When a Google Ads search campaign tanks overnight, it’s usually not “Google being weird.” It’s almost always signal quality (conversion rate and event integrity) plus bidding feedback loops reacting to new inputs. In your case, performance was stellar—~€1.95 CPA with ~800% ROAS—until a mobile banner removal degraded conversion rate. Smart Bidding (tCPA) kept chasing the old target on a worse funnel, learned the wrong lessons, and your CPA ballooned to ~€8. The good news: this is fixable with a structured reset.
Below is the exact plan I use with clients to stabilize, relearn, and scale back profitably.
1) Triage First: Stop the Bleed and Re-Prime Learning
Switch bidding to Maximize Clicks (temporarily) with a CPC cap.
Why: Smart Bidding (tCPA) needs consistent conversion rate and accurate event signals. When CVR drops, it overpays in the wrong auctions trying to hit a target that’s no longer realistic. Max Clicks with a strict bid cap prevents overpaying while you repair the funnel.
How to set it up
- Campaign → Settings → Bidding → Maximize Clicks
- Set a Max CPC bid limit (benchmark from your historic average CPC + 10–20%).
- Maintain your best-performing keywords and pause obvious waste (poor match queries, low-intent themes).
Guardrails during triage
- Keep daily budget steady (don’t starve learning).
- Limit locations to your top-converting regions.
- Tighten ad schedule to known high-intent hours (use historic hour-of-day data).
- Prefer Exact/Phrase for core money terms; Broad only when necessary and well-guarded with negatives.
2) Fix the Landing Page & Conversion Rate (The Real Root Cause)
Your bidding is only as good as your conversion rate. If a mobile banner, hero section, or above-the-fold element change dropped CVR, fix that before you ask algorithms to “be smart.”
Checklist
- Rollback or redesign the mobile banner to a compact, value-dense hero.
- Ensure the primary CTA is visible without scrolling (button contrast, sticky bar).
- Form UX: reduce fields, enable autofill, show trust badges, and confirm error handling.
- Speed: measure mobile LCP/INP; compress images, lazy-load below-the-fold, serve WebP/AVIF, use CDN.
- Message match: headline and first paragraph should mirror the keyword/ad promise.
- Add social proof near the CTA (logos, reviews, “as seen on”).
- Implement a thanks page with a unique URL (clean conversion tracking).
Only once you see CVR improve in analytics (even 20–30% rebound), proceed.
3) Ensure Tracking Signals Are Clean and Rich
If your conversion tag fires late, fires twice, or misses events on mobile, tCPA will mislearn.
Do this
- Validate conversions in Google Ads & GA4 with tag assistant and debug mode.
- Use enhanced conversions (hashed email/phone) for better attribution and bidding.
- Consolidate duplicate events; keep your primary conversion clear (e.g., “Qualified Lead” or “Purchase”).
- Consider micro-conversions (scroll depth, click to call, “Add to Cart”) for diagnostics—but do not optimize bidding to micro signals.
4) Clean Your Query Stream: Search Terms & Negatives
When performance dips, non-converting queries creep in.
Steps
- Pull last 60–90 days Search Terms.
- Add hard negatives for low-intent modifiers (free, jobs, DIY, meaning, definition, how-to in B2B lead gen, etc.).
- Split high-intent themes into focused ad groups (3–5 tightly related keywords).
- Prefer Exact for core money terms; use Phrase where volume justifies it; Broad only with strong negatives and audience signals.
Result: higher CTR, better Quality Score, cheaper CPCs.
5) Rebuild RSAs & Assets for CTR + Quality Score
Responsive Search Ads (RSAs)
- 12–15 headlines, 4 descriptions.
- Pin only when necessary (e.g., compliance); otherwise let Google assemble.
- Craft for message match: {Keyword: Primary Value} + proof point + call-to-action.
- Add sitelinks (pricing, features, case studies, FAQs), callouts, structured snippets, call extension if applicable.
- Location extension if you have local presence; lead form extensions for certain verticals.
A 0.5–1.0pp CTR lift can lower CPCs 5–15% via Quality Score—critical during recovery.
6) The Three-Stage Bidding Plan (From Safe → Smart → Sharp)
Stage A — Maximize Clicks (7–14 days)
Goal: stabilize traffic and rebuild healthy CVR & QS.
- Hold CPC cap.
- Monitor CVR, CTR, Avg CPC, and Impression Share (Top & Abs. Top).
- Do not chase volume yet; fix quality first.
Stage B — Target CPA (wide) (next 10–14 days)
Goal: retrain Smart Bidding with realistic targets.
- Switch to tCPA with a looser target (e.g., 20–30% higher than current blended CPA).
- Keep budgets stable; avoid abrupt changes (>20% in 48h).
- Let it learn for at least 50–100 conversions if possible (or 2 weeks).
Stage C — Target CPA (tight)
Goal: re-compress CPA once CVR holds.
- Reduce target CPA in small steps (5–10% per 5–7 days).
- Watch Impression Share (lost to rank), Avg CPC, and lead quality.
- If volume collapses, bump target back up slightly and hold.
Pro Tip: If purchase values vary widely (e-commerce), consider tROAS for better profitability control once tracking is clean.
7) Budgets, Ad Schedule, Locations, Devices
- Budget: maintain enough daily budget to get at least 10–15 conversions/week; starving learning prolongs recovery.
- Schedule: keep ads live during hours your team can respond (lead decay is real).
- Locations: tighten to the top 70–80% performing regions first, then expand.
- Devices: if mobile UX was the issue, re-check mobile metrics post-fix; apply device bid adjustments cautiously.
8) Competitive & Seasonal Factors
Your drop was triggered by your own UX change, but always verify external pressure:
- Auction Insights: new entrants or aggressive bid shifts?
- SERP changes: more LSAs, PMAX spillover, or Shopping units pushing organic below the fold?
- Seasonality: set seasonality adjustments only for short, predictable events (sales weekends). Do not use them as a crutch for broken UX.
9) Use Experiments (Drafts & Tests)
Never “YOLO” major changes:
- Set up an Experiment (50/50 split) for the switch from Max Clicks → tCPA and for target adjustments.
- Test one variable at a time (bid strategy, landing page variant, or RSA set).
- Run experiments for 2 weeks minimum or until you reach directional significance.
10) When to Call It and Rebuild
If the campaign has months of corrupted learning (awful CVR, wrong conversions), consider:
- New campaign shell with clean structure and conversions.
- Import best keywords & negatives only.
- Start at Stage A (Max Clicks), then Stage B → C again.
FAQ
Should I switch to Maximize Clicks?
Yes—temporarily. Use a CPC cap. The goal is to stabilize traffic quality while you fix CVR and signals. Return to tCPA in two steps (wide → tight).
Can a campaign be “permanently broken”?
Not if the underlying conversion rate and tracking are repaired. What feels “broken” is often an algorithm trained on bad data; retraining fixes it.
How long to see recovery?
Typically 2–4 weeks for stabilization and early wins (Stage A + early Stage B), and another 2–3 weeks to tighten CPA. It depends on volume and conversion cycle length.
Copy-Paste Recovery Checklist
- Switch to Max Clicks + CPC cap (7–14 days)
- Rollback/repair mobile banner and above-the-fold CTA
- Speed/UX fixes (forms, trust, sticky CTA)
- Validate conversions & enhanced conversions
- Clean search terms; add negatives; tighten match types
- Rebuild RSAs & assets; aim for CTR lift
- Move to tCPA (wide) → then tCPA (tight)
- Guardrails: budgets, schedule, locations, devices
- Monitor competitors & seasonality; run experiments
- If history is too messy: rebuild in a new campaign
Want a Pro Eyes-On Audit?
If you’re stuck at Stage A or B and CPA is still inflated, a structured audit pays for itself. Start with my free checklist or book a consult: