Maddie Lightening’s recent conversation, covered by Anu Adegbola on Search Engine Land, is a timely reminder that performance data can deceive if technical details and account design are overlooked. From a currency reporting error that effectively halved reported ROAS to legacy account structures that clash with AI-driven bidding, the lessons are practical and immediate for any marketer running paid media.
How a simple currency mismatch can change the story
One of the clearest takeaways was how a currency configuration error turned a strong-performing account into a perceived underperformer. Lightening described a situation where reporting in the wrong currency meant conversion values were translated incorrectly, making performance look half as good as it actually was. That kind of mismatch isn’t dramatic to spot in platform dashboards, but it can be catastrophic when teams make budget or strategy decisions based on the wrong numbers.
Actionable audit steps
- Confirm the account currency and billing currency match the reporting currency used in dashboards and finance systems.
- Reconcile platform ROAS with CRM or revenue data on a regular cadence—weekly for active campaigns; monthly for broader reviews.
- Log and monitor any currency conversion rules or feeds that affect conversion values (e.g., cross-border billing, multi-currency stores).
Legacy account structure versus modern automation
Lightening also recounted working with a travel client that ran a highly granular, years-old account structure with thousands of campaigns—an approach that used to work but now conflicts with AI-driven bidding and aggregated data strategies. When the account needed restructuring, the team delayed the changes to avoid disrupting peak season, and when performance dropped they were forced into rushed corrective work.
How to restructure without wrecking performance
Restructuring is rarely quick, but it can be staged to preserve machine learning signals and control risk:
- Plan a phased migration: test new structures on low-risk campaigns first.
- Use experiment campaigns (where available) to measure lift before shifting full budget.
- Keep a historical link: maintain access to legacy performance data for a full learning window after migration so models can be retrained without losing context.
Regaining control when AI-driven bids feel out of hand
Despite the rise of automation, Lightening offered a tactical fix that is immediately useful: applying a max CPC cap through portfolio bidding strategies. This approach allows automated bidding to run while bounding worst-case costs—an important lever when CPCs rise unexpectedly.
She also pushed back against blanket bans on AI tools, arguing that refusing automation can put teams at a disadvantage. Instead, marketers should learn to shepherd AI: give it clear boundaries, continuous feedback, and better inputs.
Practical controls to combine with automation
- Set conservative max CPC caps via portfolio bidding or campaign bidding strategies to limit cost shocks.
- Use performance targets tied to revenue (not just clicks) when configuring automated bidding.
- Monitor spend and conversion quality closely in the first 7–14 days after any major bidding change.
Make better AI outcomes by engineering better prompts
Lightening emphasized that the output quality of AI tools depends heavily on input. As she put it, the team’s philosophy is “test and learn,”—a concise way to summarize the iterative approach good AI use demands. Vague prompts produce weak outputs; detailed prompts with clear goals, audience context, and constraints produce actionable ideas.
Complementing that perspective, a diagnostic guide from Pixis recommends that marketers “stop treating GA4 vs Google Ads discrepancies as ‘bad data,’ and start treating them as different versions of the truth.” That framing helps teams decide when platform-reported metrics are tactical signals (use Google Ads for real-time bidding) and when to rely on consolidated analytics for financial reporting (use GA4 or CRM reconciled revenue).
Prompt-engineering checklist (quick)
- Start with the goal: optimization (ROAS), traffic, conversions or awareness.
- Provide audience and product context (demographics, SKUs, price ranges).
- List constraints: tone, brand voice, character limits, regulatory restrictions.
- Ask for multiple variations and include a clear ranking criterion (e.g., expected CTR, relevance score).
Operational recommendation: a weekly ROAS audit
Inspired by the Pixis framework, a predictable weekly ROAS audit can spot discrepancies before they force knee-jerk decisions. Compare Google Ads ROAS with CRM-reconciled revenue and GA4 for identical date ranges, validate attribution windows and models, and flag any differences that exceed your established thresholds (for example, 15–25%).
Where substantial gaps appear, prioritize technical checks: tag firing, ad-blocker impact, timestamp/timing mismatches, or cross-device attribution gaps. Where technical issues are ruled out, use the discrepancy to inform whether to trust Google Ads for optimization vs. GA4/CRM for financial planning.
Final thoughts and next steps
Maddie Lightening’s examples are practical reminders: small technical oversights and old account design can undermine excellent strategic work. By auditing currency and conversion settings, planning measured restructures, enforcing bidding constraints, and improving how we prompt AI tools, organizations can turn confusing metrics into reliable inputs for smarter investment decisions.
Read the original Search Engine Land coverage by Anu Adegbola: https://searchengineland.com/maddie-lightening-speaks-on-misreported-roas-account-structure-chaos-ai-mistakes-474151
.png)
