Short answer
A click on a tel: link shows that someone activated the link, but they can cancel, encounter a busy line or reach voicemail. Measuring answered calls requires phone records and a definition of a conversation; the web event alone does not establish that.
Key verification: Test a cancelled click, unanswered call and answered call as separate cases.
Sources and limitationsWhen someone taps the phone number on a website, the browser may open the dialler or a compatible app. That click shows an interaction with the link. The person may cancel, hit a busy line or reach voicemail. To confirm that a conversation took place, you need data from the phone system.
The opposite can also happen: someone copies the number, dials it from another device or calls later. You will then have a call without the click you expected. That is why you should not divide calls by clicks and present the result as an individual connection rate unless you have been able to link them.
Four stages worth separating
| Stage | What it shows | What it does not show | Data source |
|---|---|---|---|
| Phone click | Interaction with the link | That the number was dialled | Web tag (GA4, Tag Manager) |
| Logged call | Attempt received by the system | That a person answered it | Phone system or call tracking log |
| Answered call | Connection as the phone system defines it | That there was a useful enquiry | Call status and duration |
| Qualified contact | Business criteria met | A confirmed sale | Team or CRM review |
A phone system may treat a call as answered as soon as it connects to a menu or voicemail. Check the system’s definition and, where it matters, separate technical connection from human conversation. The “answered” status must be read according to the implementation.
Spain now has a useful legal reference for that distinction. Article 10.3 of Law 10/2025 on customer service states that a call is not effectively attended if it does not let the caller state the reason for calling and ask for an operator. It applies to providers of basic services and to companies above 250 employees, €50 million turnover or €43 million balance sheet, with twelve months to adapt from its entry into force on 28 December 2025. Even if your business is not covered, the criterion helps separate “picked up by a machine” from “attended”.
The full journey and where it can break
The diagram explains why the two totals do not match: the click loses cases to the right and the phone system gains cases from the left. On desktop the loss is usually larger, because a tel: link may open a desktop app or nothing at all. Always split by device before comparing.
How to measure the click without giving it another meaning
If you use Tag Manager, set up a “Just Links” trigger that fires on some clicks whose URL starts with tel:, plus a GA4 event tag such as phone_click. Check in preview mode that it fires on the specific link and does not fire two tags for one interaction. Google documents click triggers and their filters. Click triggers.
Identify the link’s position with a generic label, such as “header” or “contact”, rather than sending the full number. Also check which variables your tag includes. Capture must respect consent: the Spanish data protection authority’s cookie guide (May 2024 version) does not treat analytics cookies as exempt, so some real clicks will go unrecorded when visitors refuse measurement.
You can mark the event as a key event in GA4 if that helps you monitor a page, but that does not turn it into a conversation. Google Ads makes the same distinction: its conversion type for clicks on a number on a mobile website tracks only the click, not the call, whereas calls from ads or to a forwarding number let you set a minimum duration. Call conversions in Google Ads. To set those actions up, follow the guide to measuring Google Ads calls in Spain (in Spanish).
Worked example: 50 clicks and 30 calls
Sample data for one month on a services page. GA4 shows 50 phone_click events and the phone system shows 30 calls to that page’s number. Subtracting 50 − 30 and saying that “20 people gave up” would be wrong. This is how to reconcile them with records that can be linked by time and tracking number:
| Item | Count | How it is obtained |
|---|---|---|
| Click events in GA4 | 50 | Events report |
| Duplicate events from double tagging | 4 | Same second and same session |
| Unique clicks | 46 | 50 − 4 |
| Clicks followed by a call within 2 minutes | 24 | Matched against the call log |
| Clicks with no linked call | 22 | 46 − 24 |
| Calls with no prior click | 6 | 30 − 24 |
| Calls the system marks as answered | 26 | Phone system status |
| Answered by menu or voicemail only | 3 | Listening or agent-leg duration |
| Human conversations | 23 | 26 − 3 |
| Qualified contacts | 11 | Sales review |
With these figures, click-to-call is 24 / 46 = 52.2%, not 30 / 50 = 60%. The human conversation rate on logged calls is 23 / 30 = 76.7%, against the 86.7% (26 / 30) that the technical status suggests. And 11 of 23 conversations, 47.8%, were qualified. Each ratio answers a different question and uses its own denominator.
Edge case: if the page gets a lot of desktop traffic, the 22 clicks with no call may be almost entirely browsers that did not open a dialler. Before changing the design, split clicks by device; if nearly all the clicks that go nowhere are on desktop, the problem is what the link leads people to expect, not visitor interest. When traffic sources do not match across platforms, the section on Google Ads clicks and GA4 sessions, further down, covers the reconciliation that comes first.
A simple test that makes the difference clear
- Open the test page and check that authorised measurement is active.
- Tap the phone number and cancel the dialler before calling.
- Check the click event and the call log. The expected result is a click with no call; write down what you actually observe.
- Dial the test number manually from another device. Check whether a call is logged without a linked web click.
- Make one unanswered and one answered attempt. Check how voicemail, the menu and the conversation are recorded.
This is a proposed protocol, not a test run here. It should use a controlled number and be excluded from business results. If the first step does not record the click, fix the tag; do not conclude that people are not interacting with the phone number.
Which stage to use for each decision
| Decision | Stage that answers it | Why the click is not enough |
|---|---|---|
| Whether a redesign makes the phone number more visible | Phone click, by device | Here it is enough: it measures interaction with the link |
| Whether to add call cover in a time slot | Logged call versus human conversation | The click cannot see busy lines or voicemail |
| Whether a campaign brings useful customers | Qualified contact | A cheap click may never become a conversation |
| Which value to feed into Google Ads bidding | Call with minimum duration or imported outcome | Optimising for clicks rewards intent, not outcome |
If clicks rise but conversations do not, review the call routing and call handling; the guide to queue abandonment rate helps measure that part. If conversations rise without opportunities, review the offer and qualification. To join each call to its session and its CRM outcome, see how to connect calls, GA4 and a CRM (in Spanish).
Google Ads clicks and GA4 sessions: what each tool counts
Ads counts clicks on advertisements. GA4 groups interactions into sessions; several clicks can fall within an existing session. Google’s help on click and session differences, consulted on 27 October 2026, lists causes in both directions: Ads removes clicks it considers invalid, Analytics may group repeat clicks within 30 minutes into one session, anyone who leaves before the tag loads produces a click without a session, and redirects can strip the gclid or UTM parameters. With auto-tagging off, paid visits can appear as organic, direct or “(not set)”; with it on, returning users can produce more sessions than clicks.
The current GA4 session definition sets a default of 30 minutes of inactivity, adjustable in the tag settings, and explains that campaign data travels with the session start event. Closing a tab is not sufficient evidence that a new session began.
Clicks and sessions that do not match: diagnostic matrix
| Situation | Likely effect | Check |
|---|---|---|
| Several clicks within 30 minutes | More clicks than sessions | Review intervals rather than require equality |
| Page fails to finish loading | More clicks than sessions | Response, loading and events in an authorised test |
| Redirect strips gclid or UTM | Paid visits under organic or direct | Compare initial and final URL |
| Auto-tagging switched off | Paid sessions shown as (not set) | Check the setting in the Ads account |
| Returning users | More sessions than clicks | Separate new and returning users |
| Consent choice or blocking | Different observation between systems | Respect the choice and document the implementation |
| Very recent data | Provisional figures | Wait 48 hours and export again |
Before investigating, save exports using the same reporting window, campaigns, account, property and time zone, and use the session campaign dimension, not the first-user one. Separate networks or interaction types that do not lead to your website. Record filters, observed or modelled data and export time. If the accounts were linked mid-period, earlier days are not backfilled.
Common mistakes and how to spot them
- Double tagging. Symptom: more clicks than page visits on some days. Check: tap once in preview mode and count the events; look for one tag in the code and another in the tag manager.
- Badly formatted link. Symptom: the event is recorded but the dialler shows an incomplete number. Check: the target should be
tel:+34followed by the nine digits, with no spaces or duplicated prefixes. - Click used as a bidding conversion. Symptom: Google Ads shows many conversions while the team receives few calls. Check: review which actions are set as primary and whether any of them is only a click.
- “Answered” that includes voicemail. Symptom: a high answer rate alongside complaints from customers who cannot reach anyone. Check: compare call status with the duration of the agent leg.
- Different periods or time zones. Symptom: daily totals do not match but monthly totals almost do. Check: align the GA4 and phone system time zones before matching records.
- Consent ignored in the analysis. Symptom: fewer clicks than web-sourced calls. Check: estimate how much traffic refuses measurement before concluding that the button does not work.
From the author's experience
In telephony integrations, the first thing I always separate is a call that is technically answered from a conversation actually handled by an agent. With 11 years in implementation, support and software development for contact-centre systems (.NET, C#, APIs, at Vocalcom), in an integration I look for the call identifier and the routing and agent-connection events, not a generic "answered" field, which can mean almost anything depending on how it was implemented.
On the time window between a click and a call: without a shared identifier, a two-minute window can work as an initial approximation, but never as reliable attribution. I would validate it by cross-checking number, time and duplicates, and flag ambiguous matches instead of forcing them to fit. To compare mobile and desktop, I would build the comparison from click events and phone records for the same period, splitting by device and repeat sessions, before drawing any conclusion.
I also would not count a potential sale just because a call runs past some arbitrary duration threshold: as a starting point I would analyse duration brackets and a sample classified by outcome, because an urgent issue can be resolved quickly and a long call is not necessarily a useful one. When conversions and calls do not add up, I first check that both reports are counting the same thing — click, call started, connection or human attention — and only then review time zone, numbering, duplicates and attribution rules, because mixing stages produces mismatches even when the systems themselves are working correctly.
Sources and limitations
Documentary review: . Content type: Conceptual guide.
- Law 10/2025 on customer servicewww.boe.es
- Click triggerssupport.google.com
- Spanish data protection authority’s cookie guidewww.aepd.es
- mark the event as a key event in GA4support.google.com
- Call conversions in Google Adssupport.google.com
- click and session differencessupport.google.com
- current GA4 session definitionsupport.google.com
Official sources, explained calculations and clearly labelled examples. Read about our methodology and use of AI in writing.