Fewer numbers · better decisions

Most analytics measures everything and tells you nothing.

There is a dashboard somewhere in your business with forty tiles on it, and nobody has changed a decision because of any of them. Sessions went up. Bounce rate went down. Nobody knows whether that was good, or what to do next.

Useful measurement starts from the other end — with the decisions you actually make, then working backwards to the smallest set of numbers that informs them. Usually that is four or five figures, not forty, and half of what most sites track can be switched off without anyone noticing.

Hover a source to trace it
all trafficenquiry rate
Starts fromYour decisions
TracksOutcomes, not clicks
Reported asOne page
Honest aboutWhat it cannot know

Diagnosis

Six reasons your reports go unread

Almost every analytics setup we inherit has at least four of these. None of them are exotic and all of them are fixable in days rather than months.

01

Nothing is a conversion

Traffic is measured; outcomes are not. Nobody knows how many enquiries came from the site last month, so every other number is uninterpretable.

Fix: define what counts as a result, then track that.

02

Your own visits are counted

Staff, the agency, the developer refreshing a page thirty times. On a low-traffic site this is not a rounding error — it can be a meaningful share of everything reported.

Fix: internal traffic excluded, bots filtered.

03

The form is not tracked properly

A submission counted on page load rather than on success, or counted twice, or a thank-you page indexed and visited directly. The number is wrong and nobody has reconciled it.

Fix: reconcile against what actually reached the inbox.

04

Phone calls are invisible

For a great many local businesses the main outcome is a phone call, and it appears nowhere in analytics. The channel driving calls looks like it does nothing.

Fix: call tracking, or at minimum tap-to-call events.

05

Everything is averaged

One conversion rate for the whole site. Mobile and desktop, homepage and landing pages, brand and non-brand traffic all blended into a single figure that describes nobody.

Fix: segment before drawing any conclusion.

06

The report has no verdict

Forty charts and no sentence saying what happened, what caused it and what to do. A report that requires interpretation will not get interpreted.

Fix: a written conclusion, first, in plain language.

Interactive

One customer, six models, six different heroes

A real-shaped journey: eighteen days, six touchpoints, one enquiry. Switch the attribution model and watch the credit move between channels — same customer, same behaviour, entirely different conclusion about what worked.

Look at what happens under last-click, which is the default in most reports and the basis of most budget decisions. All the credit goes to Direct — a channel you cannot buy, cannot scale, and which in this case simply means the customer already knew your name because of something earlier in the list.

the journey18 days · 6 touchpoints · 1 enquiry

Interactive

Start from the question, not the dashboard

Six questions businesses actually ask. For each: the number that answers it, and the number people reach for instead.

The right-hand column is not made up of useless metrics. They are all real, all reported prominently, and all the wrong tool for that particular question — which is how a business ends up confidently acting on a figure that was never about what they thought.

Coverage

Whatever you already have, made to answer something

Google Analytics is not the problem and is rarely configured. Where privacy or simplicity matters more than depth, a lighter platform is often the better answer, and we will say so.

Analytics platforms

Google Analytics 4
Plausible
Matomo
Fathom
PostHog
Mixpanel

Tagging & behaviour

Tag Manager
Search Console
Hotjar
Cloudflare Analytics
WordPress
Shopify

Channel & campaign data

Google Ads
Meta Ads
LinkedIn
Instagram
TikTok
Mailchimp

Warehouse & dashboards

BigQuery
dbt
Looker
Looker Studio
Metabase
PostgreSQL

The work

Six things worth measuring properly

Short list on purpose. Each one exists because a decision depends on it.

Measure 1 of 6
01 — Outcomes

Enquiries, calls and orders

The actual results, counted accurately and reconciled against what reached your inbox, phone or order system — not what a tag thinks happened.

  • Form submissions verified against inbox
  • Calls tracked, not assumed
  • Duplicates and test entries removed

Everything else is meaningless without this

02 — Source quality

Which channels produce customers

Not which produce sessions. A channel sending a tenth of the traffic and a third of the enquiries is the one to spend more on, and raw session counts hide that completely.

  • Conversion rate by source
  • Assisted paths, not just last click
  • Segmented by device

Where budget decisions actually come from

03 — Page performance

Which pages do the work

Entry pages ranked by what happens after them. Usually two or three pages carry most of the outcomes and a long tail of pages carry none, which tells you where to spend effort.

  • Conversion by landing page
  • Drop-off points identified
  • Compared like for like

Frequently reorders the content plan

04 — Search visibility

What you are found for

Impressions, clicks and position by query group from Search Console — the only place with real search data, and a source most businesses have connected and never opened.

  • Grouped by intent, not query strings
  • Brand separated from non-brand
  • Movement tracked over quarters

Free, already collecting, rarely read

05 — Experience

Speed and friction, on real visits

Core Web Vitals field data and where people abandon forms. Both change conversion measurably and neither shows up in a standard traffic report.

  • Field data by page type
  • Form abandonment by field
  • Errors that visitors hit silently

Connects to speed and conversion work

06 — Trend

Direction over quarters, not days

Month-to-month noise causes more bad decisions than almost anything else. The useful view is a rolling comparison against the same period last year, with the known causes annotated.

  • Seasonality accounted for
  • Changes annotated on the timeline
  • Reported quarterly as well as monthly

Stops reactions to random variation

Honesty

What analytics genuinely cannot tell you

Web measurement is less precise than the interfaces imply, and a report that admits its limits is worth more than one that does not.

01

Your numbers are an undercount

Ad blockers, privacy settings, consent declines and browser restrictions all remove visits from the data. Every analytics figure is a sample, and the sample is not random.

02

Attribution is a modelling choice

As the panel above shows, the same journey supports six different conclusions. Pick a model, understand what it over-credits, and stay consistent so the trend is meaningful even if the level is not.

03

Cross-device breaks the chain

Someone finds you on a phone at lunch and enquires from a laptop that evening. Without a login, those are two unrelated people as far as the data is concerned.

04

Small numbers are mostly noise

Twelve enquiries versus nine is not a trend, and a test on a few hundred visits proves nothing. A great deal of confident marketing analysis is reading randomness.

05

It cannot tell you why

Analytics describes what happened. Why it happened comes from talking to customers, watching session recordings and asking the sales team — none of which is in a dashboard.

06

Consent has consequences

Under Canadian privacy law you need a lawful basis and meaningful notice for tracking. Doing this properly reduces the data you collect, and that trade-off is not optional.

Scope

What a measurement engagement includes

The first session is a conversation about decisions, not tools. Everything after that follows from it.

01

Decision mapping

What choices you actually make — where to spend, what to write, what to fix — and what evidence would change each one. This determines everything else.

02

Tracking audit

What is currently collected, what is broken, what is double-counted and what is quietly measuring your own staff.

03

Conversion setup

Every meaningful outcome tracked properly — forms on success, calls, bookings, orders — and reconciled against the source system.

04

Consent and privacy

A consent mechanism that works, configured so declining is respected, with a written note of what is collected and why.

05

Channel connections

Ad platforms, Search Console and email joined up so channel performance can be compared on the same basis.

06

The one-page report

A single view carrying the four or five figures that matter, with last period alongside and a written conclusion at the top.

07

Annotation discipline

Campaigns, launches, outages and price changes marked on the timeline, so next quarter's spike has an explanation attached.

08

Handover and training

Your team taught to read it and to answer the routine questions without calling anyone.

09

Review rhythm

A scheduled conversation about what the numbers changed. A report nobody discusses is a report nobody needed.

Questions

Analytics, answered

We already have Google Analytics. Isn't that enough?

You have the collection. Whether you have measurement depends on whether conversions are defined, internal traffic is excluded, forms are tracked on success and calls are counted at all.

In most audits we find at least three of those missing, which means the data has been accumulating for years without being answerable.

Is GA4 worse than the old version?

Different, and genuinely less intuitive for simple questions. It is also more capable once configured, and free at a scale most businesses will never exceed.

If your needs are modest and privacy matters, a lighter platform may suit you better. We would rather set up the simpler thing well than the complex thing badly.

Can you track phone calls?

Yes, and for local service businesses it is often the most valuable single addition. Dynamic numbers attribute calls to the channel that produced them.

Cheaper alternative: track taps on the phone link, which undercounts but is free and still tells you which pages generate calls.

How often should we look at this?

Monthly for operational figures, quarterly for direction. Weekly reporting on a site with modest traffic mostly produces reactions to noise.

The exception is during a campaign or after a change, where a short intense look is warranted and then should stop.

What about cookie consent?

You need a lawful basis and meaningful notice, and consent tools should actually prevent tracking when someone declines rather than just displaying a banner.

Configuring it properly reduces your data. That is the correct outcome, and the reduction is smaller than most people fear.

Can you build us a dashboard?

Yes, and we will push for a small one. A single page with four or five figures, last period alongside, and a written conclusion beats a wall of tiles.

If a dashboard needs explaining every time it is opened, it is not finished.

Our numbers dropped and we do not know why.

Check the tracking first — a surprising share of dramatic drops are a broken tag, a consent change or a site update that removed a script.

If the data is sound, segment by channel, device and landing page. The drop is almost always concentrated somewhere specific rather than spread evenly.

Do you do this as an ongoing service?

The setup is a defined project. After that some clients want a monthly report and a conversation; others take the one-page view and run it themselves.

Both are fine. The second is a legitimate outcome and we will hand over everything needed for it.