We publish the numbers, not the names

That did not come from clever copy. It came from the work before the copy: dedicated sending domains, warmed mailboxes, and a verified list built to a tight ICP rather than bought in bulk.

Three hundred hand-picked contacts on a four-step sequence is not a volume play, it is phase one, and it is deliberately expensive. What it buys is a readable result: which ICP replied, which angle landed, which offer produced meetings rather than polite answers.

The winners then get templatised, the research and drafting get automated, and volume goes up on the version that earned it.

That is unusual. Efficiency normally decays as budget rises and most accounts accept it as physics.

The real problem sat underneath: conversion tracking was counting actions that carried no value, and the bidding was trained on them. The algorithm was finding more of the wrong customer, efficiently.

Three things had to be true before any budget moved. The measurement was rebuilt first, the audience narrowed to travellers arriving in the UAE, and the landing page was fixed, which returned about 20% on its own.

The kind of place where walking in is most of the pitch. Except almost every customer met a phone screen first, on a street, deciding where to go.

On a mobile connection a score of 57 means people left before the page finished deciding what to show them. No bounce report tells you those visitors existed. They were gone before the analytics had anything to record.

Within two weeks of launch the owner reported that new customers were arriving with a different expectation of the business. Same room, same team, same prices. The site had been arguing against the business, and losing.

They were sourcing through the same job boards as every competitor, which creates a problem the boards will never fix: the same candidates apply through several agencies at once.

By the time a CV reached a client it had often already been sent that morning by someone else. In a placement-fee business, ownership of the candidate is the entire commercial basis of the thing.

So we changed which auction they were in. Job boards were excluded at keyword level, campaigns were rebuilt around what candidates actually search for, and applications landed on funnels the agency owned.

A conversion improvement applies to every visitor from every channel, forever, including the ones you have not bought yet. A budget increase applies only to the visitors it buys, and it stops the day you stop paying.

Which is why order matters. If your page converts at one percent and you double the budget, you have bought twice as many people who leave. The only number that moved was the cost.