The Google Ads account structure we default to for new eCommerce clients

When we take over an underperforming Google Ads account, the problem is rarely the ad copy or even the bid strategy. It's almost always the account structure: campaigns organized around how the marketing team thinks about the business instead of how Google's algorithm actually optimizes, which starves individual campaigns of the data volume they need to perform well.

This shows up constantly in inherited accounts: twenty campaigns each spending eight dollars a day, none of them ever accumulating enough conversion data in a rolling thirty-day window for Google's automated bidding to actually learn anything useful. The fix usually isn't more budget, it's consolidating that same budget into fewer, better-structured campaigns.

The instinct is to build one campaign per product category, matching the site's navigation. That feels organized but usually spreads budget too thin across campaigns that each get too little data to optimize well. We structure campaigns around margin tiers and purchase intent instead: high-margin bestsellers in their own tightly targeted campaign, lower-margin or exploratory products pooled together where volume matters more than precision.

For a typical eCommerce client with a couple hundred SKUs, this often collapses fifteen category-based campaigns down to four or five structured around margin and funnel stage. Fewer campaigns sounds like less sophistication, but it's the opposite: each surviving campaign now gets enough conversion volume for automated bidding to actually do its job, instead of guessing on thin data the way it was before.

Performance Max campaigns, useful as they are, can quietly cannibalize a well-performing Search campaign by bidding on the same high-intent queries with less transparency into what's actually working. We run them in parallel with clear budget caps and check search term reports weekly during the first month of any new Performance Max campaign specifically to catch this early.

The cannibalization is easy to miss because Performance Max doesn't show search term data the way Search campaigns do, so a client can watch Performance Max conversions rise while a well-tuned Search campaign's volume quietly drops, and mistake it for overall growth. Using brand exclusions and campaign-level priority settings correctly prevents Performance Max from simply capturing conversions the Search campaign was already generating at a lower cost.

Before recommending any structural change, we pull ninety days of search term reports, segment conversion rate by device and by campaign, and check whether conversion tracking is actually counting the right events, since a meaningful share of accounts we inherit have tracking misconfigured in ways that made every other decision downstream unreliable. Fixing tracking alone, before touching bids or structure, has moved the needle more than any creative refresh we've run.

We've inherited accounts double-counting conversions because both a GA4 import and a native Google Ads tag were firing on the same purchase event, inflating reported ROAS by nearly double. Every optimization decision made against that inflated number for the prior six months had been chasing a metric that didn't reflect reality, which is a far more expensive mistake than any single campaign's underperformance.

A negative keyword list built once at campaign launch degrades in value every month afterward, because new irrelevant search terms keep entering the auction as Google's matching broadens over time, especially under broad match. We review search term reports weekly for the first month of any campaign and monthly after that, adding negatives as genuinely irrelevant terms show up rather than trying to predict every bad match in advance.

Accounts we inherit frequently have negative keyword lists that were built once, years earlier, and never touched again. Wasted spend on irrelevant clicks accumulates slowly enough that nobody notices the total until we show them a ninety-day search term report with the wasted spend added up in one number.

Broad match paired with Smart Bidding is the default Google pushes hard, and it can work well for accounts with a lot of conversion history feeding the algorithm. For newer accounts or ones with thin conversion volume, broad match without close monitoring tends to bleed budget into loosely related searches faster than the algorithm has enough data to correct for. We start conservative, phrase and exact match for the core campaigns, and only widen into broad match deliberately, on a portion of budget, once we can watch the results closely.

The test we run before widening match types is simple: carve out ten to fifteen percent of a campaign's budget into a broad match experiment running in parallel with the existing phrase and exact match structure, and compare cost per conversion between the two after two full weeks. If broad match holds up within a reasonable range of the tighter match types, we widen it further. If it doesn't, we've limited the damage to a small slice of spend instead of finding out the hard way across an entire campaign's budget.

We don't flip a switch on account structure overnight. The first two weeks are audit and tracking fixes, no bid or structure changes yet, because changing structure on top of broken tracking just means optimizing against bad data faster. Weeks three and four are the actual campaign consolidation, done gradually with the old campaigns paused rather than deleted, so there's a clean rollback path if something doesn't perform as expected.

Clients understandably want to see the new structure live immediately, but a rushed restructure on top of an unaudited account is how a lot of well-intentioned agencies make an underperforming account actively worse before it gets better. By the time we hand a restructured account back for the client's own team to run day to day, usually somewhere around week six, it's been through a full reporting cycle with the new structure in place, which gives everyone a real number to compare against the account's prior performance instead of a guess based on a partial month.