Web analytics: measuring what matters

What is web analytics?

Web analytics covers the measurement, collection, analysis and reporting of a website’s data. It answers simple questions. Who comes, and from where? To do what? What stops visitors from getting to the end?

People also talk about audience measurement when the goal is limited to counting and describing visitors. Web analytics goes further. It connects behavior on the site to business results: sales, contact requests, sign-ups.

Tool family	What it measures	Example use
General-purpose analytics	Traffic sources, pages, user journeys, conversions	Steering marketing and the website
Product analytics	Feature usage, cohort retention	Steering an app or a SaaS
Behavioral analysis	Click maps, scrolling, session recordings	Understanding a blocker on a page
Search data	Queries, impressions, clicks from the search engine	Steering SEO
Tag management	Deploying trackers without touching the site’s code	Maintaining the tracking plan
Consent management platform	Each visitor’s choice about trackers	Complying with the law and making collection conditional

On Elev8 Lab, measurement will rely on Google Analytics 4, complemented by other tools from this list as needed (search data, behavioral analysis, tag management). The chosen setup will be documented on this site once it is in place.

Concrete examples of web analytics

Web analytics supports very concrete decisions every day. Here are four typical examples, presented as illustrations:

• An e-commerce site spots the funnel step where carts are abandoned, then simplifies that step.
• A consulting firm compares the pages that generate quote requests and strengthens the ones that convert.
• A media outlet identifies the articles that bring readers back and adjusts its editorial calendar.
• A marketing team compares the cost per request of each channel before allocating its budget.

Where to start: goals and key performance indicators

An analytics account without goals produces reports that nobody reads. So the method starts at the top. First the business goal, then the website’s goal. Next come the key performance indicators (KPIs), then the diagnostic metrics.

1. Goal: write the website’s goal in one sentence. For example, generating qualified quote requests, selling online, gaining subscribers.
2. KPIs: pick three to five that measure this goal. Number of requests, conversion rate, revenue, cost per request.
3. Diagnosis: list the metrics that explain the KPIs. Traffic sources, landing pages, funnel steps.
4. Target: set a value to reach and a review frequency. Every week, every month or every quarter.
5. Owner: name one person per KPI. That person proposes an action when the indicator drifts.

Type of site	Main KPI	Secondary KPIs
E-commerce	Online revenue	Conversion rate, average order value, repeat purchase rate
Lead generation	Qualified requests	Form conversion rate, cost per lead, share of leads signed
Media, blog	Engaged readers, sign-ups	Pages read per visit, returning visitors, newsletter subscribers
SaaS	Trials converted into subscriptions	Activation, cohort retention, churn

Which web analytics metrics should you know?

Definitions vary from one tool to another. A “session” or a “user” is not calculated the same way everywhere. Before comparing two tools or two periods, check the exact definition in the documentation.

Metric	What it counts	Common pitfall
Users	Identified browsers or devices	One person on two devices counts twice
Sessions (visits)	Periods of activity on the site	Splitting rules specific to each tool
Page views	Page displays	A reloaded page counts twice
Engagement rate or bounce rate	Share of visits with or without interaction	Definition varies widely by tool
Conversions	Actions that matter to the business	A poorly defined conversion skews all decision-making
Conversion rate	Conversions ÷ sessions or users	Always state the denominator
Source / medium	Origin of the visit (search engine, email, social network)	Untagged campaign links misclassified

The conversion is the most important metric, and the one most often set up badly. In Google Analytics, any collected event can become a “key event”. It only has to measure an action that matters to the business (Google Analytics Help).

The choice of event matters. Tracking a button click instead of a form that was actually submitted inflates the results. It also misleads the advertising tools that rely on it.

The tracking plan: turning questions into data

The tracking plan (or tagging plan) lists everything the website must measure. For each event, it specifies the trigger, the parameters and the use.

It is the document shared between marketing, developers and analysts. Without it, each vendor adds its own trackers and nobody knows what is being measured anymore.

Event	Trigger	Parameters	Question answered
quote_request	Quote form submitted successfully	service_type, source_page	Which pages generate requests?
guide_download	Confirmed click on the file	guide_name	Which content attracts contacts?
phone_click	Click on the displayed number	location	Is the phone replacing the form?
newsletter_signup	Confirmed sign-up	form	Which forms bring in subscribers?

• A single naming convention: lowercase, words separated by underscores, consistent verbs.
• A precise trigger: “form submitted successfully”, not “click on Send”.
• A “consent required” column for each tracker.
• Acceptance testing before each release, and a check after each website redesign.

Campaign tags and tag management

UTM parameters are added to campaign links: email, social networks, partnerships. They show where the visit comes from. Google Analytics always requires utm_source, utm_medium and utm_campaign. An incomplete link produces “(not set)” values (help on campaign URLs).

Most analytics tools read these same parameters. Tag the links of your email marketing campaigns and your social posts every time.

A tag manager centralizes the deployment of trackers. It speeds up changes. It is only useful if each published tag matches a line of the tracking plan.

Auditing data quality

A wrong figure is worse than a missing one, because it drives decisions. A simple audit, twice a year, prevents most drift.

• Compare the tool’s conversions with actual data (orders, requests received, CRM) over one month.
• Check that no page loads the tag twice, and that no key page is missing it.
• Exclude internal traffic and identified bots.
• Monitor the share of “direct” or “(not set)” traffic. A sudden rise often signals broken tagging.
• Test each event in the tracking plan after a website update.

Consent: what the law requires for audience measurement

In France, placing or reading a tracker on a visitor’s device requires, in principle, their prior consent. This is Article 82 of the French Data Protection Act (loi Informatique et Libertés). Consent must be freely given, specific, informed and given before the tracker is placed. Refusing must be as easy as accepting (CNIL, rules on trackers).

Audience measurement benefits from an exemption, under strict conditions. According to the CNIL, trackers can then be placed without consent. They must be used only to measure audience, exclusively on behalf of the publisher, and produce anonymous statistics. They must not be combined with other data or passed on to third parties. Nor may they track browsing on other websites (CNIL, audience measurement tools).

CNIL recommendation for the exemption	Value
Maximum lifespan of trackers	13 months, with no automatic extension
Maximum retention of collected data	25 months
Use	Anonymous statistics, for the publisher only
Information	Mention in the privacy policy

The exemption depends on the configuration, not only on the tool. The vendor Matomo, for example, documents the settings to apply. It also lists the features not to enable without consent: user ID, heatmaps, session recordings, full e-commerce tracking (Matomo exemption guide).

Missing data, modeling and transfers

When a visitor refuses, their data is not collected, or is collected in a reduced form. Some tools offer mechanisms to take this into account.

Google’s consent mode passes the visitor’s choice on to the tags. In basic mode, nothing is sent before consent. In advanced mode, cookieless signals are sent and used for modeling (help on consent mode). This setup does not replace a compliance review.

Another point to check: where the data goes. On July 10, 2023, the European Commission adopted an adequacy decision for US companies that adhere to the Data Privacy Framework (European Commission). Check that your provider adheres to it. Mention these transfers in your privacy policy.

Attribution: which channel gets credit for the conversion?

A customer often has several touchpoints before buying: a search, an ad, an email. Attribution splits the credit for the conversion between these touchpoints. No model is neutral: each one favors certain channels.

Model	Rule	Channels favored
Last click	All the credit to the last touchpoint	Branded search, reminder emails
First click	All the credit to the first touchpoint	Discovery channels (social networks, display)
Linear	Credit shared equally	None in particular, but none stands out
Data-driven	Split calculated by the tool from observed journeys	Depends on the available data and the tool’s algorithm

Models only see measured touchpoints. Word of mouth, trade shows and phone calls escape them. To settle a budget question, an incrementality test often gives a more solid answer than a change of model. The principle: switch off a channel in one area or for one period, then compare.

For paid search, the link between attribution and automated bidding is covered in our SEA guide. For organic search, the reference data remains Google Search Console. For AI search engines, see measuring your GEO visibility.

Dashboards and reporting: driving decisions

A dashboard is there to make decisions, not to decorate a meeting. It fits on one screen and starts with the KPIs. It compares them with a target and a reference period. Diagnostic metrics come next, to explain the gaps.

• One page per audience: management (KPIs and trends), marketing team (channels and campaigns), website team (pages and journeys).
• Each chart answers a question written as its title.
• Annotations date the events that explain a change: redesign, campaign, outage, new consent banner.
• A commented review every month, with three actions decided and followed up.

The choice of indicators follows from the digital strategy. A KPI with no link to a business goal has no place in the dashboard.

Web analytics also feeds conversion optimization. Spotting where visitors drop off is the starting point of CRO. Finally, it connects to customer data to measure retention, the subject of our CRM and customer loyalty guide.