Google Ads account audit and structural rebuild

Twelve months of change history gets read before a single performance chart. The Google Ads API exposes change events for the last thirty days only, so the twelve-month record has to be exported from the interface by hand, and the tedium pays for itself: most underperforming accounts are not badly built, they are the residue of eleven decisions taken by five people for reasons nobody wrote down. Knowing which edit preceded which decline removes weeks from the diagnosis. First week: The change export, ninety days of search terms and campaign data pulled with GAQL through the API into BigQuery, every conversion action listed with its designation and attribution setting, and the bidding strategies with their edit history. Then two conversations: one with whoever runs the channel daily, one with whoever owns the number the channel is judged against. Those two people describe different accounts often enough that reconciling the two descriptions is itself part of the finding. What we change: Campaign count, downward, nearly always. A bidding strategy starved across nine campaigns learns nothing in any of them, so consolidation is not tidiness, it is giving the model a population to work from. Single-keyword ad groups get merged, because that structure was a bid-control device and bid control at ad group level no longer exists. Brand and non-brand get separated so the account stops reporting existing demand as acquisition, and in India we will usually argue with you about whether the brand campaign should run at all, given how cheaply a competitor can sit on your name. Performance Max and Search are examined for the specific overlap where PMax absorbs branded queries and reports them as prospecting; brand exclusions go in and the effect is measured before anything else moves. How the change ships: In drafts and experiments where volume supports a read, and in staged edits with a written rollback where it does not. We do not rebuild into a fresh account. A clean start discards conversion history and bidding learning, and the two most damaged accounts we have inherited were both fresh starts sold by a previous agency. Changes go live in sequence with one variable moving at a time, so at the end of it you can say which change did what, rather than pointing at a chart. Deliverable: Not a deck. A change list ordered by expected effect, each line carrying an owner, a date and a stated risk, alongside a do-not-touch list explaining which apparent problems are deliberate and why. Where the data did not support a defensible conclusion, the line says so instead of filling the space with a recommendation. Not part of this: Creative production, feed management and Merchant Center remediation, though we will name any of them that turn out to be the actual constraint. Media buying outside search and shopping. And any finding that amounts to increasing the budget, because if more spend were the answer you would not have commissioned an audit. We turn down: Accounts whose conversion tracking is unresolved. Restructuring on top of conversion data nobody trusts produces confident nonsense, and we will send you to fix measurement first even when that means losing this engagement. Accounts below roughly twenty-five thousand dollars a month, where structure is rarely the binding constraint and the honest answer costs less than our fee. And audits commissioned to settle an argument with an incumbent agency; we have been used as ammunition once and will not be again. If the relationship is the real question, ask that question directly and stop paying for analysis.