form event as a key event, which is an event that measures something particularly important to your business, and reports will split those enquiries by source. Two traps come up on existing setups:

1. Counting the wrong moment. An event that fires on the thank-you page or on the button click records a conversion even when the message never arrives. Fire it on a confirmed send, and test by submitting the form yourself and following the message to the inbox.
2. Double counting. A page refresh, a back button or a second submit because the first seemed slow all inflate a small number. Keep a plain-email copy of every submission going to a shared address so the count can be checked against it.

For calls, which analytics cannot see, use call tracking (a separate number on the website that forwards to your real one) or ask callers "how did you find us" and note the answer in the same spreadsheet. The second is less rigorous and costs nothing.

What to do with the result

Once you know the split, you can judge each source on its own. Ads can be judged by cost per real lead, organic by the pages that bring leads, and the profile by calls and direction requests. The section above on counting enquiries shows how to start with no tools at all.

If you would rather not do this by hand, it is built into our plans. Every Pagecog plan includes a client dashboard, a first party tracker that records phone taps and form sends with their source, and a weekly report. The base website plan is $298 a month, and the Ads management add-on is $549 a month. See pricing, the analytics service and the Google Ads management service, or get started.