Python script: export Google Ads performance to a warehouse

The Google Ads API has a learning curve, but daily campaign extraction is a solved pattern. Here's the shape we deploy, trimmed to its teachable core.

The query

Google Ads uses GAQL, a SQL-ish query language:

QUERY = """
    SELECT
      segments.date,
      campaign.id,
      campaign.name,
      campaign.advertising_channel_type,
      metrics.cost_micros,
      metrics.impressions,
      metrics.clicks,
      metrics.conversions,
      metrics.conversions_value
    FROM campaign
    WHERE segments.date DURING LAST_7_DAYS
      AND campaign.status != 'REMOVED'
"""

Pull seven days on every run, not one. Ads conversions are restated for days after the click (conversion lag), so re-fetching a rolling window self-heals your history.

The extraction loop

from google.ads.googleads.client import GoogleAdsClient

def fetch_campaign_performance(client: GoogleAdsClient, customer_id: str):
    service = client.get_service("GoogleAdsService")
    rows = []
    for batch in service.search_stream(customer_id=customer_id, query=QUERY):
        for row in batch.results:
            rows.append({
                "date": row.segments.date,
                "campaign_id": row.campaign.id,
                "campaign_name": row.campaign.name,
                "channel": row.campaign.advertising_channel_type.name,
                "cost": row.metrics.cost_micros / 1_000_000,  # micros → currency
                "impressions": row.metrics.impressions,
                "clicks": row.metrics.clicks,
                "conversions": row.metrics.conversions,
                "conversion_value": row.metrics.conversions_value,
            })
    return rows

Loading: idempotent or nothing

Write to the warehouse with a merge/upsert keyed on (date, campaign_id) — never a blind append. Combined with the rolling window, this makes the pipeline safely re-runnable: crash it, rerun it, run it twice — the data is identical.

MERGE INTO ads.google_campaign_daily AS t
USING staging.google_campaign_daily AS s
  ON t.date = s.date AND t.campaign_id = s.campaign_id
WHEN MATCHED THEN UPDATE SET
  cost = s.cost, clicks = s.clicks,
  conversions = s.conversions, conversion_value = s.conversion_value
WHEN NOT MATCHED THEN INSERT ROW

Production checklist

• Retries with backoff on InternalError and quota exceptions — transient failures are routine.
• Alert on zero rows. Silence is a failure mode; an account that returns nothing usually means auth expired.
• Log request counts. The API has daily operation limits; know your headroom before you add ten more accounts.
• Store cost_micros division in one place. Half the "our numbers are 1,000,000x off" tickets in history come from doing this twice.