Your Google Ads ROAS is usually undercounting because the conversion signal is incomplete, not because performance collapsed overnight. If ROAS dropped right after algorithm changes or a switch to pLTV bidding, the system is bidding on different math with reset history.
Check tracking first, then judge bidding. A clean signal tells you what actually broke and what just needs time.
Why is my Google Ads ROAS undercounting?
Undercounting means Google Ads sees the spend but misses part of the revenue. Consent mode gaps, missing enhanced conversions, broken tags after a site change, and mismatched analytics and ads conversion actions all shrink reported revenue while real sales stay flat. The ratio looks worse, Smart Bidding then bids more cautiously, and volume drops next.
Run the basics in Free Tools before you touch bids or budgets. Check what is actually firing on your key pages and compare against profit, not just the ratio in the interface:
- Conversion actions — Confirm primary purchase actions fire on every checkout and lead thank-you page.
- Enhanced conversions — Confirm hashed first-party data is sent to recover conversions that cookies miss.
- Consent handling — Confirm consent mode preserves measurement when visitors reject cookies.
- Analytics match — Confirm ads and analytics agree on direction before you trust the ROAS number.
If you judge by profit-first metrics to judge results by, undercounting becomes obvious fast: spend steady, banked revenue steady, reported ROAS down.
Our Google Ads ROAS dropped after algorithm changes, what happened?
When Google Ads changes how it models conversions, values bids, or attributes credit, your history stops predicting the next auction. Campaigns re-enter learning, exploration spikes, CPCs and impression mix shift, and reported ROAS swings even if your site and offers did not change.
Treat it as a signal reset. Do not stack more changes on top. Freeze structure, budget, creative, and conversion actions for now. Segment the drop by date, campaign, match type, and device to see if it is account-wide or isolated. Account-wide and timed to the change points to modeling and attribution. Isolated points to budget limits, lost impression share, or thin conversion volume in that campaign. Fix measurement in that window first, because Smart Bidding can only relearn on what it can now see.
Our ROAS dropped after switching to pLTV bidding. How long should we wait before judging results?
pLTV bidding trades immediate ROAS for predicted lifetime value. It will bid up for new customers that look valuable long-term and bid down for one-time buyers, so early ROAS almost always looks worse. That is expected, not proof it failed.
Wait until you have meaningful volume on the new objective before you judge results. Learn how long Smart Bidding’s learning phase lasts and do not grade the test during that window. For most accounts that means at least two to three conversion cycles and stable volume, often 2-4 weeks, longer if volume is thin. Judge on predicted value calibration, new-customer mix, contribution margin, and CAC payback, not day-to-day ROAS. If volume never stabilizes, the value rules or conversion volume are too thin for pLTV to learn.
Fix the drop without guessing
Work in order. Do not change bidding to fix a tracking problem.
- Verify measurement. Check tags, enhanced conversions, consent mode, and primary conversion actions.
- Lock the test. No budget, structure, or creative changes while you diagnose.
- Split signal from performance. Compare reported conversions, analytics revenue, and banked revenue.
- Reset or ride out learning. Fix the signal, then give Smart Bidding clean, stable volume before you judge.
Start free to see where you stand.
