Smart Bidding Diagnosis: When It's Not Working and Why (2026)

Why isn't my Smart Bidding working?

Smart Bidding underperforms for one of seven reasons in 2026: (1) below conversion threshold (under 30 for tCPA, 50 for tROAS, 15 for Maximise variants — the most common cause), (2) learning period turbulence (first 7 days after any change show unstable bids), (3) conversion data quality issues (duplicates, missing values, missing key event marking, wrong currency), (4) broken Consent Mode V2 signals (V1-only or stripped signals defeat the modelling that fills consent-denied conversions), (5) attribution gaps (gclid not preserved through user journey, payment gateway returns dropping click ID), (6) budget-limited campaigns (Smart Bidding can't optimise when daily budget exhausted by 2pm), (7) target value misalignment (tCPA target unrealistic for the campaign's actual market). The diagnosis order matters — checking conversion volume before consent signals before attribution gaps catches root causes faster than random investigation.

The three-step diagnostic workflow

When Smart Bidding underperforms, work through these in order. Each step takes 10–30 minutes and rules out a class of issues.

Step 1 — Volume and threshold check (10 min)

The fastest diagnostic. Most Smart Bidding underperformance is volume-driven.

1. Google Ads → Campaigns → check the Status column. "Limited" or "Eligible (Limited)" indicates data sufficiency.
2. Click into the campaign → Settings → Bid strategy. If a learning phase note is showing, the algorithm is still calibrating.
3. Check the conversion count for the past 30 days vs the threshold for your bid strategy: 30 for tCPA, 50 for tROAS, 15 for Maximise variants.
4. Check the time-of-day impressions chart. If campaigns hit budget and stop serving by midday, you're budget-limited regardless of bid strategy.

If steps 1–4 reveal volume issues, fix those first. No amount of attribution tuning fixes a campaign with 12 conversions when tCPA needs 30.

Step 2 — Conversion data quality check (15 min)

Once volume is sufficient, validate the conversions Smart Bidding is learning from are clean.

1. GA4 Admin → Events → check key event marking. Every conversion event you want feeding Smart Bidding must be marked as a key event (formerly "conversions" toggle).
2. GA4 Admin → Product links → Google Ads. Confirm conversion import is enabled.
3. Run the BigQuery duplicate transactions query (see *Duplicate Transactions in GA4*). Duplicates inflate conversion counts and break Smart Bidding's value learning.
4. GA4 Monetisation → check currency consistency. Mixed currencies break tROAS optimisation.
5. Check for missing values. Conversions with value: 0 or missing values count toward volume but not toward tROAS learning. They drag down the average value the algorithm targets.

The cleanup priority: fix duplicates first (they affect everything), then missing values (specifically affects tROAS and Maximise Conversion Value), then currency consistency.

Step 3 — Attribution and consent check (30 min)

Once conversions are volume-sufficient and clean, verify they're being correctly attributed to ad clicks.

1. Pick a recent purchase. Trace its journey. Was the user originally from a Google Ads click? Did gclid persist through to conversion? Tools: GA4 Realtime + Source/Medium reports + your CRM.
2. Check Enhanced Conversions match rate. Google Ads → Tools → Conversions → conversion action → Diagnostic. Below 70% indicates implementation issues.
3. Verify Consent Mode V2. GA4 Admin → Data Streams → Configure tag settings → Consent settings. Both indicators should be green. If V1 only, modelled conversions aren't filling the consent-denied gap.
4. Test the gclid pipeline. Click your own ad in incognito, complete a conversion, verify gclid persisted through the entire journey to the conversion event in GA4.
5. Audit payment gateway flow. Stripe, PayPal, Klarna returns can strip gclid — typical 3-5% conversion attribution loss per gateway.

If steps 1–5 reveal attribution issues, those compound conversion volume losses — fixing them unlocks both raw conversion count AND quality of Smart Bidding learning.

The seven failure modes in detail

Failure 1 — Below conversion threshold

Most common single cause. Symptoms: "Limited - data" status, conservative bidding, volume capped below budget, high learning-period CPA.

Fix paths:

• Aggregate conversion events into a single "lead" or "purchase" event
• Switch to Maximise Conversions (15 threshold) instead of tCPA (30)
• Loosen targeting to increase conversion volume
• Wait for natural growth or stay on Enhanced CPC as last resort

Time to see fix impact: 14 days (7-day learning period + 7 days of stable performance).

Failure 2 — Learning period turbulence

Symptoms: unstable CPA/ROAS, volume fluctuating, bids visibly changing day-to-day. First 7 days after any significant change.

Fix path: wait. Don't make changes during learning period. Annotate the change date in dashboards.

Time to see fix impact: 7-14 days for stabilisation.

Failure 3 — Conversion data quality issues

Symptoms: GA4 revenue doesn't match source-of-truth, items reports show duplicate transaction_ids, currency mismatches in Monetisation report.

Fix path: deploy duplicate detection (server-side dedup), enforce currency on every event, fix missing items arrays. See *Duplicate Transactions in GA4* and *Item Array Integrity*.

Time to see fix impact: 7 days for Smart Bidding to learn from cleaned data.

Failure 4 — Broken Consent Mode V2 signals

Symptoms: GA4 admin shows red consent indicators, modelled conversions absent, GDPR-region performance worse than non-EU.

Fix path: implement V2 properly with all four signals, verify with DevTools. See *TC-019: Consent Mode V2 ad_user_data Parameter Missing*.

Time to see fix impact: 24-48 hours for V2 indicators to turn green; 14 days for modelled conversions to start filling the consent-denied gap meaningfully.

Failure 5 — Attribution gaps

Symptoms: paid traffic showing as Direct, gclid missing on conversion sessions, Enhanced Conversions match rate below 60%.

Fix path: audit gclid preservation through the user journey, fix payment gateway returns, ensure cross-domain tracking is configured, verify Enhanced Conversions implementation.

Time to see fix impact: 14-30 days as Smart Bidding incorporates the recovered conversions.

Failure 6 — Budget-limited campaigns

Symptoms: campaigns hitting budget early in the day, "Limited - by budget" status, Smart Bidding can't pursue high-CPA periods.

Fix path: increase budget if business case supports it; tighten targeting if budget is fixed; switch to Maximise Conversion Value if budget concerns matter.

Time to see fix impact: 1-7 days.

Failure 7 — Target value misalignment

Symptoms: tCPA campaign consistently exceeding target, tROAS hitting only 60-70% of target, volume well below capacity.

Fix path: investigate market reality. If you've set a £20 tCPA target but the realistic CPA is £45, no algorithm can hit your target. Either accept higher CPA, narrow targeting, or accept lower volume.

Time to see fix impact: 14 days post-target adjustment.

What stakeholders should expect

A few hard truths to set with stakeholders before Smart Bidding work:

• Smart Bidding is not magic. It optimises against the data you give it. Garbage data → garbage optimisation.
• First 7-14 days after any change look bad. This is normal. Don't react to it.
• Some businesses don't have enough volume for tROAS. Maximise Conversions or manual bidding is genuinely better at low volumes.
• Compliance and Smart Bidding are linked. V2, Enhanced Conversions, offline import all matter. Treat them as a system, not separate projects.
• Smart Bidding can be wrong. The algorithm optimises for what you ask. If your conversion event is wrong (lead-fill vs closed-deal), the algorithm optimises for the wrong thing reliably.

The narrative for stakeholders: "We're going to fix the inputs. The algorithm will then do its job." Don't promise Smart Bidding will solve underlying data quality problems.