The situation
A UK telehealth business operating in one of the most tightly regulated corners of digital marketing, where the ASA, MHRA and GPhC all have a say in what an ad or a landing page is allowed to claim. New customer volume had fallen by around half, and it had fallen during a period when several things changed at once: an agency transition, compliance-driven changes to the website, and a series of edits to tracking. Everyone had a theory. Nobody had evidence.
I was brought in on contract to own paid acquisition end to end, reporting directly to the founder and CEO. The brief was simple to say and hard to do: find out what actually broke, fix it, and get growth moving again without stepping outside the rules.
The diagnosis
I started with measurement, not media. When performance falls and the causes are tangled, the platforms' own numbers are the first thing to distrust, because bidding algorithms learn from whatever the account feeds them, right or wrong.
The audit found three compounding faults. Soft page events were sitting in the primary conversion column, so the headline conversion figures looked healthy while real sales fell. Multiple overlapping purchase tags were counting the same sale more than once, inflating reported revenue and training bidding on fiction. And a cross-domain tracking break between the marketing site and the checkout subdomain meant the majority of paid revenue, 58% of it, was landing in analytics as direct traffic with no campaign attribution at all. From the platform's point of view, paid media looked like it had stopped working. In reality, the account could no longer see what paid media was doing.
The work
Rebuilding the measurement layer came first. I fixed the cross-domain configuration so sessions survived the move from marketing site to checkout, restructured conversion actions so bidding optimised to deduplicated real purchases rather than page events, and put consent handling in place properly, with a clear-eyed view of what compliant consent would cost in reported volume so nobody mistook a reporting dip for a trading collapse.
With the signals trustworthy, I transitioned the account to value-based bidding with guardrails agreed in advance: floor targets derived from cohort revenue and confirmed product margins, and a standing rule that nothing would be changed that put the account back into a learning phase without advance notice to leadership. Bidding changes stopped being surprises.
Alongside that I built the monitoring to keep everyone honest. Custom Google Ads scripts audited the account continuously: URL health across every ad and keyword, disapprovals, invalid click rates, and spend on keywords that had never produced a sale. One 30-day pass found over £35,000, around 16% of spend, going to zero-sale keywords, which turned a vague sense of inefficiency into a specific, cuttable list. A daily purchase heartbeat flagged any day of spend without sales. Weekly trading reports went to the CEO, written to state the action being taken on each number, not just describe the movement.
The agency relationship moved onto the same footing: weekly sessions run on evidence, written actions after every call, and change history reviewed so that account changes were made deliberately and on the record. And because the business kept moving, I also led the compliant paid launch of a newly MHRA-approved treatment across Google and Meta while the recovery work ran.
The results
The account went from optimising to inflated signals it could not trust, to bidding on deduplicated sales with attribution intact and agreed guardrails. Tens of thousands of pounds of monthly waste was identified and removed. Leadership went from competing theories to a single evidence-led view of the funnel, and the recovery plan was presented at board level with the numbers to support every line.