The situation we inherited
Our client is a boutique asset manager based in London, running a wealth-management product targeted at UK and Channel Islands high-net-worth individuals. The business had been running a Google Ads programme for two years, managed by a generalist agency, with a monthly spend in the mid-five figures. Our diagnosis on takeover was that the programme had drifted in the way most multi-year programmes drift: a structure that had once made sense had been added to in small increments until it no longer cohered.
Specifically: 380 keywords spread across 23 ad groups, the majority on Broad match; conversion tracking misconfigured to the point that two of the four key actions were not informing the bidding signal; landing pages running on a third-party form provider that threw a 5xx error roughly 8% of the time; no negative-keyword discipline; brand and non-brand campaigns sharing budgets and bidding strategies. The cost per qualified lead reported by the agency was substantially below what the client was actually seeing in their CRM, because most of the conversions being credited were not high-quality enough to be classified as a lead by the sales team.
The rebuild
We rebuilt the programme in three months under a fixed-fee retainer, working alongside the client’s in-house analytics lead.
Month one: structural rebuild. A new campaign architecture — brand, generic-evergreen, intent-led, and a small Performance Max test — with tightly themed ad groups. Match types tightened: 60% Phrase, 30% Exact, 10% Broad with audience signal anchoring. Landing pages migrated to the client’s own infrastructure, with first-party form handling and server-side conversion firing. Conversion actions reconfigured: one primary-for-goal action defined as a CRM-qualified lead, with three secondary actions for upper-funnel signals. Enhanced conversions implemented via the API to recover signal from cookie-deleted journeys.
Month two: creative and copy. Ten new RSAs across the live ad groups, written by the client’s in-house copy lead with our editorial input. Sitelinks and structured snippets rebuilt. Three months of historical search-term data classified into negative-keyword lists at account, campaign and ad-group level — 1,400 negatives applied at takeover, 60 to 100 added per week thereafter as the search-term report stabilised.
Month three: bid strategy and scaling. Once the rebuilt campaigns had accumulated 30+ conversions each, we moved bid strategies from Manual CPC to Maximise Conversions (no target), and then — once the conversion data was stable — to Target CPA against a client-set target. Spend was scaled in 15% increments, with a hold step after each increment to verify CPA stability.
What we changed structurally
- Brand and non-brand are now separate campaigns with separate budgets and bidding strategies. Brand exclusion applied at PMax level.
- Conversion tracking is one primary action (CRM-qualified lead), three secondary actions (form submission, brochure download, calculator engagement). Bidding signal reflects business value, not form completions.
- Negative-keyword sweeps run weekly. The cumulative negatives list is treated as an asset and version-controlled in our tooling.
- Reporting reconciles weekly to the client’s CRM via a server-side import of CRM stage transitions. The reported CPL is now the same number whether the client looks at Google Ads or at their CRM.
The outcome
By end of month three the cost per qualified lead had declined 48% versus the trailing-twelve-month baseline at takeover. Qualified-lead volume had increased 3.1× over the same window. Return on ad spend — calculated against actual closed-won revenue from the lead cohort, six months out — was 3.9× the at-takeover figure.
The client extended the engagement for a second year, expanded scope to include Paid Social, and increased the monthly retainer in line with the new programme’s strategic importance to the business.
What we’d do differently
One thing in retrospect: we should have run a more aggressive landing-page experimentation programme from month two onwards, rather than waiting for the rebuilt campaigns to stabilise. Landing-page work is largely independent of campaign structure and we were over-cautious in sequencing it. Estimated cost of the cautious approach: 10% to 15% additional CPL improvement that we left on the table for two months.