Pro analytics

Campaigns, locations, events, outbound links, downloads, site search and scroll depth. All Pro, all gated at the service layer rather than in the UI.

Campaigns

Links tagged with utm_source (and optionally utm_medium, utm_campaign, utm_term, utm_content) are attributed to the session, not the pageview - a campaign brings somebody to the site once, not once per page they then read.

Values are lower-cased and trimmed, so Newsletter and newsletter stay one campaign rather than becoming two.

UTM and click-id parameters are stripped from the recorded path. They describe how somebody arrived, not which page they arrived at, and they are captured separately. Leaving them on would split one page into a row per campaign and inflate path cardinality for data recorded elsewhere. gclid, fbclid, msclkid and friends are stripped for the same reason - but they are not read as campaigns, because a click id identifies an ad click, not a source, and inventing a campaign from one would be making data up.

Attribution models

Model	Credit
last-click (default)	The most recent touch takes it all
first-click	The touch that first brought them takes it all
linear	Split evenly across every touch

The model only matters for a session with more than one touch, which is uncommon - with a single touch, every model gives an identical answer.

One session is always credited as exactly one session. Under linear, a session with three touches contributes 0.333 to each, so the column can be fractional and the total still equals the sessions you actually had. An attribution report that credits a session to three campaigns in full is claiming traffic the site never received.

What this cannot do, and why

There is no cross-session attribution window for anonymous visitors, and we will not fake one.

The Tier-1 visitor hash is re-salted every 24 hours and the old salt destroyed, so there is no way to know that today's visitor is the person who clicked an ad last week. That is the same property the banner-free claim rests on - you cannot have both. Attribution is therefore within a session.

Consented (Tier-2) visitors keep a durable identifier and can be followed across sessions; if multi-touch attribution over weeks matters to you, that is what it costs. See privacy.

Locations

Setting this up needs a database file you install yourself - see Locations & the geo database.

Country and region, resolved from a database on your own server. No lookup service is called, nothing leaves the machine, and the address is resolved and discarded inside a single call frame - only the country code and region name are stored (C5, C7).

php craft craft-analytics/geo/install --url=<db-ip lite url>
php craft craft-analytics/geo/status

Nothing downloads on its own. The command exists so that fetching a database is a decision an operator makes.

Database	Licence	Notes
DB-IP Lite (recommended)	CC BY 4.0	Free and redistributable. Attribution to DB-IP is required, and the CP shows it for you.
MaxMind GeoLite2	MaxMind's own	Free with an account. We can read the file; we can't ship it for you.

Both are monthly releases - re-run the install command to update.

Events

craftAnalytics.event('signup');
craftAnalytics.event('purchase', { value: 49.99 });

Values are clamped and event names stripped of control characters server-side. The client can say anything; an event worth 10²⁰ is not a sale.

Outbound links and downloads

Tracked automatically when enabled. Clicks are caught in the capture phase, so they are recorded even when the page's own handlers stop propagation or navigate away. A click is genuinely a separate event from the pageview, so it sends its own beacon - there is no way to report a click on the way out of the page it left.

Downloads are same-host links to a file with one of the extensions in downloadExtensions.

Scroll depth

Four buckets - 25/50/75/100 - because "how far down did people get" has four useful answers and 101 useless ones.

It rides the pageview beacon rather than sending its own request, via an extension point in the tracker. One request per pageview is the budget, and it stays the budget. Reaching a depth counts once per pageview, and reaching 75% means they also passed 25% and 50%.

Site search

Read from the URL of your search page - no JavaScript required, so it works for visitors who have it turned off:

'trackSiteSearch' => true,
'siteSearchPath' => '/search',
'siteSearchParam' => 'q',

Terms are lower-cased and trimmed so one search stays one search. Whether a search found anything is something only your site knows; report it yourself with craftAnalytics.event('search:zero-results').

Segments

Traffic broken down by whatever your own site knows about a visit - the plan somebody is on, whether they were signed in, which role they hold. Sessions, views, bounce rate and average duration for each group.

This one has no settings. It appears only once a module on the site declares what to segment by, and the Segments nav item is absent until then. Doing that is a few lines in your module: see Extending.

It stores counters, not people. A segment value is a dimension shared by everybody who matches it, so a visit in a segment costs one row per day however many pages it covered - and nothing on the screen can be traced back to an individual, because nothing about an individual was kept.

Two things the screen says itself, worth knowing before you read it:

• The tables do not add up to your total sessions. A visit counts in every segment it belongs to. "Plan" and "Role" are two questions about the same visits, not two sets of visits.
• A visitor is whatever they were when they arrived. Somebody who signs in halfway through a visit is still counted as a guest for that session, which is the honest answer to where the session came from.

The scripts

File	Loaded when	Size (gzipped)
tracker.js	always	1,837 bytes (budget 2,048)
consent.js	Pro and consent enabled	1,258 bytes
pro.js	Pro and events or segments enabled	2,380 bytes

Split deliberately: a Lite site downloads only tracker.js, and pays nothing for features it doesn't have. Only tracker.js is under the C3 budget, because only tracker.js is what every visitor to every site gets.

Settings

All of these are on the What gets tracked section of the settings screen (Settings -> Plugins -> Craft Analytics), and every one can also be set in config/craft-analytics.php to differ between environments.

Setting	Default
enableCampaigns	true
attributionModel	last-click
enableGeo	false (needs the database installed)
enableEvents	true
trackOutbound / trackDownloads / trackScroll	true
trackSiteSearch	false