Turning insights into actions to improve campaign performance
Turning insights into actions to improve campaign performance
This domain tests whether you can convert measurement findings into controlled changes that improve outcomes. Expect questions on bid strategies, budget allocation, search-term and audience actions, creative testing, and experiment design.
Core rules
- Diagnose the funnel before changing a lever. Low impression share with lost IS (budget) points to budget; lost IS (rank) points to bids, quality, or targeting. High CTR with low conversion rate points to offer, landing page, or audience mismatch. Low CTR points to creative or keyword intent.
- Choose the bid strategy that matches the goal and the data. Target CPA needs reliable conversion volume; Target ROAS needs accurate values; Maximize conversions and Maximize conversion value work with limited controls. Portfolio strategies pool data across campaigns. Do not switch strategy and budget at the same time.
- Use experiments to prove causality. Google Ads campaign experiments and ad variations split traffic; hold budget and targeting stable during the test. Wait for the learning period and enough conversions before declaring a winner.
- Act on search terms, audiences, and assets. Add irrelevant search terms as negatives; separate branded and non-branded reporting; adjust bids or exclusions by audience; test RSA assets and landing pages one variable at a time.
- Allocate budget at the margin, not the average. A campaign with a higher average CPA may still deserve more budget if its next dollar is efficient. Use conversion value and tROAS when revenue matters.
Worked example
Campaign A has CPA $40 and ROAS 3.0; Campaign B has CPA $70 and ROAS 5.0. The target is profitable growth, not lowest CPA. B returns more value per dollar, so shifting budget toward B may improve total profit, but only if B can absorb spend without degrading ROAS. The team runs a budget experiment, raises B by 20%, holds A steady, and monitors marginal CPA and ROAS over two conversion cycles. Reasoning: the decision uses value, not CPA alone, and the experiment protects against assuming the average holds at higher spend.
Traps
- Applying recommendations or optimization score changes without checking whether they match the campaign goal.
- Judging a bid strategy during its learning period or after changing conversion actions mid-test.
- Optimizing to secondary or low-value conversions because they are easy to count.
- Cutting brand campaigns when brand ROAS looks high but incrementality is low, or scaling prospecting on last-click ROAS alone.
- Shifting budget weekly on noisy data without allowing for conversion lag.