Insights
Practical notes on reliable systems.
Marketing and business data, warehouse reliability, and API/workflow automation. Applied AI is a supporting theme.
Scoping a Useful First Version of Automated Business Reporting
Choose initial questions, define CAC and customer value, connect accessible sources, and agree a useful first reporting output without hiding its limitations.
Connecting BigQuery to Claude: Prepare the Reporting Layer Before MCP
A practical example of connecting AI to business reporting: shared metric definitions, scoped access, freshness, and answer checks over BigQuery data.
Recomputing BigQuery Reporting Models with Dataform After Inputs Change
Changed historical dates, attribution settings, refresh checkpoints, and safe replacement: practical recomputation design for BigQuery reporting models.
Recovering Missing Meta Ads Data in BigQuery: Reconnects, Pagination, and Backfills
How to identify incomplete advertising history, recover a bounded account/date scope, and validate source completeness before rebuilding reporting models.
Ads Data Hub: How User-Provided Data Matching (UPDM) Actually Works
A technical reference for Ads Data Hub user-provided data matching. Hashing requirements, match table architecture, match rate calculation using is_updm_eligible, privacy thresholds, EEA consent, and practical tips for improving match rates - sourced from Google's official documentation.
DV360 Reporting and BigQuery: Understanding the Three Data Paths
How Display & Video 360 data actually reaches BigQuery. The BigQuery API Connector, Reporting Data Transfer, and Bid Manager API serve different purposes - here is what each gives you and how to use them together.
Freelance Data Engineer vs Data Engineering Agency: What Marketing Teams Should Know
An honest comparison of hiring a freelance data engineer versus a data engineering agency for pipeline, warehouse, and reporting projects. When each makes sense.
Building a Multi-Agent Marketing AI System with LangGraph and Gemini
How to architect a LangGraph system that routes 40+ marketing performance rules across parallel AI domain workflows using a 3-tier classifier and Gemini.
Dataform vs dbt for Marketing Analytics: Differences That Actually Matter
A practical comparison of Dataform and dbt for marketing analytics warehouses on BigQuery. When to choose each, what the real differences are, and what does not matter as much as the articles say.
Why We Used a Deterministic State Machine Instead of an LLM for Issue Tracking
How to design a reliable issue lifecycle system for AI-detected marketing problems - and why the state transitions must not be LLM decisions.
Ad Creative Clustering with Embeddings, Ward Hierarchical, and HDBSCAN
How to group ad creatives into meaningful approach clusters using vector embeddings, hierarchical clustering, HDBSCAN, and approach ID lineage across periods.
How to Connect Google Ads and Meta Ads Data to BigQuery
A practical guide to moving Google Ads, Meta Ads, GA4, LinkedIn Ads, and DV360 data into BigQuery for marketing analytics. What each connection looks like and what breaks.
How a 6-Phase AI Pipeline Generates Creative Briefs from Ad Performance Data
Architecture of a production pipeline that turns BigQuery marketing performance signals into structured creative briefs through 6 sequential AI and deterministic phases.
What a Marketing Data Pipeline Project Costs
An honest breakdown of what marketing data pipeline projects cost, what drives the price up or down, and how to scope a project before asking for a quote.
Multi-Platform Ads Reporting to Google Sheets: What Breaks and How to Fix It
How to automate Google Ads, Meta Ads, LinkedIn Ads, and Bing Ads reporting into Google Sheets. Platform-specific challenges, rerun safety, and when Sheets stops being enough.
Prompt Guardrails: When the Data Model Is the Real Fix
How ambiguous evidence can confuse AI recommendations, and how clearer issue definitions, structured inputs, and output checks help.
Why Marketing Dashboards Break When Spend Is Modeled at the Wrong Grain
A practical note on campaign, ad, creative, order, and CRM grain problems in marketing warehouses, and how I model them before they reach dashboards.
What a Creative and Campaign Intelligence Data Platform Needs Before AI
The ingestion, modeling, output validation, and data-serving work behind useful AI-assisted campaign and creative analysis.
When Google Sheets Automation Is Enough, and When You Need a Warehouse
How I decide whether a reporting workflow should stay in Google Sheets and Apps Script or move into BigQuery, a database, or a custom dashboard.
Building a Custom Amazon Seller Analytics Dashboard: What Generic BI Tools Miss
Why Amazon vendor and seller reporting workflows often need a custom dashboard. How email-based report ingestion, MySQL storage, and a Next.js reporting app solves what Seller Central and generic BI tools cannot.
Normalizing Affiliate Network Data: Awin, Rakuten, and TradeTracker
How to normalize affiliate network transaction data from Awin, Rakuten, TradeTracker, and similar networks into one reporting-ready dataset. The schema differences, status lifecycles, and validation checks that matter.
Building a Reliable DV360 and Ads Data Hub Pipeline with BigQuery, Cloud Run, and AWS SQS
Engineering lessons from building a reliable marketing data pipeline with DV360, Ads Data Hub, BigQuery, Cloud Run, Cloud Tasks, and AWS SQS.
Data Engineering & Automation: A Complete Guide for Growth Teams
Learn how data engineering and automation can transform your business operations. Expert insights from Ahmad Humayun on building scalable data solutions.
Meta Ads to BigQuery: The Reporting System Behind the Pipeline
Meta advertising ingestion, BigQuery reporting layers, recovery, attribution-aware transformations, and campaign analysis within a production analytics product.