Technical scope

The home page explains why. This page explains what. It is written for someone who wants to check the depth before writing a first email – if that is not you, skip it.

This describes method, not specific engagements. The details of client projects – their architecture, the problems found and the results of audits – stay with them.

Campaigns: account architecture, not just optimisation

Google Search, display remarketing, video, Meta Lead Generation, Performance Max – run and optimised across multiple European markets.
Consolidating campaigns where fragmentation degrades the signal automated bidding depends on.
Matching the bidding strategy to real conversion volume rather than to whatever the platform recommends.
Managing tCPA, keyword and negative structure, and budget allocation by measured efficiency.
Scaling decisions based on CPA across several time windows and on CPM, CPC and CTR trends – not on a single week.
Catching the methodological faults that quietly distort campaign assessment: duplicated spend, search and display mixed in one number, badly defined KPIs.

The warehouse as the single source of truth

BigQuery as the place where GA4, ad platforms, CRM, payments and production data meet.
Queries, views and mart layers built so you can trace a path from ad spend to actual revenue.
Reconciliation: comparing what a channel reports against what actually happened in the system.
Quality control – data freshness, deduplication, deterministic joins, and independent validation run directly in the warehouse.

Tracking and conversion measurement

Google Tag Manager, server-side GTM, GA4, Google Ads, Meta Pixel and Conversions API.
Difficult environments: multiple subdomains, single-page applications, measurement driven by history changes.
Custom conversion events, dynamic lead values, Advanced Matching, Conversion Linker, improving Event Match Quality.
Cleaning up UTM conventions – wrong separators, mismapped sources and hardcoded identifiers distort reporting in ways only visible once the data reaches the warehouse.

Attribution

Multi-touch models joining ad, product, CRM and subscription data.
Attribution model changes run in parallel with the existing one and validated against data before replacing it. Never swapped overnight.
Channel classification and cross-device identity resolution – particularly where in-app browsers anonymise the click.

Data pipelines and closing the loop

Production ETL and reverse-ETL pipelines connecting ad platforms, CRM, payments and the warehouse.
Offline conversions: sending valuable CRM events and closed deals back to the ad platforms through Enhanced Conversions and the Conversions API.
Deterministic identity resolution on CRM identifiers and addresses – rather than probabilistic matching on company name, which looks like it works right up until the first audit.
Audience pipelines segmented by sales pipeline status, with employees, test accounts and existing customers excluded automatically.

Production deployment

Rollback snapshots, versioned builds, independent data validation after release, changes synchronised with the repository.
Irreversible operations guarded by dry runs, assertions and an explicit activation step.
This part is usually missing from marketing work – and it is the part that decides whether a change to measurement can be undone once it turns out to be wrong.

Reporting

A reporting layer joining ad, product, CRM and financial data – from spend through to revenue.
Reports that state the limits of the data openly: traffic and conversions that cannot be attributed, because of missing consent, blockers or technical constraints, shown separately rather than absorbed into a total.
Reconciliation of channel figures against the real number of accounts or transactions, so the gap is visible instead of hidden.

Consent, privacy and data governance

Consent gating for marketing tags, removing stray cookie-writing code, verifying tracking behaviour in a clean session.
Auditing tag containers for what actually fires, and when.
On the data side: reviewing cloud permissions, catching over-privileged service accounts, taking inventory of personal data held in the warehouse.
Data lifecycle – whether a deletion in the source system propagates through to the analytics layer, and where pseudonymisation can be applied.

Tooling

Advertising: Google Ads · Meta Ads
Analytics and tagging: GA4 · Google Tag Manager · server-side GTM · Meta Pixel · Meta CAPI
Data: BigQuery · ClickHouse · SQL
CRM and revenue: HubSpot · Stripe
Cloud and pipelines: Google Cloud Platform · Cloud Run · Cloud Storage · ETL and reverse-ETL
Measurement: attribution modelling · offline conversions · Enhanced Conversions · Customer Match
Governance: consent management · IAM · data quality · PII governance
Engineering: Git · production deployment · validation · rollback · monitoring

If any of these describes a problem you have, that is a good place to start a conversation.

Get in touch