Measure what matters, then improve it.

• Start with analytics implementation when nobody trusts the numbers.

  If reports disagree with CRM or order records, or no key events exist, every later decision would rest on data that cannot be defended.

• Start with conversion optimization when measurement holds and visitors do not act.

  Reliable key events and steady traffic with few completions mean the question has moved from what happened to why it happened.

• Start with behavior analysis when you know where but not why.

  A funnel that shows the step where visitors leave, without explaining it, is exactly what recordings, heatmaps and usability review are for.

• Start with dashboards when data exists and nobody reads it.

  When the tools are configured but decisions are still made without them, the gap is reporting that answers the questions the team actually asks.

• Look to SEO, AEO and GEO when too few visitors arrive.

  Conversion work improves what happens after the visit. When the problem is being found in the first place, search visibility work comes first.

A website should track as conversions the actions that represent value to the business, such as a submitted inquiry, a booking or a purchase, and mark those as key events in Google Analytics. Supporting actions that lead toward them, such as starting a form or adding an item to a cart, are tracked as events that explain the path.

In Google Analytics, a key event is an event that measures an action particularly important to the success of a business. What Analytics once called conversions are now key events, while the word conversion remains in use in Google Ads, so keeping the two terms distinct avoids confusion when both reports are read together.

Lead tracking records that an inquiry happened and where it came from, never who sent it. The recommended Google Analytics event for when a visitor submits a form or a request for information is generate_lead, and it is sent after the site confirms the submission succeeded, so failed or rejected forms are not counted as leads.

E-commerce tracking follows the recommended Google Analytics events for a store, from viewing an item and adding it to a cart through beginning checkout to purchase. Each purchase carries a transaction ID, which helps avoid duplicate purchase events, and a currency whenever a value is sent.

Campaign links should carry UTM parameters written to one documented naming convention, with utm_source, utm_medium and utm_campaign on every tagged link. Values are kept lowercase and consistent, because Google Analytics treats parameter values as case sensitive, and internal links between pages of the site are never tagged.

Medium values do more than label a report. Google Analytics assigns traffic to its default channel group using rules on source, medium and, for some channels, campaign name, so a medium of email lands in the Email channel, while a value no rule recognizes can leave visits Unassigned. The naming convention is therefore written against those channel rules.

Internal links stay untagged, because campaign parameters exist to describe how a visitor arrived, not how they moved between pages once they were there. Where Google Ads auto-tagging is on, Google Analytics uses the auto-tagged values for source, medium and other traffic dimensions even when manual tags are also present.

• utm_source: where the visit came from, such as a newsletter or a named platform.
• utm_medium: the type of channel, such as email, cpc or banner.
• utm_campaign: the campaign, product or promotion, such as spring_sale.
• utm_content: which creative or link was clicked, such as one of two links in an email.
• utm_term: the paid keyword, for search advertising.
• utm_id: the campaign ID that identifies a specific campaign or promotion.

Landing page optimization changes the parts of a campaign page that decide whether a visitor acts: how closely the headline matches the advertisement or link that brought them, how clearly the offer is stated, whether one primary action stands out, and how much a form asks for. Each change is measured against the page baseline.

Building new landing pages belongs to web design and development; this work improves pages that already receive traffic. Campaign pages often sit on their own URLs, so a variant can be tested there without touching the pages that rank in search.

• Message match: the headline repeats the promise of the advertisement or link in the same terms.
• Offer clarity: what the visitor gets, what it costs and what happens next.
• One primary call to action, written specifically and repeated only where it helps.
• Forms that ask only for what the business will use, with visible labels.
• Proof placed where the decision is made, using only real evidence.
• Load speed and layout stability on the devices the campaign actually reaches.

The step where visitors drop out is found by defining the steps of a real journey as events, such as landing, starting a form and submitting it, and comparing how many visitors complete each one. Google Analytics funnel explorations show how well visitors succeed or fail at each step, and segments show which visitors lose intent.

Google Analytics distinguishes open funnels, where visitors can enter at any step, from closed funnels, where they must enter at the first step, and a funnel exploration can define up to 10 steps and apply up to 4 segments. A closed funnel describes a journey from its start; an open funnel also counts visitors who join part way through.

The last step usually sits outside analytics. Whether an inquiry became a qualified lead is recorded in the CRM, so a funnel is reconciled with those records rather than ending at the form. Connecting a website to a CRM is covered under integrations and automation.

• Visit: the landing page and the source the visitor arrived from.
• Engagement: reading or scrolling the content that answers their question.
• Intent: a click on the primary call to action.
• Form start: the first interaction with the inquiry form.
• Submission: the form sent and confirmed by the site.
• Lead: the inquiry recorded in the CRM or inbox and followed up.

Heatmaps and session recordings show what visitors do on a page, not only how many do it: where they click, how far they scroll, where they hesitate and which elements they try to use. Analytics shows that visitors leave a step; behavior tools suggest why, which turns a drop-off into a testable hypothesis.

Recordings show behavior, not intent, so they are watched in groups selected from a funnel step or a segment rather than at random, and a pattern is confirmed in analytics before it becomes a hypothesis.

Privacy settings are configured before recording starts. Some recording tools mask sensitive content by default, mask input box content in every masking mode, and never upload masked content. Whichever tool is used, it is set up so form entries, contact details and account pages are not captured.

• Click heatmaps: the elements visitors click, including ones that are not links.
• Scroll maps: how far down a page visitors get before they leave.
• Session recordings: the path through a page, including hesitation and repeated clicks.
• Form interaction: the fields visitors pause on, correct or abandon.

A monthly analytics report contains what the decisions it supports need: key events against their baseline and the same period last year, the channels, campaigns and landing pages behind them, search performance from Search Console, what changed on the site that month, what the data cannot show, and the actions agreed for next month.

A dashboard is built from KPI definitions agreed in writing, so a figure means the same thing every month. A Looker Studio report can combine Search Console and Google Analytics data, and because Search Console clicks and Google Analytics sessions are calculated differently, a good report explains why those two numbers never match.

Charts in a report carry a text summary of what they show, because complex images such as charts need a long description conveying the essential information. A report only its author can read does not change a decision.

• Traffic and acquisition: visits by channel, source and campaign.
• Key events and leads: totals and rates against the baseline.
• Funnel performance: completion at each step, by segment.
• Campaign performance: key events per campaign under one naming convention.
• Search performance: Search Console clicks, impressions and queries by page.
• Website performance: Core Web Vitals field data where enough of it exists.

Measurement turns into improvement through a loop of eight stages, from the visit that arrives to the report that decides what changes next. The loop is simplified, and a weak stage undermines every stage after it.

• A written measurement plan comes before any tag, and every report is built on the key events it defines.
• Google Analytics 4, Google Tag Manager and Google Search Console are the default stack, and other tools are added only when a question needs them.
• Personal data is kept out of analytics by design, because Google Analytics does not permit personally identifiable information to be sent to it.
• Consent is technical implementation of a decision the business makes with its legal adviser, never legal advice or a compliance promise.
• Every tag is justified against its cost to page performance, and unused tags are removed rather than left in place.
• Controlled tests run only where traffic can detect the expected effect, with the metric and stopping rule agreed first.
• No benchmark, uplift figure or client result is quoted, because a number from another business says nothing about yours.