Yasmina Rodríguez Pérez

The problem

A search-arbitrage (RSoC) programme: we bought traffic on Meta and native platforms and earned revenue from search ads on partner pages. It started at about €90 a day and was losing money: in April, ROI was −62% and only 7% of days were profitable. The goal was to grow spend without giving the margin away.

What I did

• Structured the account by vertical and country, with a naming standard (country, keyword, platform, device, campaign type, channel) so anyone on the team could read performance quickly and act on it.
• Tested at volume: 4,046 campaigns, about 3,700 of them tests. Around 13% of the campaigns that spent became profitable, so budget moved to those and the rest were cut early.
• Scaled winners by adding new creatives to winning keywords, extending to other devices and countries, and testing related angles with broad targeting. I stopped scaling when new angles came back at breakeven.
• Tested Max Value against Target ROAS bidding in Meta after noticing a partner using it. On the right niche it improved profit by about 25%.
• Convinced a partner to give us a test domain. The same campaigns went from barely breakeven to 120% ROAS.
• Pushed back on a request for 100 new campaigns a day. With most new campaigns failing at that stage, that volume would have burned budget, so I proposed building a profitable base first and scaling as the success rate improved.

Results

Daily spend: €90 to €2,300+ The Red Club, average per day by month, 2026Hover or tap for details

Monthly figures are average spend per day. "Late Aug" is the average of the last week of August.

What I learned

Protecting spend when tracking breaks

Found a partner tracking bug that affected all of their clients, before the partner did.

In arbitrage, revenue is reported by the monetisation partner. If their tracking breaks, you are spending blind.

How I spotted it

I compared Meta conversions against the partner's reported conversions and noticed a small drop in revenue that did not match what the ad platform showed. I escalated it twice. The partner investigated and found a bug that was affecting all of their clients.

Handling a reporting outage

• When the partner's revenue reporting stopped while campaigns kept spending, I cut budgets on the weakest campaigns first.
• Paused campaigns close to breakeven after two days without reliable data.
• Prepared campaigns in advance so we could relaunch quickly, then recovered spend in phases once the data was trustworthy again.
• Used a minimum conversion threshold per channel and country before trusting a signal, so decisions were not made on noise.

What I learned

Let's connect (opens my booking page in a new tab)

The problem

New paid acquisition accounts starting from €0. The goal was to find profitable keywords and angles quickly, then scale them.

What I did

• Started with market research: checked what competitors were running in the Google Ads Transparency Center, looking for patterns in their ads and angles.
• When a pattern appeared, searched Google and analysed competitors' landing pages to find the keywords they were targeting.
• Expanded keyword ideas with Google search suggestions and Keyword Planner.
• Used partner data from Visymo on which keywords were getting the most traffic, with revenue per click (RPC), revenue and reported conversions, to decide what to launch first.
• Launched paid acquisition accounts from €0, tested them to profitability, and handed stable channels over to new specialists.
• Ran performance reviews, training and issue resolution for the team, and set KPIs and optimisation processes.
• Created an issue-report template for the IT team (error, description, how to reproduce, suspected cause, priority) so problems were fixed faster.

Results

12-month figures cover November 2024 to November 2025, excluding one month with a reporting adjustment.

What I learned

Let's connect (opens my booking page in a new tab)

The problem

Facebook spend was not reaching internal reporting, and Google pay-per-conversion tracking was lost on two accounts. With no manager in place, the issue had been stuck for about two weeks. Soon after, the internal platform stopped matching Facebook, Outbrain, Taboola and the revenue dashboard.

What I did

• Audited every account and found €3,478.51 of Facebook spend missing over 7 days.
• Escalated directly to the head of IT. After two weeks stuck, it was resolved in under 15 minutes.
• Found the root cause of a location bug: the buying platform read "MV" in campaign names as the Maldives, because it takes the location from the name. I renamed the templates to "MaxVal" and asked buyers to rename live campaigns.
• Flagged that Advantage+ shopping campaigns were not supported by the platform.
• Reconciled spend and revenue account by account: gaps of 73%, about 35% and 25% on three accounts, Outbrain and Taboola not syncing (€379 in the platform against €668 in the dashboard), removed channels still earning revenue, and revenue duplicated between campaigns sharing a channel.
• Found a buyer duplicating campaigns from Facebook instead of the platform, and re-attributed a departed buyer's revenue for January.

Results

After the fixes, the Facebook spend gap was about 2% (revenue 0.02%) and the Google Display spend gap 0.64% (revenue 0.11%).

What I learned

Let's connect (opens my booking page in a new tab)

What I did

• Built a campaign-creation workflow with ChatGPT and Claude on my own, without the tech team.
• Built a Chrome extension for the campaign platform. I pasted a list of country and keyword lines, and it detected each keyword's language (15 languages, including mixed-language keywords), then set the location, locale and matching ad template for every campaign.
• The extension reported which lines were updated, unchanged, failed or not found, so each batch could be checked before launch. I kept improving it over several versions.
• Built a Chrome extension for creative uploads. It read each asset's file name, found the exact matching keyword in the platform and filled in the keyword field for up to 30 files at a time. Files without an exact match were listed for me to check, so no creative went live under the wrong keyword.
• Built a larger extension to run the full campaign creation flow. It uploaded the creatives, matched each one to its country and keyword, applied ad categories and created the campaigns in batches of up to 60. Before submitting, it checked the budget, Facebook page capacity and matching ad template on every campaign, and stopped before creating anything if a template was missing or a keyword did not match. I used it for real launches, refined it over 36 versions and tested each release on full 60-campaign test runs before using it live.
• Built a fourth Chrome extension to audit launches. It counted the campaigns completed in the last 24 hours, listed the keywords and countries launched with a count for each, and exported the list to CSV. When the first version undercounted because the page only showed part of the queue, I rebuilt it to scroll through the whole list and keep every row it had already seen. All the analysis ran in the browser, so no campaign data left the platform.
• Kept the naming standard and pre-launch checks in place, so speed did not cost accuracy.
• Managed creative output of around 150 to 300 ad variations a day, using AI tools and briefs for designers.

Let's connect (opens my booking page in a new tab)

Problem

Campaign creation took about 70% of my working day, which left little time for optimisation and scaling.

What I did

I built a faster campaign creation process using ChatGPT and Claude.

I also set up Jira automation rules to track each campaign's preparation steps (keyword research, content, tracking links) and close tasks automatically when every step was done.

Result

Campaign creation dropped from about 70% to about 20% of my day, so I could spend more time on optimisation, analysis and scaling.

Let's connect (opens my booking page in a new tab)

Building a native team from scratch

As Team Lead at Pearch Digital (2022) I set up the native operation on Taboola and Outbrain with 3 direct reports who were new to native. I taught the basics (allowed images, verticals, how to launch) and left the creative style to them, then ran inspiration sessions. They were performing within a short time.

Coaching a loss-making buyer to profit

When I became team lead at Alegro AI in August 2024, my first goal was to make a Google Display buyer profitable. He had been losing money for a year. I set daily 9:30 check-ins with a written recap, clear targets (3 profitable campaigns from at least 10 clones), shared-screen optimisation and simple rules: optimise by 10am and no launches after 1pm. On day two, 12 of his 15 new campaigns were profitable. He was profitable within his first month with me and kept performing well while we worked closely together.

Keeping the team running without a manager

From December 2024 to March 2025 the team had no manager, or only limited management. I kept 4 buyers across Facebook, native and Google Display running: status reports to leadership, partner funding (a €20K Taboola top-up agreed with finance and the account manager), access for new people, offboarding, weekly plans, tech escalations and one-to-one support.

Weekly plans and delivery checks for native buyers

I built weekly creation plans by bid strategy, country, language and vertical for Taboola and Outbrain, then audited requested against delivered campaigns (about 80 were missing). I set a kill rule: any campaign with next-day ROI below −50% is removed.

Structure that helps a team move faster

Account structure by vertical and country, a shared naming standard, and a task-assignment structure for the IT team so requests were clear and prioritised.

Let's connect (opens my booking page in a new tab)