The Australian marketing data stack.
Modern marketing data infrastructure for Australian businesses. BigQuery as source of truth, Looker for visualisation, Vertex AI for analytics pipelines, NL2SQL for portfolio managers. The stack we run for PE-portfolio + multi-brand clients.
Six layers of the marketing data stack
Layer 1, Sources
- Google Ads accounts (multiple per portfolio)
- Meta Business Manager (multiple ad accounts)
- LinkedIn Campaign Manager (where applicable)
- GA4 (cross-domain across multi-brand portfolio)
- Klaviyo + HubSpot + Salesforce (CRM event data)
- Booking platforms (HotDoc, Best Practice, Cliniko for healthcare)
- Shopify / WooCommerce customer events
- Xero invoice data (for ad-spend pass-through verification)
Layer 2, Pipelines
Vertex AI pipelines + scheduled BigQuery transfers for ad-platform data + GA4 BigQuery export + Klaviyo / HubSpot / Salesforce CRM event piping. Refreshed daily, australian-region data residency where available.
Layer 3, Storage (BigQuery)
BigQuery as the source-of-truth data warehouse. Multi-region for redundancy where required. Schema design for portfolio-level + brand-level + campaign-level analysis. Cost optimisation via partitioning + clustering.
Layer 4, Models + transformations
dbt-style transformations for marketing-analytics-ready datasets. Attribution modelling (multi-touch where volume justifies). Cohort retention modelling. Anomaly detection (Vertex AI on spend velocity + conversion-rate trends).
Layer 5, Visualisation (Looker)
Looker Studio dashboards consolidating the BigQuery layer for portfolio-level + brand-level + executive-level views. Quarterly portfolio CEO review preparation runs against this layer. Daily refresh.
Layer 6, Self-service (NL2SQL)
Natural-language query interface against BigQuery for non-technical portfolio managers. "How many GP appointments did we book in Brisbane last month vs prior month?" → SQL → answer in seconds. Removes the agency-as-data-team-bottleneck.
What most Australian brands miss
Three layers most brands under-invest in:
- Layer 3 (BigQuery as source of truth). Most brands rely on platform-native dashboards stitched into a deck. Without the BigQuery layer, attribution is platform-locked + can't be audited cross-platform.
- Layer 4 (transformations + modelling). Most brands skip the attribution-modelling + cohort-retention modelling layer. The data exists in BigQuery; the models that turn it into operational intelligence don't.
- Layer 6 (NL2SQL self-service). Most brands have data team bottlenecks because non-technical portfolio managers can't query the data directly. NL2SQL changes this.
Read more:
Want a serious marketing data stack?
Book a Discovery