I've kept my clients' branded campaigns running too, but I've built a monitoring layer that watches for Google's "helpful" suggestions. If you're not auditing manually, the algorithm will happily cannibalise your own brand terms into PMax - it happened to a colleague who lost sight of a £50k branded budget in under two weeks.
I don't trust Shopping AI Max or the broader PMax black box. The reporting is garbage for understanding where marginal spend actually goes. Here's the Python snippet I run weekly to flag any branded impression share drops below 90 % - triggers an alert if performance max has silently started serving on those queries:
import pandas as pd
from google.ads.googleads.client import GoogleAdsClient
# fetch branded campaign stats
def check_branded_share(client, campaign_id):
query = """
SELECT campaign.name, metrics.impression_share, metrics.search_impression_share,
segments.keyword.info.text
FROM keyword_view
WHERE campaign.id = {0}
AND segments.keyword.info.text CONTAINS 'client_brand'
""".format(campaign_id)
response = client.service.google_ads.search(query=query)
for row in response:
share = row.metrics.impression_share
if share and share < 0.9:
print(f"Warning: branded share at {share:.2%} - possible PMax overlap")
If those numbers dip, I go in and check campaign exclusions, audit the PMax asset groups, and often find Google has started bidding on brand terms despite explicit negatives. Manual management still beats the AI hype when the client's ROI depends on it.
