Zero conversions in three days — and the right call was to serve almost nothing
A seasonal campaign launch showed zero conversions everywhere. The same zero meant three different things — and the fix was pausing keywords, not adding them.
A consumer brand I look after built a campaign around one moment: a big summer football tournament. The product genuinely shines in a situation the tournament creates, so the whole thing was designed as an ambush — search campaigns in several markets, a shopping campaign, creatives, landing page, all pointed at a few weeks of unusual demand.
Three days after launch I pulled the numbers over the API, the same way I did for that first Meta MCP audit. Four figures spent. Zero conversions from the new campaigns.
The novice reading of that line is panic: kill it, or double down. Both are the wrong lever. What the data actually showed was three completely different situations wearing the same zero.
The leak was hiding in two small markets
The two smallest markets were burning almost all of the wasted money. The mechanism is worth understanding, because it's one of the most common ways launches bleed.
The campaigns ran on Maximize Conversions — Google's smart bidding — with no conversion history in those markets. Google's own documentation says the algorithm needs around 50 conversion events to calibrate, and that prior conversion data is what speeds the learning phase up. With nothing to learn from, the system explores. Expensively.
And it was exploring on the wrong keywords. Alongside the tournament-specific terms, the ad groups contained generic product keywords. Those generic searches were landing on the tournament promo page — a mismatch Google punishes hard. The worst keywords sat at a Quality Score of 1 out of 10, with landing page experience rated below average, and clicks costing many times what a healthy account pays. Smart bidding without history plus generic intent plus the wrong landing page is a money incinerator with a respectable-looking dashboard.
The biggest market was spending nothing — and that was fine
Meanwhile the largest market's search campaign had spent exactly zero. Every keyword flagged "rarely served." The tournament-specific searches it targets simply didn't exist yet, because the tournament hadn't properly started for that country.
This is the part that looks like failure and isn't. The campaign was in position, costing nothing, waiting for the moment it was built for. The wrong move — the one the platform's own recommendations nudge you toward — would have been to "fix" it by adding broader keywords. That's how you turn a sniper into the two small markets I just described.
The fix was to serve less, not more
So the intervention went the other way. Every generic product keyword in the leaking markets: paused. Every use-case keyword tied to the tournament moment: kept live, even the ones barely serving. One market now serves almost nothing at all, and that is the deliberate, correct state — a burn stop, not a retreat.
The shopping campaign got left alone entirely. It was mid learning phase, the account's other campaigns were still converting normally (so tracking wasn't the problem), and touching a campaign during calibration resets the clock. Restraint is also a decision.
A launch campaign's job is not to spend. Its job is to be in position when the moment arrives.
This is the same lesson as the brand search post from a different angle: the dashboard number means nothing without the structural story underneath it. Zero conversions meant "wrong intent" in two markets, "too early" in another, and "still learning" in the fourth. Same number, three diagnoses, three different actions — and one of those actions was doing nothing.
AI made the diagnosis fast. Pulling keyword-level Quality Scores, cost splits and serving statuses across four markets took minutes instead of an afternoon. But the tooling didn't know that near-silence was the goal state for a tournament campaign before kickoff. That's the judgement layer, and it's the whole argument of the death of generic AI: the levers are cheap now. Knowing which one not to pull is what you're actually paid for.
The tournament starts properly this weekend. The campaign is quiet, cheap, and exactly where it should be.
Sources & further reading
External
Google Ads Help — Learning phase and duration · Google Ads Help — About Quality Score for Search campaigns · Store Growers — How to use Maximize Conversions bidding
Related posts
Your brand search isn't the hero your dashboard thinks it is · I pointed the Meta MCP at a real ad account · The death of generic AI