The Signal Room · Analytics Field Manual

Lesson 01

What you are actually looking at

A Google tag sends events from your website to an Analytics property. An event names an action; parameters describe its context. A key event marks an action you consider important. Reports, funnels and paths arrange those observations.

Users are measured identities, not a headcount of humans. Cookies, devices, consent and reporting identity affect the count. A session is a visit that normally times out after thirty minutes of inactivity; GA4 does not reset it at midnight.

• Engaged session: longer than the configured engagement timer, a key event, or at least two page or screen views. The default timer is ten seconds.
• Engagement rate = engaged sessions ÷ sessions. Bounce rate is its inverse.
• Ask one question before opening a report.

Lesson 02

Plant the tag

Sign in to analytics.google.com with an account authorised for your own property. In Admin, open Data streams and select or create a Web stream. Copy its Measurement ID. Follow your platform’s integration, Tag Manager or Google tag installation instructions.

For hand-built pages, place the base tag in the head of every intended page, once. Replace the example ID with yours. A tag in a comment is inactive; a text search alone cannot prove installation.

• Test a live page with Tag Assistant and inspect event parameters in DebugView.
• Use Realtime as a quick collection check, allowing processing time.
• Check consent state, blockers and duplicate tags if collection fails. Standard reports can take twenty-four to forty-eight hours.

<!-- Example only: replace G-XXXXXXXXXX in both places. -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-XXXXXXXXXX"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag("js", new Date());
gtag("config", "G-XXXXXXXXXX");
</script>

Lesson 03

Build the clean room

Configure measurement before interpreting it. Choose retention appropriate to your purpose: standard properties offer two or fourteen months for event data. Retention affects explorations and funnel reports, not standard aggregate reports. Increasing it preserves eligible data that has not already been deleted.

Define internal traffic for addresses you control, then test the exclusion before activating it. An active exclusion permanently removes matching new data. A private browser window does not exclude you.

• Review unwanted referrals for payment handoffs; exclusions do not join identities across domains.
• Audit enhanced measurement: page views, ninety-percent scroll, outbound clicks, search, eligible YouTube videos, file-link clicks and forms.
• Set the reporting time zone and currency for your studio.
• Link verified Search Console ownership with Editor access, then publish its report collection if needed.

Lesson 04

The language: events, parameters and key events

Automatic and enhanced measurement events cover common behaviour. Recommended events use Google’s defined names and meanings. Custom events describe actions specific to your site. Use generate_lead for a submitted request; use contact_click for a mail or telephone link. Clicking does not prove a message was sent.

Keep names consistent: lowercase letters and underscores, starting with a letter. Event names are case-sensitive. Describe galleries with parameters, rather than separate event names.

• Respect reserved names, forty-character event and parameter names, and twenty-five parameters per event. Most parameter values are limited to one hundred characters.
• Web streams have no five-hundred-event-name collection cap.
• Mark a small, purposeful set of key events and choose once per event or once per session intentionally.

Lesson 05

Wire your own pages

Measure the doors you deliberately built. Track contact menus as contact_revealed, contact links as contact_click, gallery handoffs as portfolio_opened, and external creative-platform links as platform_click. Label each with a stable surface or project category.

For tutorials and creative releases, send content_copied only after copying succeeds; section_expanded when a section opens; and sequence_toggled when a sequence changes. These are useful actions, not proof of a purchase, saved track or client decision.

• For long pages, optional twenty-five, fifty and seventy-five percent scroll milestones describe depth, not time spent looking.
• For guides, measure submitted searches and search_no_results using safe topic labels; never transmit unrestricted personal text.
• Use the snippet forge for click, page load, visible element and manual triggers. Review its selector and test the output in DebugView.

Lesson 06

Custom dimensions: make parameters usable

Sending a custom parameter does not automatically add it to every report. In Admin → Custom definitions, create an event-scoped dimension with the exact parameter key. After processing, add it to a relevant report or import it into an exploration.

Register the labels that answer actual questions. Reuse built-in dimensions where available. Custom definitions are not a recovery mechanism for older unregistered report data, although separately retained raw exports may contain it.

• Useful labels: project, surface, platform, contact method, content type, content name and scroll percent.
• Allow twenty-four to forty-eight hours after collection and definition creation.
• Standard properties offer fifty event-scoped custom dimensions. Avoid unique visitor IDs, private gallery names and unbounded values.

Lesson 07

Cross the border to your gallery

If your portfolio lives on Pixieset or another domain, check whether that product supports your GA4 tag. Use the same web stream and Measurement ID on both sites, then add the exact intended domains under Configure tag settings → Configure your domains.

Cross-domain measurement passes identifiers in a _gl linker parameter when consent permits. Test the handoff, preserved redirects and destination events. A decorated URL is useful evidence, but verify that the destination tag actually receives the journey.

• If the gallery cannot run your tag, measure portfolio_opened as a departure only. Activity beyond that point remains unobserved.
• Configured cross-domain links do not generate enhanced outbound-click events; retain your deliberate handoff event.
• Investigate unexpected self-referrals, cookie configuration and redirects. Referral exclusions alone cannot repair continuity.

Lesson 08

Reading the standard reports

Five views are enough to start. Realtime tests collection. Traffic acquisition describes session sources and their outcomes. Pages and screens compares content use. Events shows actions. Tech and user attributes help you inspect device and broad geographic patterns.

Use Session source / medium alongside channel groups. Compare key-event counts and rates with their denominators. In Pages and screens, consider views, engagement time and next actions together. Engagement time measures page focus.

• Compare similar page purposes: a quick contact page can succeed while a long tutorial holds attention.
• Start with twenty-eight days and compare an equivalent period; use year-over-year comparisons for seasonal work.
• Check real mobile behaviour. City is approximate, and your client market cannot be diagnosed from a top-three ranking.

Lesson 09

Funnels: where the route breaks

A funnel checks a journey you define. In Explore → Funnel exploration, name the steps and specify events or page conditions. A closed funnel requires entry at the first step; an open funnel allows entry later. Choose the form that answers your question, not whichever produces larger numbers.

For an inquiry route, try arrival → work viewed → contact opened → confirmed request. For a gallery route, try portfolio page → handoff → tagged gallery page. For educational content, try section viewed → resource copied or downloaded.

• Specify sequence, direct or indirect follow, and time constraints. Check whether the journey should remain within one session.
• Break down by device or a suitable dimension, and inspect drop-off segments.
• The drop-off reader practises a closed funnel. Its largest percentage loss suggests a place to investigate; it cannot diagnose motivation.

Lesson 10

Paths: where they wander

A path exploration examines recorded routes rather than a prescribed sequence. Open Explore → Path exploration, choose Start over, and select a starting event or page. Expand nodes to see what followed. Page paths help distinguish content; event names reduce clutter.

For a backward path, choose an ending point such as generate_lead or contact_click. Inspect preceding pages for navigation hypotheses.

• Paths can span sessions and aggregate many event streams.
• Use a relevant segment and a longer period when traffic is sparse.
• An exit page is not automatically a failure: someone may have found your opening hours or left to call.

Lesson 11

Segments, audiences and comparisons

Comparisons split a report into groups. Exploration segments select users, sessions or events within the available analysis period. Choose user scope for a journey across visits and session scope for behaviour during a visit; avoid treating those as interchangeable.

Try four practical groups: people who explored contact or a gallery, visitors in your service region, users with sustained measured engagement, and returning visitors. These groups do not establish buying intent.

• Add conditions under Segments in an exploration, name the group and apply it to the analysis.
• Audiences maintain membership rules over time and support reporting as well as advertising integrations.
• Audience reporting differs from retroactive segment analysis; some integrations can prepopulate lists from eligible recent data. Create audiences when you have a purpose.

Lesson 12

The weekly dispatch

Create a small report you will actually read: sources, attention and confirmed inquiries, with contact attempts reported separately. Administrators can schedule standard and custom GA reports through Share this report → Schedule Email. Select authorised recipients, a date configuration, PDF or CSV, and a useful cadence.

Allow for processing delay; scheduled-report guidance allows up to seventy-two hours. Review expiry and status in Admin → Scheduled emails. Saved eligible funnel reports offer another reporting route; an exploration is not automatically a recurring email.

• A one-page Looker Studio dashboard can combine report fields, but its GA connector does not reproduce explorations, segments or comparisons.
• Use custom insights for meaningful anomalies and property-user notifications. Tune thresholds for your volume.
• Use the weekly template to record observations, one action and its review date.

Lesson 13

Origin trace: UTM discipline

Referrers can be absent from apps, messages, privacy settings and redirects. Direct traffic means Analytics lacks a clear referral source; it is not a reliable count of people typing your address. Attach consistent UTM labels to external campaign links you publish.

Use source for the platform, medium for the channel, campaign for the initiative, content for the placement or creative version, and term when keyword detail is relevant. Prefer social or email as the medium and bio, story or image_a as content.

• Keep a naming register and use consistent lowercase values.
• Avoid campaign tags on internal navigation: GA4 does not start a new session, but campaign data can still confuse interpretation.
• The link builder preserves existing query parameters and anchors. Read results in Session source / medium and Session campaign.

Lesson 14

Traps and misleading signals

Measurement is incomplete and its mistakes are not always undercounts. Before acting, inspect collection, volume and alternative explanations. There is no universal visitor-loss percentage, minimum sample or conversion benchmark that makes a report trustworthy.

Separate focus time from interest and contact intent from confirmed business. Attribution depends on dimension scope: session acquisition uses session attribution; event-scoped key-event credit follows the property’s reporting model. Neither reveals the full history of every customer.

• Collection traps: your own testing, blockers, consent choices, missing or duplicate tags, and fragmented cross-device identities.
• Reading traps: small samples, missing values, high-cardinality other rows, thresholding, mismatched tools, mixed audiences, and averages across different page purposes.
• Check the data-quality indicator, unusual bot or referral activity, and attribution scope. Modelling and User-ID have conditions.
• Avoid optimising only what is measurable. Keep making valuable work that has no immediate metric.

Lesson 15

The ritual

Spend twenty minutes each week reviewing a consistent window. Read the dispatch, separate inquiries from contact clicks, check source changes and comparable pages, and review safe zero-result search topics. Write what you learned and what you will investigate. Doing nothing is a valid decision when evidence is weak.

Monthly, inspect the inquiry funnel and backward paths, compare your service-region group, audit new pages and handoffs, and test the newest page on a phone. Make one focused change and record the expected result.

• Quarterly: review scheduled-email access and expiry, event usefulness, internal-traffic rules, retention and seasonal comparisons.
• Keep a dated decision log with the change, reason, measurement and review date.
• Reset checklists when starting a new review cycle. Browser progress records learning, not completion of your actual GA setup.

Lesson 16

Decoder

Dimensions describe; metrics measure. Parameters describe events. Scope identifies event, session, user or item context.

A property contains collected data; a stream feeds it; a Measurement ID points the web tag at its stream. DebugView checks test events. An exploration asks a custom analytical question. A funnel checks a sequence; a path examines recorded routes.

• Filter both term names and definitions; clear the search to restore every term.
• Read attribution, audience, cardinality, direct traffic, engagement, retention, thresholding and UTM definitions before interpreting them.
• Consult the sources for changing interfaces and limits.

Lesson 17

Clearance exam

The ten-question clearance exam tests decisions: distinguish contact attempts from leads, choose backward paths, recognise small-sample claims, and explain what tags and UTMs measure.

Answer every question, then submit for a score and an explanation for each answer. Revisit modules and retry. A passing practice score is a learning milestone; it is not professional certification or proof that your tracking is correct.

• Seven out of ten earns practice clearance; every question remains reviewable.
• Resetting the exam clears answers and feedback, with no effect on your analytics account.
• Write one question your data answers and one it cannot.