In 2026, the PMax learning phase typically lasts 2 to 6 weeks. During this time, performance fluctuations are normal. However, if the campaign remains stuck in “Learning” or fails to exit it with stable results, you need to diagnose the cause.
1. Identify the “Learning” Status
To see if your campaign is officially still in this phase:
Go to the Campaigns tab.
Look at the Status column.
Hover over the “Learning” status: A tooltip will appear showing one of three reasons:
New strategy: The campaign was recently created.
Setting change: A major change (like budget or bid target) was made.
Composition change: You added/removed asset groups or listing groups.
2. The “Data Threshold” Check
The AI cannot learn without “success signals” (conversions). If you don’t have enough data, the phase will drag on indefinitely.
The 2026 Benchmark: PMax requires 30 – 50 conversions per month minimum to stabilize.
The Diagnosis: If your campaign has been live for 14 days and has fewer than 10 conversions, the learning phase is “stalled” due to low data.
The Fix: Temporarily switch from “Target ROAS” to “Maximize Conversions” (with no target) to let the AI spend freely and find those first 30 converters.
3. Identify “Learning Resets” (What restarts the clock?)
Technical writers must warn users that certain actions “reset” the AI’s memory, effectively starting the 2-6 week period over.
Major Resets: Changing the bid strategy (e.g., from Max Conversions to tROAS), changing conversion goals, or changing the campaign’s structural settings.
Minor Disruptions: Adjusting the budget by more than 20% or swapping out more than 2-3 creative assets at once.
Diagnostic Table: Is it “Normal Learning” or a “Problem”?
Scenario | Status | Action |
High Spend, 0 Conversions (Days 1-7) | Normal. The AI is testing expensive placements. | Wait. Do not touch the campaign. |
Status stays “Learning” after 14 days | Problem. Likely a “Low Volume” issue. | Broaden targeting or lower the ROAS target. |
Status is “Active” but CPA is 3x higher than goal | Normal. The “Post-Learning” calibration. | Give it another 7 days to stabilize. |
Daily Spend is highly inconsistent | Normal. AI is pacing based on daily auction volume. | Monitor the weekly average instead. |
Pro-Tip for your Guide: The “90-Day Conversion Window”
In 2026, many customer journeys will take weeks. If your PMax status is “Active” but you see no results, check your Conversion Lag report.
Go to Insights & Reports > Search Insights. Look at the “Days to Conversion” data. If your average customer takes 14 days to buy, you cannot judge a PMax campaign’s learning phase until day 28.
How to “Speed Up” Learning
To help the AI exit the learning phase faster, provide it with Audience Signals immediately upon launch:
First-Party Data: Upload your customer email list.
Search Themes: Add 10 – 15 high-intent keywords that you know work in your Search campaigns.
Merchant Center: Ensure your product feed is 100% error-free before launch.

Salesforce

