google-ads-learning-phase-tracker

Learning Phase Tracker

Complete guide for monitoring, understanding, and optimizing Google Ads Smart Bidding learning phases for faster and more stable performance.

See decision-trees.md for details.

Learning Phase Fundamentals

What Is Learning Phase?

LEARNING PHASE EXPLAINED
═════════════════════════

WHAT HAPPENS:
├── Google's AI collects data about your account
├── Analyzes which signals lead to conversions
├── Builds a prediction model
├── Bids are dynamically adjusted
└── Performance may fluctuate

WHY IT IS NECESSARY:
├── Every account is unique
├── AI must learn your specific patterns
├── Seasonality, audience, products vary
└── Generic models are suboptimal

LEARNING DURATION:
────────────────────────────────────────
Typical: 7-14 days (or ~50 conversions)

May take longer with:
├── Low conversion volume
├── Inconsistent tracking
├── Insufficient budget
├── Very niche targeting
└── New conversion actions

May be shorter with:
├── High volume accounts
├── Consistent historical pattern
├── Stable seasonality
└── Experience with similar campaigns

Learning Phase Signals

SIGNALS SMART BIDDING USES
═══════════════════════════

REAL-TIME AUCTION SIGNALS:
├── Device (mobile, desktop, tablet)
├── Operating system
├── Browser
├── Location (geographic, granular)
├── Time of day
├── Day of week
├── Search query (exact + intent)
└── Ad format/placement

HISTORICAL SIGNALS:
├── User's past behavior
├── Conversion patterns
├── Seasonal trends
├── Time-of-day performance
└── Device performance history

AUDIENCE SIGNALS:
├── Remarketing list membership
├── Optimized Targeting signals
├── Customer Match segments
├── Demographics (age, gender, income)
└── Affinities & in-market

CONTEXTUAL SIGNALS:
├── Ad relevance
├── Landing page quality
├── Ad position factors
├── Competitive auction dynamics
└── Query-ad match quality

IMPORTANT:
The AI learns which COMBINATIONS of signals
lead to conversions in your specific situation.

MCP Tool Integration

STEP 1: Check bid strategy status across all campaigns
google_ads_run_gaql(query="
  SELECT
    campaign.name,
    campaign.status,
    campaign.bidding_strategy_type,
    bidding_strategy.status,
    bidding_strategy.name,
    bidding_strategy.type,
    metrics.conversions,
    metrics.cost_micros,
    metrics.cost_per_conversion
  FROM campaign
  WHERE campaign.status = 'ENABLED'
    AND segments.date DURING LAST_30_DAYS
  ORDER BY metrics.cost_micros DESC
")

STEP 2: Conversion volume per campaign (estimate learning progress)
google_ads_run_gaql(query="
  SELECT
    campaign.name,
    campaign.bidding_strategy_type,
    segments.date,
    metrics.conversions,
    metrics.cost_micros
  FROM campaign
  WHERE campaign.status = 'ENABLED'
    AND segments.date DURING LAST_30_DAYS
  ORDER BY segments.date DESC
")

STEP 3: Check campaigns potentially stuck in learning
google_ads_run_gaql(query="
  SELECT
    campaign.name,
    campaign.bidding_strategy_type,
    campaign.start_date_time,
    metrics.conversions,
    metrics.cost_micros,
    metrics.search_impression_share
  FROM campaign
  WHERE campaign.status = 'ENABLED'
    AND campaign.bidding_strategy_type IN (
      'TARGET_CPA',
      'TARGET_ROAS',
      'MAXIMIZE_CONVERSIONS',
      'MAXIMIZE_CONVERSION_VALUE'
    )
    AND segments.date DURING LAST_30_DAYS
  ORDER BY campaign.start_date_time ASC
")

Learning Phase Monitoring

Checking Status

WHERE TO FIND LEARNING STATUS
══════════════════════════════

LOCATION 1: Campaign Level
─────────────────────────
Campaigns → Columns → Modify → Bid strategy
→ Add: "Bid strategy type" and "Bid strategy status"

LOCATION 2: Bid Strategy Report
──────────────────────────────
Tools → Shared Library → Bid strategies
→ See "Status" column per strategy

LOCATION 3: Recommendations Tab
──────────────────────────────
Recommendations → Scroll to "Bidding"
→ Alerts about learning issues

STATUS INTERPRETATION:
─────────────────────
┌─────────────────────┬─────────────────────────────────────────┐
│ Status              │ Meaning                                 │
├─────────────────────┼─────────────────────────────────────────┤
│ Learning            │ Actively learning (7-14 days)           │
├─────────────────────┼─────────────────────────────────────────┤
│ Learning (limited)  │ Insufficient data to learn              │
├─────────────────────┼─────────────────────────────────────────┤
│ Eligible            │ Learning complete, full                 │
│                     │ optimization active                     │
├─────────────────────┼─────────────────────────────────────────┤
│ Misconfigured       │ Setup issue (check conversion           │
│                     │ tracking)                               │
├─────────────────────┼─────────────────────────────────────────┤
│ Limited             │ Not learning-related (budget/bid)       │
└─────────────────────┴─────────────────────────────────────────┘

Learning Duration Tracker

HOW LONG DOES LEARNING TAKE?
════════════════════════════

BASELINE EXPECTATIONS:
──────────────────────
New account/campaign:
├── Minimum: 7 days
├── Typical: 10-14 days
├── Max (normal): 21 days
└── Guideline: ~50 conversions needed

Existing account (change):
├── Bid strategy switch: 7-14 days
├── Target adjustment >20%: 7 days
├── Small target change <10%: 3-5 days
└── New conversion action: 14+ days

LEARNING DURATION CALCULATOR:
─────────────────────────────
Estimated duration = 50 / (conversions per day)

Example:
├── 10 conv/day → ~5 days (fast!)
├── 3 conv/day → ~17 days (average)
├── 1 conv/day → ~50 days (long)
└── 0.5 conv/day → 100+ days (too long)

If >21 days in learning:
├── Evaluate conversion volume
├── Check budget allocation
├── Consider higher-funnel conversion
└── Or different bid strategy type

What Resets Learning Phase?

Major Resets (Full Restart)

ACTIONS THAT FULLY RESET LEARNING
═════════════════════════════════

AVOID THESE DURING LEARNING:
─────────────────────────────

1. CHANGING BID STRATEGY
   └── From tCPA to Max Conversions = RESET
   └── From tROAS to tCPA = RESET
   └── Portfolio to individual = RESET

2. LARGE TARGET ADJUSTMENT (>20%)
   └── tCPA from $50 to $40 = RESET
   └── tROAS from 400% to 300% = RESET
   └── Exception: Adding first target

3. CHANGING CONVERSION ACTION
   └── Different primary conversion = RESET
   └── Changing conversion value = RESET
   └── Linking new conversion action = RESET

4. PAUSING CAMPAIGN >7 DAYS
   └── Longer pause = learning data becomes stale
   └── Reopening = essentially a fresh start

5. DRASTIC BUDGET CHANGE (>50%)
   └── Halving budget = RESET
   └── Doubling budget = Partial reset

Minor Changes (No Reset)

SAFE ACTIONS DURING LEARNING
═══════════════════════════

THESE ACTIONS ARE SAFE:
───────────────────────

TARGETING:
├── Adding keywords (small batches)
├── Adding negative keywords
├── Expanding locations
├── Device bid adjustments
└── Adding audiences (observation)

ADS:
├── Adding new ads
├── Pausing ads
├── Changing RSA headlines/descriptions
├── Adding/changing extensions
└── Optimizing ad copy

BUDGET:
├── Budget increase <20%
├── Budget decrease <10%
└── Small shared budget adjustments

LANDING PAGES:
├── Changing landing page URL
├── Content updates
└── A/B test URL variants

GENERAL RULE:
Small, incremental changes = OK
Large, structural changes = RESET

Gray Zone Actions

ACTIONS WITH POTENTIAL IMPACT
═══════════════════════════

HANDLE WITH CAUTION:
──────────────────────

1. BUDGET CHANGE 20-30%
   └── May slow down learning
   └── Usually not a full reset
   └── Monitor closely

2. TARGET CHANGE 10-20%
   └── May cause short "re-calibration"
   └── Not always a full reset
   └── Wait-and-see approach

3. ADDING MANY KEYWORDS AT ONCE
   └── >50 keywords = may disrupt learning
   └── Better: batches of 10-20
   └── Monitor impression share

4. ADDING NEW AD GROUPS
   └── 1-2 ad groups = no problem
   └── Large-scale restructure = problematic
   └── Test small first

5. CHANGING AUDIENCE TARGETING
   └── Observation to targeting = impact
   └── Adding new audiences = OK
   └── Removing audiences = minor impact

See decision-trees.md for details.

Learning Phase Best Practices

Pre-Launch Checklist

BEFORE LAUNCHING SMART BIDDING
══════════════════════════════

□ CONVERSION TRACKING VERIFIED
├── Test conversions fired
├── Correct attribution window
├── Values correct (if used)
└── No duplicate conversions

□ HISTORICAL DATA PRESENT
├── Minimum 30 days of data
├── Preferably 50+ conversions in history
├── Consistent tracking throughout the period
└── No large gaps in data

□ ADEQUATE BUDGET
├── Minimum: 10x target CPA per day
├── Recommended: 20x target CPA per day
├── Buffer for learning fluctuations
└── 30-day budget commitment

□ REALISTIC TARGET SET
├── Target CPA: Historical average + 20%
├── Target ROAS: Historical average - 20%
├── DO NOT start too aggressively
└── Room to learn

□ STAKEHOLDERS INFORMED
├── Expect fluctuations in first 2 weeks
├── No panic on CPA spikes
├── Focus on trend, not daily numbers
└── Review after 4 weeks, not earlier

During Learning Protocol

DURING LEARNING PHASE
══════════════════════

WEEK 1: HANDS-OFF
─────────────────
□ Daily monitoring (no action)
□ Log performance in spreadsheet
□ Expect: Higher CPA, volatility
□ DO NOT: Change budget, adjust targets

WEEK 2: OBSERVE PATTERNS
────────────────────────
□ Analyze day-of-week trends
□ Check device performance
□ Note metrics becoming stable
□ Still: No major changes

WEEK 3: FIRST EVALUATION
────────────────────────
□ Compare with pre-learning baseline
□ Check learning status in UI
□ If "Eligible": Small optimization OK
□ If "Learning": Keep waiting

WEEK 4: FORMAL REVIEW
──────────────────────
□ Full performance analysis
□ Learning complete? → Optimize
□ Still learning? → Evaluate cause
□ Document learnings

MONITORING DASHBOARD:
─────────────────────
┌───────────────────────────────────────────────────────────┐
│ Metric          │ Pre-Learning │ Week 1 │ Week 2 │ Week 3│
├─────────────────┼──────────────┼────────┼────────┼───────┤
│ Conversions     │ baseline     │ -20%?  │ stable │ +10%? │
│ CPA             │ baseline     │ +30%?  │ -10%?  │ target│
│ Conv Rate       │ baseline     │ varies │ stable │ stable│
│ Impression Share│ baseline     │ varies │ varies │ stable│
└─────────────────┴──────────────┴────────┴────────┴───────┘

Learning Acceleration Tactics

HOW TO ACCELERATE LEARNING
══════════════════════════

TACTIC 1: CONSERVATIVE TARGETS
───────────────────────────────
Start with more relaxed targets:
├── tCPA: +30% of desired target
├── tROAS: -30% of desired target
├── More auctions = more data = faster learning
└── Tighten after learning is complete

TACTIC 2: ADEQUATE BUDGET
──────────────────────────
Budget formula:
├── Daily budget >= 15x target CPA
├── Or: Enough for 15+ conversions/day
├── More spend = faster learnings
└── Investment in the learning period

TACTIC 3: PORTFOLIO STRATEGIES
───────────────────────────────
Bundle campaigns:
├── Same conversion goal
├── Shared learnings
├── Aggregated data
└── Faster optimization

TACTIC 4: BROAD MATCH + SMART BIDDING
──────────────────────────────────────
During learning only:
├── Broad match = more impressions
├── Smart bidding selects the best
├── More data for AI
└── Later: Add negatives

TACTIC 5: SEASONALITY ADJUSTMENTS
──────────────────────────────────
For known seasonal events:
├── Tools → Bid strategies → Advanced
├── Add seasonality adjustment
├── Informs AI about expected change
└── Prevents unnecessary "re-learning"

See detailed-reference.md for details.

Output: Learning Phase Status Template

# Learning Phase Status Report

## Campaign Overview
- **Campaign name:** [Name]
- **Bid strategy:** [tCPA/tROAS/Max Conv/etc.]
- **Learning status:** [Learning/Limited/Eligible]
- **Days in learning:** [X]

## Current Performance
| Metric | Baseline | Week 1 | Week 2 | Current |
|--------|----------|--------|--------|---------|
| Conversions | [X] | [X] | [X] | [X] |
| CPA | $[X] | $[X] | $[X] | $[X] |
| Conv Rate | [X]% | [X]% | [X]% | [X]% |
| Spend | $[X] | $[X] | $[X] | $[X] |

## Learning Analysis

### Data Volume Check
- Conversions past 30 days: [X] (needed: 50+)
- Average per day: [X]
- Estimated time to eligible: [X] days

### Blocking Factors
- [ ] Budget too low
- [ ] Target too restrictive
- [ ] Conversion tracking issues
- [ ] Insufficient impressions

## Recommendations

### If Status = "Learning"
- [ ] Wait at least [X] days
- [ ] No major changes
- [ ] Daily monitoring

### If Status = "Learning (Limited)"
**Priority action:** [Describe solution]

Options:
1. [Option 1 - e.g., increase budget]
2. [Option 2 - e.g., relax target]
3. [Option 3 - e.g., consolidate]

### If Status = "Eligible"
- [ ] Evaluate performance vs targets
- [ ] Consider target tightening (max 10%)
- [ ] Begin optimization cycle

## Next Steps
1. [Action 1 + deadline]
2. [Action 2 + deadline]
3. [Action 3 + deadline]

## Review Schedule
- Daily check: [Time]
- Weekly deep-dive: [Day]
- Formal evaluation: [Date]

</output>