The instinct when CPA is too high is to cut the campaigns that look worst and double down on the ones that look best. Sometimes that is correct. More often, there is a structural problem that campaign-level culling does not fix, and you end up cycling through the same issues at a smaller scale. Here is how I actually diagnose high CPA and what the levers usually are.
Before optimising towards a number, you need to know whether the number is real. High CPA might mean you have an acquisition problem. It might also mean your attribution is broken and you are missing conversions, or that your tracking is double-counting and the real CPA is lower than it looks.
Things to check: Are conversion events firing correctly on all browsers and devices? Is there deduplication logic in place if you are running both a pixel and CAPI? Are you including assisted conversions or only last-click? Is your conversion window appropriate for your sales cycle?
I have seen accounts where "high CPA" turned out to be a tracking gap, not a performance gap. Fix the measurement before you change the campaigns.
Once you are confident your measurement is sound, CPA is a function of three things: click efficiency (how much you pay for relevant traffic), conversion rate (how many of those visitors convert), and offer clarity (whether people understand what you are selling and why it is worth buying). Most CPA problems are heavy on one of these and light on the others.
Click efficiency problems show up as high CPCs or poor CTRs relative to benchmarks. Usually a creative or targeting issue. The fix is better ad creative that qualifies intent before the click, or tighter audience definition.
Conversion rate problems show up as decent traffic with poor on-site behaviour: high bounce rates from paid landing pages, short sessions, low form completion. This is almost never an ads problem — it is a landing page problem. The ad made a promise the page did not keep, or the page has friction the user did not expect. Paid acquisition optimisation cannot compensate for a bad landing page.
Offer clarity problems are subtler. Traffic arrives, people look around, but nothing converts. Often the product is fine but the messaging does not resonate: the benefit is unclear, the price-to-value relationship is ambiguous, the call to action is vague. This requires honest creative and messaging work, not bid adjustments.
Beyond the three levers, there are structural improvements that often yield CPA gains without touching creative or landing pages.
Keyword or audience quality: many accounts have legacy targeting that made sense at launch but now pulls in marginal traffic. A quarterly audit of search terms, placements, and audience segments often surfaces meaningful waste.
Bidding strategy alignment: if your conversion volume is low and you are on a Target CPA or Target ROAS strategy, the algorithm is working with too little data to converge reliably. Moving to Maximise Conversions or a higher-funnel conversion event can stabilise performance while volume grows.
Budget timing: in some categories, conversion rates vary significantly by day of week or time of day. If your budget is spread evenly across a week and 70% of your conversions happen on three days, you are wasting money on the other four. Ad scheduling can be a simple efficiency win.
Cutting budget rarely fixes a CPA problem — it just shrinks the volume. Pausing everything and restarting from scratch resets algorithm learning and usually makes things worse before they improve. Adding more campaigns does not address a structural issue in the existing ones.
If your CPA is not responding to the obvious fixes and you want a fresh set of eyes, that is exactly the kind of work I do.