← All field notes
Operating evidenceField observation / July 20263 min read

Analytics belong in the release plan—not the cleanup list

For a conversion-focused release, loading the page is only the start. The team also needs to see whether the intended path works, where it fails, and what to change next.

A technically successful launch can still be operationally blind

A site can launch quickly, meet accessibility standards, and produce no visible errors while still leaving the business unable to answer a basic question: did visitors complete the action the site was built to support? For a conversion-focused release, that is a significant gap.

The sequence is understandable: teams prioritize copy, design, forms, infrastructure, and the production cutover, then postpone analytics because it feels secondary to the launch. The gap becomes visible as soon as someone asks whether a campaign produced qualified visits, a form submitted correctly, or visitors reached the intended page. At that point, the team has no reliable answer.

Define the decision before defining the event

Instrumentation is useful only when it supports a decision. A page-view count may confirm traffic, but it cannot explain whether the right visitor reached the right destination or completed the intended action. Start with the decision the team will need to make, then work backward to the smallest set of events and details that can support it.

  • What action represents meaningful progress rather than passive activity?
  • Which source, campaign, or referral path needs to survive through the journey?
  • What does success look like in both the browser and the receiving workflow?
  • Who reviews the signal, and what will they change when it moves?

Validate the whole path, not only the tag

A firing browser event does not prove that the conversion worked. The form must accept the submission and show an accurate success state. Campaign attribution must survive the handoff, the receiving team must get enough context to act, and the analytics platform must record the event exactly once.

Release validation should follow one real submission across every handoff: redirects, campaign parameters, production configuration, the live form, its delivery destination, the analytics event, and the report where that event will appear. Individual components may pass their own checks while the journey still breaks between them.

Keep the measurement proportional

Not every event needs a dashboard, a monetary value, or a complicated attribution model. Artificial precision creates its own form of blindness. If the business has not established a defensible value for a lead, do not invent one to make the report look mature.

A small, trusted measurement system is often more useful than a large event catalog the team does not review consistently. Measure enough to confirm that the journey works, locate failures, compare meaningful sources, and decide what to change next.

Andrew Erie leads Lavigne, joining important technology and AI projects at any stage and carrying them through delivery.

This page was first published July 31, 2026. The field date identifies when the underlying observation was recorded.

Bring the decision

What decision keeps coming
back to your desk?

Start with the technology, AI, product, vendor, or delivery decision that no one fully owns today.

Start a conversation