Service
Marketing analytics services
Measurement you can act on, and an honest account of what it cannot tell you.
Analytics work has two halves. The first is technical: events firing correctly, deduplicated, with the right parameters, server-side where the browser is unreliable. The second is interpretive: knowing which numbers can carry a decision and which cannot. Most reporting problems are actually the second half, dressed up as the first.
The problem
Why does nobody act on the reports?
Because a dashboard listing twenty metrics implies no decision. Underneath, tracking is usually part broken, and attribution gets asked questions it structurally cannot answer.
The most common analytics problem is a dashboard that lists what happened without implying anything to do. Twenty tiles, no decision. It gets built because it is easier to add a metric than to argue for which three matter.
Underneath that, tracking is usually part broken in ways nobody has checked: a conversion event firing twice on some browsers, a key event that counts page views instead of submissions, or a cross-domain handoff dropping the session so a third of conversions land under direct.
The third problem is expecting attribution to answer questions it structurally cannot. No model can tell you what a channel caused, only what it touched. Treating a last-click report as a causal claim is how good channels get cut.
Scope
What do marketing analytics services cover?
GA4 implementation, server-side tracking, validation against real interactions, a stated attribution model with its biases written down, decision-oriented reporting, and consent configuration.
Implementation, validation, and interpretation, in that order.
GA4 implementation
Event taxonomy designed before anything is tagged, key events defined against real business outcomes, and parameters chosen so the reports you need are possible later.
Server-side tracking
Server-side tagging or Conversions API where browser tracking loses meaningful volume, with deduplication verified rather than assumed.
Tracking validation
Every event tested against a real interaction on real devices. An event that fires in the tag debugger and nowhere else is a common and expensive illusion.
Attribution
A stated model, its known biases written down, and cross-checks against the platforms and your CRM. We reconcile the differences rather than presenting whichever source flatters the work.
Reporting
A small number of reports built around decisions: what to spend more on, what to fix, what to stop. Each metric has to earn its place by changing an action.
Consent and privacy
Consent mode configured properly, so the analytics setup is lawful and the data loss from consent choices is understood rather than discovered later.
Process
How does an analytics engagement run?
Audit what is firing, design the event taxonomy before tagging, implement and verify on real devices, build few reports, then review the model quarterly.
Audit what is firing
A full inventory of current events and conversions, and a test of each one against a real interaction. This normally finds at least one significant error.
Design the taxonomy
Naming and parameter conventions agreed before implementation, because renaming events after the fact orphans your history.
Implement and verify
Tagging built, then validated end to end, including the server-side path and the deduplication logic.
Build the reports
Few reports, each tied to a recurring decision, each with the caveats stated on the report itself rather than in a footnote nobody reads.
Review the model
Attribution and reporting revisited quarterly, because channel mix and tracking conditions both change underneath a model that stays still.
Coverage
What does the audit behind this work look at?
Every engagement opens with the audit, and these are the first of the areas it reads before anyone recommends a single change.
GA4 property configuration and key events
Data streams, key events, and whether the property is set up to answer the questions the business actually asks of it. A default install answers almost none of them.
Google Ads conversion actions, sources and attribution
What the ad account counts, where each action comes from, which are Primary, and what attribution setting is applied. These four decide what bidding optimises toward.
Conversion imports and offline conversion handling
Where the real outcome happens after the click, in a CRM or a phone call or a signed contract, whether it makes it back into the platform at all, and whether it arrives in time to be useful.
GTM container structure and hygiene
Tags, triggers and variables, including the ones firing on pages nobody remembers and the ones left paused since a migration. A container nobody has pruned is a container nobody can reason about.
Data layer implementation and event naming
Whether the site pushes its own events with consistent names, or whether triggers are scraping the DOM and waiting to break the next time the markup changes.
Consent mode and privacy compliance
Whether consent defaults are set before the tag loader runs rather than after it, whether choices are honoured, and whether the modelling that depends on all of it is actually working.
- 8 further areas are covered on the audit page.
All 14 are listed on the Conversion tracking audit page.
Deliverables
What do you receive?
An audit naming every broken event, a documented taxonomy, implemented tracking validated on real devices, a written attribution model with its biases stated, and reporting.
- An audit of every current event and conversion, with errors named
- A documented event taxonomy
- Implemented GA4 and server-side tracking, validated on real devices
- A written attribution model with its biases stated
- Decision-oriented reporting, not a metric wall
- Consent mode configured and documented
Start with the audit
Conversion tracking audit
What we read before recommending any of this, and what it covers. There is a free version if you would rather start there.
Related reading
GA4 conversion tracking that does not break
The long-form version of the thinking behind this service.
Related services
- Google Ads management
Accounts built to be measured, then managed against what the measurement says.
- AI marketing automation
Automate the repeatable, keep judgement where judgement is needed, and be specific about which is which.
Before you brief anyone
- What this kind of work costs
Third-party benchmark data, named and dated. We publish no rate card and the pages say why.
- Whether an agency is the right model at all
Agency against in-house and against a freelancer, each reaching a recommendation rather than a pitch.
- What is different in your vertical
Platform policy constraints, conversion definition problems and negative keyword themes by industry.
- What your own numbers support
Calculators for target cost per lead, in-house cost against an agency, and whether you can test at all.
Questions
What do buyers ask about marketing analytics?
Why GA4 and Google Ads disagree, whether server-side tracking is needed, whether existing GA4 can be fixed, and which attribution model to use.
Why do GA4 and Google Ads report different conversion numbers?
They use different attribution and different windows. Google Ads credits the ad interaction within its conversion window, and GA4 by default uses a data-driven model across all channels. Both can be correct at once. The mistake is comparing them as if they were measuring the same thing.
Do we need server-side tracking?
If a meaningful share of your audience uses blockers or privacy-restricted browsers, yes, because those losses are not evenly distributed and they skew what your bidding algorithms learn from. If you are running small spend on a niche audience, it is often not yet worth the complexity.
Can you fix our existing GA4 rather than rebuilding it?
Usually yes, and it is normally the better option because rebuilding costs you historical comparability. We rebuild only when the existing taxonomy makes the reports you need impossible to produce.
Which attribution model should we use?
Whichever one you understand the biases of. There is no correct model, only models with different distortions. We recommend picking one, documenting how it misleads, and cross-checking against your CRM for anything with a long sales cycle.
How do we know the tracking is actually right?
By testing it against real interactions on real devices and reconciling the totals against a source of truth outside analytics, normally your CRM or order system. A tag that fires in preview mode has proved almost nothing.
Start with an audit
Send us what you are running. We will tell you what we would look at first in Analytics, and whether it is the right place to start at all.