blog single

Analytics that answer a question

Analyst reviewing a small set of website metrics

Published on:

July 17, 2025

TAG:

ANALYTICS

BY:

Edgeless Team

Read Time:

5 MIN READ

post detail

Most websites have more data than they use. A tag manager sends events to two platforms, a dashboard shows thirty charts and nobody can say which number changed a decision last quarter. The problem is not a lack of tracking. It is the absence of a question that the tracking was designed to answer.

Start with a decision

Write down the decisions the website is supposed to support. Should we invest more in the paid campaign or the organic content. Is the new contact form better than the old one. Which service page deserves the next case study. Each decision needs one or two numbers, not a dashboard.

A decision also needs a time frame and an owner. A number reviewed monthly by the marketing lead is different from one reviewed weekly by the sales manager. Without an owner, a metric becomes a report. With an owner, it becomes a conversation with a next step.

Keep the list short. Five decisions cover most marketing sites for a year. When a new question appears, retire an old one or add it deliberately. A measurement plan that grows without limit becomes a maintenance burden and a source of disagreement about definitions.

Events before dashboards

Define the actions that matter as events with clear names and parameters. A form submission, a phone click, a booking, a document download, a quote request. Use one naming convention and document what each event means, including when it does not fire.

Avoid tracking everything because it is easy. Every event adds noise and a maintenance obligation, and it makes the important actions harder to find. A small set of well defined events is more useful than a complete record of every click.

Name events after the action, not the interface. A name like submit form 2 will be meaningless after a redesign. A name like quote request describes what happened and survives a new layout. Use the same vocabulary as the business so the report does not need a translation layer.

One event per meaningful action, with a name the team already uses.
Parameters for the page, the service and the source where relevant.
A documented definition for each event, including exclusions.
A test plan that confirms each event fires once, not twice.
A named owner for the tracking implementation.
A quarterly review of events that nobody uses.

Attribution is imperfect, and that is fine

No model gives a complete picture of why someone enquired. Last click credits the final search. First click credits the article they read a month earlier. A consent banner blocks some visits entirely. The goal is not perfect attribution. It is a consistent model that helps compare channels over time.

Pick one primary model and keep it stable. Changing models mid year makes every comparison invalid. If you need a second view, add it as a secondary report rather than replacing the first.

Record what the model cannot see. Offline conversations, referrals and returning customers who type the address directly all contribute. A short note in the monthly report is more honest than a chart that implies precision the data does not have.

Small dashboard focused on conversion metrics — One screen with the numbers tied to a decision is more useful than a dashboard with every available metric.

Dashboards people actually read

Put the decision numbers at the top and everything else below. Use comparisons that matter: this month against the same month last year, this campaign against the previous one, this template against the old one. A number without a comparison is hard to act on.

Annotate changes. A launch, a campaign, a holiday or a tracking fix should appear on the chart with a date and a short note. Without annotations, every spike becomes a mystery and the team spends the meeting reconstructing events from memory.

Send a short summary rather than a link to a dashboard. Three sentences on what moved, why it probably moved and what happens next is more likely to be read than a login prompt. The dashboard is the backup for anyone who wants detail.

Privacy and consent

Collect only what the measurement plan needs. That usually means page level data, event data and a coarse source, not personal identifiers. Check that the consent tool blocks tracking before consent and that the data retention setting matches the privacy notice.

Document the data flow: which script, which platform, which fields and which retention. A data flow diagram is useful during a privacy review and during a tracking audit, and it takes an hour to write.

If the business uses a customer data platform or a CRM, connect the website events to the record with a stable identifier. That connection turns enquiries into a pipeline view, which is usually the number the sales team cares about.

Campaign tracking without chaos

Agree on a small set of campaign parameters and a naming convention before the next campaign starts. Source, medium, campaign and content cover most needs. The same campaign should use the same name everywhere, including in the ads platform, the email tool and the link in a partner newsletter.

Keep a register of active campaigns with their dates and owners. When a campaign ends, its traffic stops but its parameters live on in bookmarks and old posts. A register makes it easier to separate a current campaign from a stray link that still sends traffic.

Check the landing pages, not only the campaign totals. A campaign can deliver traffic that bounces because the page does not match the promise. The landing page report tells you whether the problem is the audience or the page, and those need different responses.

The monthly review

Hold one short review each month with the people who can act on the numbers. Look at the decision metrics, the annotations and the events that changed. End with one action and one question for the next month. The action is what turns measurement into improvement.

Keep a running log of what was tried and what happened. Over a year, that log is more valuable than any dashboard because it records the reasoning behind the numbers. It also prevents the team from repeating an experiment that already failed.

Invite one person from sales or operations to the review. Their questions are usually closer to the business than the marketing team's assumptions, and their presence makes the numbers part of the weekly conversation rather than a monthly report nobody opens.

A metric earns its place when someone can name the decision it informs. Everything else is a hobby.
Edgeless Team

A one page measurement plan

Write the plan on one page: the decisions, the metrics, the events, the owner, the review date and the known limitations. Share it with everyone who touches the site. A plan that fits on a page is easy to keep current and easy to challenge.

Review it every quarter. Retire metrics that stopped informing a decision, add the one that keeps coming up in meetings and check that the events still fire after the latest release. Tracking decays silently, and a quarterly check is cheaper than a year of bad data.

We design measurement plans, implement clean event tracking and build reporting that connects website activity to pipeline.

Explore digital marketing

The point of analytics is not to know more. It is to decide better. A small plan tied to real decisions will outperform a large dashboard every time.

Analytics

Measurement

Marketing

Analytics that answer a question

Let's Talk About Your Project