GA4 Reporting Architecture

The Challenge

When BankBazaar migrated from Universal Analytics to GA4 in 2023, the team inherited a completely rebuilt data model — and lost every report, every goal, every segment built over years in UA. At the same time, the SEO team was spending roughly 10 hours every week manually pulling data from Google Search Console, Ahrefs, and the old GA dashboards into a sprawling collection of six separate Excel files, then consolidating them into a Monday morning PowerPoint for leadership.

The problem wasn't just time. It was that by Monday, the data was already 3–5 days stale. When leadership asked "why did organic traffic drop last Thursday?" nobody could answer with confidence until the following week's report. SEO decisions were being made on lagging indicators.

Strategy & Design

I approached this as a data product problem, not just a reporting problem. The goal was to build a system that: (1) required zero manual intervention to stay current, (2) surfaced the metrics leadership actually cared about, and (3) could be extended by anyone on the team without needing to know SQL or GA4 report builder syntax.

I mapped out 12 stakeholder personas across SEO, product, finance, and leadership, and ran brief interviews to understand what questions each team actually needed answered. The finding was surprising: 95% of their questions could be answered with 8 core metrics, but those 8 metrics were scattered across 4 different tools in formats that required manual reconciliation.

Execution

The most technically complex piece was building the organic attribution model. GA4's default "organic search" channel was too broad — it lumped brand queries with non-brand, and mixed "discovery" traffic (people first learning about BankBazaar via SEO) with "return" traffic (existing customers coming back from branded searches). These have fundamentally different economic value.

I wrote a BigQuery SQL view that segments organic sessions by: brand vs. non-brand query, first vs. repeat visitor, product category (loans / credit cards / insurance / investments), and device type. This view powers every dashboard and ensures all teams are looking at the same version of the truth.

The weekly anomaly report was a particular point of pride. Every Monday morning at 7:00 AM, a Google Apps Script pulls the previous 7 days of data, calculates week-on-week and month-on-month deltas for 15 key metrics, and sends a formatted Slack message to the SEO and product channels. If any metric deviates by more than 15% from the 4-week rolling average, it's automatically flagged as an anomaly — prompting investigation before the weekly standup.

"For the first time, I walked into Monday's meeting already knowing the story. Lochan's dashboards meant we were proactive, not reactive." — VP Growth, BankBazaar

Results

The 10 hours per week saved in manual reporting freed up roughly 520 hours per year — equivalent to 13 full working weeks that could be redirected into actual SEO work: content optimisation, technical improvements, and link building. Beyond time savings, the dashboards changed the nature of SEO conversations at BankBazaar: from explaining what happened last week to discussing what to do next week.

Key Takeaways

The biggest lesson was that the best SEO reporting is the one that answers the question "so what do we do?" — not just "here's what happened." Every metric in the final dashboards was chosen because it directly informed a decision. Metrics that were interesting but not actionable were cut. This ruthless focus made the dashboards actually get used, rather than bookmarked and forgotten.

The second lesson: invest in the data layer first, the presentation layer second. Beautiful Looker Studio templates built on top of messy, unreliable GA4 data produce beautiful lies. The BigQuery transformation work was unglamorous and took three weeks, but it meant every number in every dashboard was trustworthy.