Below is a practical troubleshooting flow you can use to diagnose what went wrong, fix the root cause, and get performance back without forcing weekly “strategy resets.”
1) First, Confirm You Are Seeing a Real Smart Bidding Problem
Smart Bidding issues usually show up as one of these patterns:
- Spend drops even though demand is there, impressions fall, and CPCs look “too cautious.”
- Spend spikes but CPA or ROAS tanks, as if the system suddenly got overconfident.
- A campaign gets stuck in Learning or keeps relearning after small changes.
- Performance becomes inconsistent day to day, even though your site and offer are stable.
Before you touch targets, identify what changed in the last 7 to 21 days: tracking, site, feed, pricing, lead quality, conversion actions, attribution model, budget caps, geo targeting, or creative rotations. If you skip this step, you will treat symptoms and keep reintroducing the same root issue.
2) Check Measurement Integrity (This Is the #1 Cause)
Smart Bidding optimizes to what you tell it is a conversion, and it is only as good as the data you feed it. When conversion data is wrong, delayed, incomplete, or inconsistent, bids will swing because the model is learning from a distorted reality.
What to verify quickly
A. Your primary conversion actions are correct.
Make sure you are not accidentally optimizing to micro conversions (page views, button clicks) when you really need qualified leads or purchases.
B. Conversion delay did not change.
If sales cycles got longer (or offline conversion uploads slowed down), Smart Bidding can look “worse” simply because feedback arrives later. Google notes that learning duration depends heavily on conversion cycles and how long clicks take to turn into conversions.
C. Improve measurement quality with enhanced conversions.
Enhanced conversions can improve conversion measurement accuracy by sending hashed first-party data (privacy-safe SHA256 hashing) that helps with stronger measurement and bidding signals.
D. If you use value-based bidding, check your values.
If values are missing, inconsistent, or inflated, Target ROAS and Maximize Conversion Value will make “bad” decisions very efficiently. Conversion value rules can also alter values at auction time and are used by Smart Bidding in real time, so review them if performance shifted after rule changes.
3) Look for “Learning” Triggers You Accidentally Created
If your bid strategy shows Learning, do not assume it is broken. Learning often appears after a new strategy, a settings change, or a composition change such as adding or removing campaigns, ad groups, keywords, or products.
Google also explains that learning duration is primarily influenced by:
- How many conversions you get,
- The length of your conversion cycle,
- And the bid strategy type.
A practical takeaway: if your campaign gets fewer conversions or has a long lag from click to conversion, it will take longer to stabilize, and frequent changes will keep it in permanent recalibration.
Book a free consultation with us to learn more about it.
4) Diagnose the Most Common Scaling Trap: Targets That Are Too Aggressive
In many accounts, Smart Bidding “stops working” right after someone tightens targets because they want efficiency back. The problem is that tCPA and tROAS are not just reporting goals. They directly constrain bids.
What this looks like:
- You lower tCPA and spend collapses.
- You raise tROAS, and volume disappears.
- You keep adjusting weekly, and the system never settles.
Fix approach:
- Move targets in smaller steps and hold long enough for conversion cycles to play out.
- If you recently changed attribution models or conversion definitions, recalibrate targets based on the new reality instead of forcing old targets onto new data.
Google explicitly warns that attribution model changes can shift conversion credit and reporting time lag, and recommends monitoring and updating bids and targets after switching models to avoid over- or under-bidding.
5) Check Attribution Changes Before You Blame Bidding
Attribution changes can make Smart Bidding look worse even if real performance is fine, because credit gets redistributed across campaigns and days.
Google notes three important effects when you change attribution models:
- Conversion credit can become fractional (decimals),
- Reporting time lag can increase,
- And credit can shift across campaigns and keywords.
If you switched to data-driven attribution or changed models, wait long enough to cover your average days to conversion before judging results, and then align targets to the new model.
6) Use the Two “Emergency Tools” Correctly (Most People Misuse Them)
A) Data exclusions for tracking outages
If tracking broke, tags were removed, imports failed, or the site went down, data exclusions can reduce the performance damage by telling Smart Bidding to ignore that bad period for optimization. Google recommends applying exclusions quickly, avoiding frequent or extended use, and considering conversion delay because exclusions apply to clicks.
B) Seasonality adjustments for short, extreme events
Seasonality adjustments are meant for major, short conversion rate changes, typically 1 to 7 days, and are not ideal for long periods. Google also notes Smart Bidding already handles typical seasonality, so use this only when you expect a significant temporary shift such as a big promo.
7) If You Need to “Reboot” Smart Bidding, Do It Without Nuking the Account
When performance is unstable, the goal is to stabilize inputs and reduce change frequency.
A clean recovery sequence often looks like this:
1. Fix tracking and conversion definitions first, then let data normalize.
2. Remove obvious constraints (like too tight targets or extreme budget caps) so the algorithm can find conversions again.
3. Ensure you are optimizing to a single meaningful stage with reasonable conversion volume, since Google recommends choosing a goal with a relatively short delay and at least 15 monthly conversions for value based bidding setups.
4. Only after volume returns, tighten efficiency targets gradually.
Book a Call With a Y77.ai Expert
If Smart Bidding has gone unstable and you want a clear diagnosis, not guesswork, book a call with a Y77.ai performance expert. We will audit your tracking, targets, attribution, and learning signals, then map a practical recovery plan you can implement without sacrificing scale. Book a free consultation with us.