Wasted Search Spend Forensics for Restricted Accounts

In the restricted-category search accounts we have opened cold over the past two years, telehealth, digital asset exchanges and staffing groups, the recoverable share of monthly spend has landed between 19 and 41 percent. That is a range, not a forecast for you. This engagement establishes your number and proves it with a geographic holdout, because a number nobody tested is a number your CFO is right to ignore. Where the money goes: Three leaks account for most of it, and none of them look like bad keywords in a report. Geography is the first. A Singapore account set to presence or interest pays for Jakarta and Kuala Lumpur research traffic on a device the distributor is licensed to sell only inside Singapore. We read geo_performance_view against the markets you are licensed and staffed to serve, then set targeting to presence and let the impression share fall where it falls. Query breakage is the second. Broad match and Performance Max both buy intent by inference, and in medtech that inference lands on symptom queries from patients who will never be the buyer, while in staffing it lands on job seekers when you are selling to employers. The search terms report alone will not show you this, since a slice of spend always sits on terms it never names. We rebuild the picture from search_term_view, the Performance Max search category report and server-side session data together, then map each cluster to buyer or not-buyer by hand. The third is demand you already own. Brand search, remarketing lists on search, and conquest campaigns that mostly re-buy your own returning users. This is the one that needs a test rather than an argument, so we run one. How the holdout works: We split your addressable geos into matched pairs on 13 weeks of conversion history, pause the campaign under suspicion in one half for 14 days, and read total conversions rather than platform-attributed conversions across both. Two campaigns per engagement, chosen because the money at stake justifies the risk of the pause. Everything else is reasoned, and the memo labels it as reasoned. What you receive: A written decision memo naming each leak, the dollars attached to it, and the change that closes it. A ranked change list your team or ours can execute. The holdout design and its raw result. A query-to-buyer mapping delivered as one dataset. A working session with the person who did the analysis rather than an account manager reading it back. Not included: We do not rebuild landing pages, write ad copy or expand keywords here. We do not touch attribution modelling; if your conversions are mistracked, the audit says so and stops, because cleaning spend measured through a broken signal is theatre. We do not build or maintain dashboards. We do not audit Meta, LinkedIn or affiliate spend, and we will say plainly when the leak you are chasing lives there. Who this is not for: Accounts under roughly 40,000 US dollars a month in search. The work costs more than it returns at that level and we will tell you so on the first call. Teams who cannot pause a campaign for two weeks. Without a holdout you are buying an opinion, and there are cheaper opinions. Anyone who needs the audit to conclude that the incumbent agency is at fault. Sometimes the answer is that the account is fine and the site is not converting, and we have delivered that answer before. Anyone in a category where we have no standing to judge, including large ecommerce catalogues, app install and local multi-location. We work in regulated and restricted search, where the constraint is what you are permitted to say. That is a narrow shop by choice.