Google Ads · paid search, measurement & internal tooling

Michel Cunill

Google Ads at agency scale, and the tooling underneath it.

I run paid search for a Chicago agency's client book: fintech, SaaS, and events accounts across Search, Display, and Performance Max. The part I'm known for internally is measurement. When an account looks broken, I check whether the data is real before I touch a bid. I also built the reporting tools the team uses to manage all of it.

Caracas, Venezuela/ remote, US EST hours year-round/ michelcunill1@gmail.com/ +58 424 152 0466

Scale

What I manage

My book

~100client accounts

~$39Kmonthly ad spend

The full agency book I work inside

450accounts

2,261campaigns

$2.61Mspend, Apr 25 to Jul 26

205,596conversions

28.7Mimpressions

1.74Mclicks

6.1%click-through rate

19,017assets tracked

I split these deliberately. The second set is the whole agency book, not my personal results. Every number here comes out of the reporting database, and I'll walk through how any of it was pulled.

Result

Cost per conversion, same month year over year

$13.39 → $10.03 25% lower

Jun 2025 → Jun 2026 · agency book

Spend +9%Conversions +46%

This is a team result across the whole agency book, over the same window I've been there. I was part of the group that delivered it, and I'm not going to tell you I drove it alone. Cleaner conversion tracking was a real contributor, which is the part I can speak to directly.

Day to day

What the work actually is

Build and optimize Search and Display campaigns end to end: daily pacing, bidding strategy, keyword and audience refinement, budget management across the book.
Write and test ad copy. Headlines, descriptions, responsive search ad variations, iterated against CTR and conversion data rather than opinion.
Run client meetings: performance review, what changed and why, priorities for the next cycle. Fintech, SaaS, and events clients, so the buying cycles and the definition of a good conversion vary a lot between accounts.
Handle inbound tickets and ad-hoc requests async. With a book this size, most things arrive in writing and get resolved in writing.
Review landing pages and flag the conversion problems that no bid adjustment will fix.

Measurement

The part most people skip

I own conversion tracking implementation across client sites. That means configuring and QAing tags in Google Tag Manager, validating events in GA4, and fixing tracking that's quietly wrong: a tag firing on the wrong event, a duplicate conversion inflating the count, a thank-you page that stopped existing after a site redesign.

This matters more than it sounds. Smart Bidding optimizes toward whatever you tell it a conversion is. If that signal is dirty, the algorithm spends the budget confidently in the wrong direction, and no amount of bid tuning recovers it. More than once the fix for a campaign that stopped working was a tag, not a strategy.

Stack

Google Tag Manager: container setup, triggers, QA before publish
GA4: event configuration, conversion validation, debugging discrepancies against Ads
UTM structure and conversion event design across accounts
Meta Pixel and the Conversion API, including server-side, set up on my own projects
Enough HTML and JavaScript to open a page and work out why a tag isn't firing

Tooling

I build the tools I need

Two internal systems, both built with LLM assistance. They're the reason I can hold a book this size without losing track of it, and other managers on the team work out of them too. Uncommon for someone whose title says marketing.

Designed and built, 2025

Multi-account reporting platform

Syncs every client account under the agency's manager accounts every 20 minutes and keeps about 400 days of history, then serves it through a web app: cross-account comparison, spend pacing at a glance, asset search, and a map view of campaign location targeting. Read-only by design, so it physically cannot modify ad data. The team used it daily and some of their reporting workflows depended on it. It answers "which accounts are pacing badly this week" in one screen instead of 100.

TypeScriptPostgres Reactserverless 2FA + admin approval

Built 2026, still in development

Paid search automation suite

Checks live ad destinations and sitelinks automatically. A URL is only flagged after it fails twice in a row, because sites blip and an alerting system that cries wolf gets ignored within a week. It computes a proposed fix and hands it to a person instead of editing the account. It also drafts campaigns from a client's website, and finds both the searches naming towns a client can't serve and the negative keywords quietly blocking traffic they want. Nothing applies itself: every fix goes to a person for approval. Across 100 accounts, nobody has the hours to check every destination URL by hand.

PythonFastAPI Google Ads API276 offline tests approval gate

More detail on how these were built, the decisions behind them, and the stack is on the growth engineering page.

Channels

Where I've bought, and where I haven't

My buying experience is Google-side: Search, standard and dynamic, plus Display and Performance Max, across around 100 accounts and roughly $39K a month.

I have not run campaigns in Meta Ads Manager myself. I've worked closely with a Meta team and I set up the Pixel and the Conversion API, including server-side, on my own projects. That's measurement, not buying, and I'd rather say so up front than have it surface in week two.

Before this

Background

Digital Marketing Specialist · Event Hawk, Chicago (remote) Jun 2025 – Aug 2026

Everything above.

Product Marketing Specialist · Ugami, Inc., Miami (remote) Mar 2021 – Mar 2023

Ran App Store and Google Play presence for a consumer fintech app: listings, creatives, and metadata optimized for discoverability and install conversion. Ran A/B tests and fed the results into copy and UX decisions. Built go-to-market and acquisition strategy, including in-app brand and character development. This is also where I learned to brief designers and video editors and manage creative through revisions.

Medical Interpreter, EN/ES · Propio Language Services Aug 2024 – Jun 2025

Held 88%+ quality accuracy inside the first month, among the highest in my cohort, and finished with no complaints on record. A year of interpreting live under time pressure is why I don't hedge about writing and speaking in English.

Sales Executive · Jungleworks Jun 2024 – Aug 2024

Outbound lead generation for a SaaS platform: cold prospecting, demos, negotiation, and closing with international buyers.

Psychology · Universidad Central de Venezuela Oct 2016 – Jun 2024

Completed 9 of 10 semesters before going full-time into marketing. It shows up in the work more than I expected. Most of ad copy is figuring out what someone is actually worried about at the moment they search.

Read on

Two longer versions

For PPC and paid media roles Paid search →

The full account of how I run the book, the measurement work, and the results.

For growth and marketing engineering roles What I build →

The two systems in depth: architecture, the decisions behind them, and the stack.

Sendable summary One-pager →

Everything compressed to a single page that prints clean on US Letter.