Webinar Summary
The starting point
An eight-day reporting cycle. Before the merger, getting a centralized view of marketing performance took eight people, working through Google Sheets, copying and pasting numbers out of every ad account. Eight business days for a single month-level snapshot covering 22 metrics and dimensions, and sometimes longer when manual errors meant going back and forth with the agency to correct them. Anything happening inside the month was invisible. If a sale or holiday campaign ran mid-month, the team couldn't see how it performed or how long it took to ramp up. Basic questions like "how did we do over the last seven days?" couldn't be answered at the speed the business was asking them.
What changed after the merger
Supermetrics now pulls data from more than 50 advertising, social, and Google Search Console accounts on a daily schedule into BigQuery, delivered in a production-ready format that the team reports from directly rather than raw output that needs cleaning first. Because everything comes straight from the source, the copy-paste errors are gone and the numbers are consistent every time. When a connector disconnects, Jake gets an email alert with troubleshooting steps, so keeping 50+ accounts running takes about 30 seconds of his attention instead of the full-time data quality role it would otherwise need. The reporting scope went from 22 metrics and dimensions to more than 100, and the whole change added up to 500+ hours saved every month across internal teams and agencies.
Clean data is the prerequisite for AI, not a nice-to-have.
Jake's framing was direct: garbage in, garbage out. Train agents on inconsistent spreadsheet data, and you get answers you can't trust. Now that everything lands in one warehouse in a consistent state, the team has started training their own agents. Their data sits in BigQuery, Gemini sits alongside it, and team members run an AI assistant during meetings across offices in Seattle, Japan, and Honolulu. An off-the-cuff question gets an answer grounded in real analytics without an analyst in the room writing SQL or opening dashboards. They're also currently testing the Supermetrics Gemini marketing intelligence analyst agent.
How the data has to land for AI to work
Zach walked through the technical side. Data needs to arrive flat and analysis-ready, because a model reads it row by row. Supermetrics also does the math before the data reaches the model: aggregations and calculations are handled upfront, so the LLM interprets rather than computes. That's a deliberate rule, and it's what makes the same tables readable to an analyst, a dashboard, and an agent at once.
The value spread beyond the marketing analytics team
Because the data lands in the warehouse in a usable state, other teams self-serve without knowing how to move data themselves. Hawaiian's SEO lead runs a Google Search Console dashboard built on Supermetrics tables, covering every query the sites appear for: impressions and clicks by keyword and URL, click-through rate, ranking position, plus indexing and error alerts.
Getting it past InfoSec
Two points mattered for approval. Hawaiian's setup passes data through to their own warehouse rather than storing it, with anything cached on Supermetrics servers purged within 15 to 30 minutes of delivery. And Supermetrics is SOC 2 Type II compliant. Both helped move the conversation with the data governance team along. Storage is available if you want it, but it's a choice, not a requirement.
Key takeaway: the merger didn't get solved by adding people. Getting the data foundation right meant one analyst team could absorb double the footprint, answer questions in near real time, and start building AI on top of it rather than waiting until reporting caught up.
Ready to build the data foundation your AI runs on?
See how Supermetrics brings your ad, social, and search data into BigQuery in a format your team and your AI agents can both work from, on a daily schedule you don't have to manage.
