How to measure website results without confusing visits with enquiries

Measuring the website helps decide what to improve. A visit, a contact click and an enquiry received represent different moments. The plan starts by defining those moments and the source confirming them. It then organises origins, privacy choices and comparisons so a report does not express more certainty than the data permits.

By Lyntev. Updated on .
About 5 minutes to read.

From data to the next decisionChoose the question, define the calculation and act on the result.

Translation published on . Original Portuguese guide; source revision .

By the end, you can:

  • Distinguish intent, submission and confirmed result
  • Document events and observation limits
  • Choose improvements from consistent comparisons

Choose the question before the tool

Write a decision the report should support: improve a service page, discover where the form fails or compare enquiry quality from two sources. Define the expected result and who confirms it. On a services website, this may be a valid enquiry received; in a shop, an order with confirmed payment. Interpretation must lead to an action the team can take.

Separate intent signals from results. Opening WhatsApp does not demonstrate a message was sent; clicking a payment button does not confirm a purchase; visiting the thank-you page does not, by itself, prove a new record. Keep these signals when useful but name them for what they actually observe. This avoids optimising the interface for clicks that do not correspond to useful work.

Choose the question before the tool
SignalWhat it confirmsWhat remains unknown
Contact clickIntent to open a channelWhether there was a conversation and its result
Form startedInteraction with the requestWhether it was sent and accepted
Request recordedReceipt confirmed by the systemQuality and sales follow-up
Payment confirmedStatus returned by the payment sourceSubsequent delivery, cancellation or refund

Define the exact moment of every event

Prepare a dictionary with name, trigger condition, necessary parameters and source of truth. An enquiry-received event should depend on real confirmation, not merely the click. If using a tool with recommended events, such as Google Analytics, confirm documented meaning and parameters before applying them. The same name with different rules on different pages prevents useful comparisons.

Do not send message content, email addresses or other personal data to analytics parameters. Use controlled categories helping answer the question, such as service type and form version. Also define how to prevent duplicates when a page reloads or a response repeats. The operational request identifier may be needed in the business system; that does not mean every analytics provider should see it.

Use consistent source names

For campaign links you control, establish a convention for source, medium and campaign. UTM parameters help compatible tools identify those visits. Use stable, documented names; alternating between “Instagram”, “instagram” and abbreviations can fragment the report. Do not apply campaign tagging to every internal link, confusing arrival at the website with navigation inside it.

A fictitious example would be a newsletter link to a guide: source newsletter, medium email and campaign guia-manutencao. Keep the final link and test that the destination opens correctly, including redirects. Technical attribution still has limits: device changes, link sharing and privacy choices may alter what is observed. An optional question about how the person heard of the company is another source, not a perfect substitute.

Test choices before interpreting figures

Inventory measurement technologies and purposes. Configure optional resources according to applicable choices and confirm in the browser that they do not start loading before the intended point. Test first visit, refusal, acceptance and withdrawal while observing network requests and storage. A visible refuse button does not demonstrate resources respect that decision.

Document who is represented in each dataset. A consent-based tool may observe only some visits; operational records may contain every enquiry received. Do not automatically divide one total by the other. Official European guidance explains purposes requiring consent and withdrawal; implementation and information supplied need review for the specific context, without treating tool installation as legal certification.

Fictitious example: two totals that cannot be mixed

In an entirely fictitious set, the analytics tool observes 320 eligible visits and 16 confirmed enquiries associated with those visits in the same period under the same rule. The observed ratio is 16 divided by 320, or 5%. The operational system records 24 enquiries overall. Dividing those 24 by the 320 observed visits would mix populations, attributing enquiries whose origin was not observed to the measured group.

Useful interpretation presents both figures with their scope. It then investigates quality: how many enquiries were valid, how many needed additional information and who followed them up? The team may discover an unclear field or poorly explained offer. The example does not define a desirable rate for your business or demonstrate actual performance; it shows how to keep denominator, period and source consistent.

Fictitious example: two totals that cannot be mixed
Illustrative dataValueCorrect interpretation
Eligible observed visits320Tool coverage and chosen rule
Enquiries associated with that base16Observed ratio of 5%
Total enquiries in the system24Operational volume with different coverage
Difference between sets8Investigate coverage; do not attribute a source without evidence

Validate measurement and compare similar conditions

Run an identified test journey and follow each event through to the report or debugging record. Confirm failures do not generate success, a reload does not create another result and refusing optional resources keeps the essential journey usable. Record the version and exclude tests from business reports through the tool’s appropriate procedure.

Compare complete periods with equivalent filters and note changes to offer, campaigns, form or availability. In Search Console, impressions and clicks describe search presence; in the commercial system, enquiries have another definition. Common errors include adding users from different reports, interpreting average position as a fixed ranking and attributing sales to a change just because they happened afterwards. Choose a hypothesis and seek enough evidence before the next intervention.

Use a worksheet ending in a decision

Keep the plan short enough to use and precise enough to reproduce. Every measure needs an owner, definition and possible action. If a number changes no decision, review its usefulness before adding collection. The following worksheet accompanies implementation, interpretation and the next hypothesis, connecting technical work to the website’s objective.

Worksheet for your project

Question and decision we want to support:
Business result and source confirming it:
Event / exact moment / permitted parameters:
Duplicate rules and test exclusion:
Source, medium and campaign convention:
Optional resources and choice behaviour:
Observed population and coverage limits:
Period, filters and denominator:
Validation test and result:
Observation / hypothesis / missing evidence:
Chosen change and owner:
Date and criterion of the next comparison:

Sources and criteria

Content prepared with AI assistance and research using technical sources, with checks of the relevant examples and calculations. The references support the guide’s decisions; examples identified as fictitious let you try out the method.

Wellinton Oliveira

Written by

CEO of Lyntev

Technology with context, original analysis and sources you can consult.