Analytics & Strategy

Why measurement has to come first

Most analytics problems are not really analytics problems. They are measurement problems wearing a dashboard. If the underlying data is incomplete, inconsistent, or collected without consent, then every report, insight, and strategic call built on top of it inherits those faults. That is why we treat measurement as the foundation rather than an afterthought.

Trustworthy measurement means a few specific things in practice:

• Events fire once, reliably, and mean the same thing across every page and platform.
• Key actions (enquiries, purchases, sign-ups, calls) are defined the same way in your analytics, your ad platforms, and your internal reporting.
• The numbers can be reconciled against a source of truth, such as your CRM or billing system, rather than taken on faith.
• Anyone reading a report can trace a figure back to how it was collected.

When this groundwork is solid, the rest of the work becomes faster and less contentious. Teams stop arguing about whose number is correct and start discussing what to do about it.

What a clean setup actually involves

Analytics tooling has shifted considerably. Universal Analytics was retired by Google in 2023, and GA4 is now the standard Google Analytics property. GA4 uses an event-based data model rather than the older session-and-pageview model, which gives more flexibility but also means a thoughtless migration tends to produce messy, hard-to-use data. A clean GA4 build is not a default install; it is a deliberate event and parameter design that matches how your business actually works.

A well-structured setup typically includes:

• A documented measurement plan that lists each event, its parameters, and the business question it answers.
• GA4 configured with clear conversion (key) events, sensible naming, and filtered internal and bot traffic.
• A tag management layer (commonly Google Tag Manager) so changes can be made and tested without code releases for every tweak.
• Server-side tagging where it adds value, moving tag execution to a server container to improve data quality, reduce reliance on browser conditions, and give you more control over what data is shared with third parties.
• Consent handling wired in from the start, so tags respect a visitor's choices and data collection aligns with privacy obligations.

The goal is a setup that is legible. Six months later, someone should be able to open the configuration and understand it without reverse-engineering it.

Turning data into decisions people read

A dashboard nobody opens is a cost, not an asset. The common failure is reporting that shows everything and explains nothing: dozens of metrics, no narrative, and no indication of what changed or why it matters.

We build reporting around the decisions it is meant to support. That usually means different views for different audiences:

• A leadership view that answers "are we growing, where is the money coming from, and what needs attention" in a single screen.
• An operating view for marketing and product teams with the detail needed to act, such as channel performance, funnel drop-off, and campaign results.
• A diagnostic layer for analysts to investigate when a headline number moves unexpectedly.

Good reporting also states its assumptions. It notes the date range, the attribution approach, and any known data gaps, so readers are not misled by a clean-looking chart that hides a caveat. Where it helps, we add short written commentary alongside the numbers, because a sentence of context often prevents a wrong conclusion better than another chart.

Proving what works through experimentation

Opinions about what will improve performance are cheap and plentiful. Evidence is harder, and more valuable. Experimentation is how you separate changes that genuinely move the needle from changes that merely feel productive.

A disciplined approach to testing looks like this:

• Start from a clear hypothesis tied to a metric you can measure, not a vague hunch.
• Decide in advance what success looks like and how long you will run the test.
• Run a fair comparison, whether that is an A/B test, a phased rollout, or a before-and-after analysis when a controlled split is not possible.
• Read the result honestly, including the tests that fail or come back flat, because those save you from scaling a bad idea.

Not everything can or should be tested with a formal experiment. Low-traffic pages, long sales cycles, and one-off campaigns often need other methods, such as cohort analysis or holdout groups. The point is to be honest about what the data can and cannot prove, and to make decisions accordingly.

The strategy layer that ties it together

Measurement, reporting, and experimentation are inputs. Strategy is where they connect to commercial goals and to the other channels you are investing in. Without that layer, you get accurate data that nobody knows how to act on.

The advisory work focuses on a few questions:

• Where is the next unit of effort or budget best spent, given what the data shows?
• How do paid, organic, content, email, and on-site experience reinforce or undercut each other?
• Which metrics genuinely indicate progress, and which are vanity numbers that look good but change no decisions?
• What is the sequence of work over the coming quarters, and how will we know it is working?

This is also where analytics stops being a standalone service and starts coordinating the rest. The same measurement that proves a campaign worked should inform the next content brief, the next round of SEO priorities, and the next budget conversation. Strategy keeps those threads aligned so the channels pull in the same direction rather than competing for credit.