We Had 99 City-Level Google Ads Campaigns. We Killed Them.
At CodaPet we run in 170+ markets, and our Google Ads account once had 99 city-level campaigns to match. We killed almost all of them — and performance improved. Why granular structure is often sabotage, and the rule I use instead: segment by data sufficiency, not by map.
There's a version of account structure that looks like control and performs like sabotage.
At CodaPet we operate in 170+ markets across the country, and at one point the Google Ads account reflected that literally: city-level campaigns, each with its own budget, its own bids, its own tiny slice of data. Ninety-nine of them. On paper it was precision — every market managed on its own terms. A dashboard that made you feel like a portfolio manager.
We killed them. Almost all of them. And performance got better, not worse.
Why the "precise" structure loses
The logic of geographic segmentation sounds unassailable: markets differ, so campaigns should too. Denver isn't Dallas. Why would you let them share a budget?
Here's what that logic misses: the algorithm doesn't learn from your org chart. It learns from conversions. Every campaign you spin up divides your conversion data into a smaller pond. Smart Bidding needs volume to find patterns — split 99 ways, most campaigns never accumulate enough signal to optimize anything. They sit in perpetual learning mode, bidding on statistical noise, while you interpret the noise as "market differences" and adjust budgets to chase it.
You end up with the worst of both worlds: the feeling of granular control and the reality of an algorithm flying blind in 99 separate cockpits.
There's a compounding human cost too. Ninety-nine campaigns means ninety-nine budgets to balance, ninety-nine sets of search terms to review, ninety-nine places for a mistake to hide. Nobody actually manages that. They skim it. The structure that promised control guarantees neglect.
What we did instead
We restructured the account around performance tiers, not geography. Markets got grouped by how they actually behave — conversion volume, economics, maturity — so the campaigns aligned with how the data clusters rather than how the map looks. High-signal groupings gave the algorithm enough conversions to genuinely learn. Budget flowed to performance instead of being rationed evenly across a hundred line items.
And we let Performance Max do what it's built for — it dramatically out-converted our Search campaigns once it had consolidated data to work with. That's the part most teams miss about automation: PMax and Smart Bidding aren't good or bad in the abstract. They're exactly as good as the volume and honesty of the conversion data you feed them. Consolidation isn't surrender to the algorithm — it's the price of admission for the algorithm working at all.
The blended result of the broader restructure: our paid CPA came down by roughly half, while the paid engine kept feeding a business that drives about 80% of its demand through channels we own.
The rule I actually use
Segment by data sufficiency, not by map.
A segment deserves its own campaign when it clears two bars at once: it generates enough conversions to exit learning mode and stay out of it, and it behaves differently enough that sharing a campaign would genuinely mislead the bidding. Different geography alone clears neither bar. A city is not a strategy.
Most accounts I look inside fail this test in one direction: over-segmented by geography, product line, or someone's internal reporting needs — structures built for the humans reading the dashboard, not the system spending the money. If your campaigns exist so your reporting looks organized, you're paying an efficiency tax on every dollar to make a spreadsheet prettier.
When would I go the other way? When a market truly earns it — enough volume to stand alone and a real economic difference (radically different competition, pricing, or service constraints). Then split it, watch it, and be willing to merge it back. Structure is a hypothesis, not furniture.
The uncomfortable part
Killing the 99 campaigns felt like giving up control. That's worth sitting with, because it's the actual obstacle — not the mechanics, the psychology. Granular structures survive in accounts everywhere because they feel rigorous, and consolidation feels like abdication. But control you can't statistically exercise isn't control. It's ritual.
The teams that win with paid right now are the ones who put their control where it still works: in the offer, the creative, the conversion data's honesty, and the economics — and let consolidated volume do the bidding. That's not less discipline. It's discipline applied where it pays.