YEAR:
2025
BRAND IDENTITY
CLIENT:
VANTA STUDIO, PARIS
IDENTITY · MOTION · CAMPAIGN
INTRODUCTION
The dashboards said the campaigns were working. The recruiters said they weren't. When the people closest to the outcome disagree with the reporting, the a is usually measuring the wrong thing — and in this case it had been measuring the wrong thing for years, cheaply and consistently enough that nobody had reason to look.
CHALLENGE
Recruiters noticed first. Applications were arriving in volume, and the wrong people were attached to them — candidates who selected "Other" as their occupation, candidates without a license, candidates who clicked once and left before the page finished loading. Meanwhile the reporting looked healthy. Conversions were up. Cost per conversion was low. Every dial pointed the right direction.
We traced the full path from ad click to submitted application, and the contradiction resolved itself. In the highest-spending campaigns, as much as 60% of budget was reaching search terms that couldn't be tracked at all — invisible, unattributable, and impossible to argue with. The conversion mix was dominated by a single low-intent action, a job-opportunity search, which told the platform to go find more people who browse. The landing page took twenty seconds to load. The application form asked for eleven fields and never named the role it belonged to. A platform migration midway through the year had broken tracking badly enough that the year's own baseline couldn't be trusted.
None of these was unusual on its own. Together they formed a closed loop: the algorithm optimized toward the cheapest available conversion, the cheapest conversion was the least meaningful one, and every dollar spent trained it to find more of the same. The account wasn't underperforming. It was performing precisely, in a direction nobody had chosen on purpose.
STRUCTURE
On January 31, 2025, the account stopped optimizing for volume. Bidding moved from Maximize Conversions to Maximize Conversion Value, which meant the platform could no longer treat a browse and a completed application as the same event. Low-intent actions were demoted to secondary — still measured, no longer chased. We rebuilt the campaign structure around condensed, exact-match keyword sets, so that budget flowed to terms we could see, attribute, and defend.
Switching to a value-based strategy means assigning a number to every action, and the first set of numbers we assigned was wrong. We had carried over an assumption that Apply Now form views were worth $50, close to the value of the applications themselves. In practice, form views were abundant and applications were scarce, so the algorithm did the rational thing and went hunting for views. In hindsight the values should have been set correctly in January rather than corrected in June, and this is the part of the project we'd run differently. On June 14 we rebuilt the value structure from the outcome backwards: qualified leads to $600, completed applications to $400, form views down to $30. The incentive to farm cheap intermediate actions disappeared, and the metrics that mattered started climbing within the month.
Alongside the account work, we rebuilt what candidates actually encountered. Programmatic display creative moved to a quarterly refresh cycle, ending more than a year of fatigue on the same assets. LinkedIn video ads were cut from interviews with practicing optometrists, on the reasoning that the most credible argument for a career is made by someone already inside it. Native placements extended reach beyond the search moment. And a recruiter enablement kit — templates, branded headers, platform guidance — brought the organic hiring effort into the same voice as the paid one, so a candidate met one message instead of four.
RESULTS
Click-through rate rose 73% to 7.85%. The share of budget reaching trackable search terms went from 43% to 70% — the same money, now visible and arguable. Of every hundred completed applications, 78 came from qualified candidates, and after the June correction that number rose to 82.
The headline figure needs stating plainly: primary conversions fell 11%. That was the intention. We deliberately stopped counting the actions that had been inflating the total, which makes the drop a design decision rather than a decline. Measured across the full picture including secondary search activity, conversions rose 112%, and combined search actions grew 140% on a 66% budget increase — the clearest evidence that demoting an action doesn't suppress it, it only stops it from steering.
A year on, the pattern held. Across the twelve months to March 2026, click-through rate stayed above 7% and conversions ran roughly 6,800 ahead of the preceding period. That durability matters more than any single quarter's lift, because a spike is a campaign and a plateau at a higher level is a system.
The caveats belong here too. Budget rose 66% year over year. Cost per click rose 53% as we bid into more competitive exact-match terms. The 2024 baseline was suppressed by a platform migration that broke tracking for a full quarter, which flatters every comparison drawn against it. The strategy worked, and it worked under conditions we've described rather than conditions we've arranged.
