Common Issues with Custom Bidding Strategies in Google Ads - Diagnosis and Fixes
A technical guide for US performance teams to identify why custom bidding underperforms and practical steps to restore accurate, revenue-focused results.

Image via 123RF
In This Article
Root cause first
Use proxy conversions
Align with revenue
Why custom bidding strategies in Google Ads fail
Custom bidding strategies can be powerful for scaling paid search and performance campaigns, but they rely on accurate signals, sufficient conversion volume, and disciplined experiment design. This section explains the most common issues with custom bidding strategies in Google Ads and how to detect them in US eCommerce and B2B contexts.
1. Poor or delayed conversion data
If conversions are missing, delayed, or double-counted, the bidding algorithm optimises toward wrong targets. Common causes include client-side tracking failures, blocked cookies due to CCPA consent, and mismatched conversion actions. In US stores using Shopify + Stripe, ensure server-side or enhanced conversions are implemented to reduce attribution gaps.
2. Low conversion volume or noisy signals
Custom bidding needs a minimum amount of reliable conversion data to learn. Campaigns with fewer than a few dozen conversions per week will show unstable bids and wide ROAS variance. For low-volume funnels, consider conversion grouping, using a higher-funnel proxy conversion, or running experiments at campaign aggregation levels.
3. Misaligned attribution and business metrics
Platform-reported conversions may differ from revenue-focused business metrics (MER, LTV-adjusted ROAS). If Google Ads is optimising to a last-click conversion action while your finance team measures revenue-based LTV, bids will diverge from profitability goals. Use consistent conversion definitions and map critical revenue events into Google as server-side conversions when possible.
4. Overfitting to short-term signals
Aggressive target CPA or ROAS targets can push the bidding algorithm to favour narrow audiences or high-variance placements, reducing scale. Overfitting shows as sudden traffic drops with stable conversion rates but falling absolute conversions. Monitor impression share and volume alongside CPA/ROAS to detect this.
5. Incorrect goals or constraints
Setting unrealistic maximum CPCs, bid limits, or mixing portfolio bid strategies with incompatible campaign goals can throttle learning. Ensure constraints are aligned with forecasted CAC and lifetime value assumptions and provide the algorithm room to explore.
Conversion tracking diagram
| Event | Client-side | Server-side |
|---|---|---|
| Page view | gtag.js / gtag events | Server logs / proxy pixels |
| Purchase | Conversion tag, may be dropped by blockers | POST to GTM Server -> Google Ads conversion |
| Enhanced conversions | Email hashing client-side | Server-side hashed user data for attribution |
Funnel and testing implications (TOF → MOF → BOF)
- TOF: Broad reach and signals - low immediate conversion volume makes custom bidding unreliable here.
- MOF: Engagement metrics and micro-conversions - useful proxies when purchases are sparse.
- BOF: Purchases and revenue - primary signal for bidding when volume is sufficient.
As you diagnose issues, document your current conversion mapping and check the Prebo Digital services overview for how tracking and optimisation fit into a scalable growth system: Services overview.
For a high-level view of how bidding fits into an integrated growth stack, review Prebo Digital’s homepage explanation of the agency’s approach to measurement-first performance: Prebo Digital homepage.
Fixes and best practices for custom bidding strategies in Google Ads
Addressing common issues with custom bidding strategies in Google Ads requires both engineering and strategy: clean data, realistic targets, and an experimentation cadence. Below are practical, experience-based steps used by US-focused eCommerce and B2B teams to stabilise bidding and prioritise profitability over vanity metrics.
1. Improve data quality and attribution
Move critical purchase and revenue events to server-side tracking (GTM Server) and enable enhanced conversions where permitted. This reduces losses from ad blockers and cookie restrictions and aligns Google’s signals with backend revenue. If your store’s finance team tracks revenue differently, map the same events into Google to avoid divergent optimisation.
2. Use proxy conversions and aggregation
When purchase volume is low, use higher-volume proxy conversions (add-to-cart, initiated checkout, qualified lead) during the learning phase. Group similar conversion actions into a single bidding target to meet learning thresholds, then phase in the primary conversion once volume grows.
3. Set realistic targets and controlled constraints
Avoid overly tight ROAS or CPA caps that force the algorithm to overfit. For example, a $50 target CPA for a product with $80 gross margin and $25 average CAC may be too aggressive; instead, start with a range and tighten after stable performance. Regularly reconcile Google targets with business KPIs like MER and LTV-adjusted CAC.
4. Experiment with structure and bidding types
Run A/B style experiments where one campaign uses custom bidding and a control uses portfolio or manual bidding. Monitor learning duration (usually 7-28 days for stable signals) and compare on revenue, not just conversion rate. Document wins and failures so strategy decisions are evidence-based.
5. Monitor signals beyond conversions
Track impression share, average CPC, and auction insights alongside CPA/ROAS. A rising CPA with dropping impression share signals budget or bidding constraints. Use analytics tools and clean data pipelines to combine Google Ads data with backend orders for accurate reporting.
Consider scheduling a focused audit of tracking and goal alignment if you see persistent underperformance; fixing attribution issues is often the fastest path to improving bidding outcomes.
Practical US scenario: fixing a volatile bidding campaign
Example: A Shopify brand in the US saw CPA variance from $40 to $120 week-to-week with a $60 CPA target. Steps taken: implemented GTM Server for purchases, switched bidding to a portfolio strategy with an aggregated conversion action (initiated checkout + purchases), relaxed the CPA target to $75 during learning, and ran a 14-day experiment comparing outcomes. Results (estimates) showed a steady reduction in CPA variance and a 12-18% increase in weekly conversions after 6 weeks.
Compliance and consent considerations
In the United States, CCPA and state privacy rules affect tracking. Ensure consent flows do not block critical server-side events and document your consent model. Consult legal or privacy experts for compliance decisions and follow platform guidance for enhanced conversions.
If you want to know how an integrated measurement and optimisation program could look for your team, learn more about Prebo Digital’s approach and team experience: About Prebo Digital. For a scoped review of tracking and bidding options, request a direct discussion: Contact Prebo Digital.
Checklist: quick diagnostics
- Are purchases arriving in Google within a consistent delay window?
- Is conversion volume above the learning threshold (dozen+ conversions/week)?
- Do campaign constraints allow exploration (no overly strict caps)?
- Are business revenue metrics and Google conversion actions aligned?


