添加HostName/pagePath维度后GA4数据数值偏高，求精准获取方法

问题本质 #

GA4中activeUsers、sessions这类用户/会话级指标，和eventName、pagePath这类事件/页面级维度组合查询时，会因为一个用户/会话对应多个事件/页面记录，导致指标被重复统计。比如一个用户访问3个页面，每条页面数据都会携带该用户的标识，最终求和时同一个用户会被计算3次，造成数值虚高。

解决方法 #

方法1：给用户/会话指标指定UNIQUE聚合类型 #

直接在请求中给activeUsers、sessions这类需要去重的指标添加aggregationType: "UNIQUE"，让GA4在计算时自动去重统计：

{"dimensions":[{"name":"date"},{"name":"eventName"},{"name":"hostName"},{"name":"pagePath"}],
 "metrics":[{"name":"activeUsers", "aggregationType": "UNIQUE"},
            {"name":"eventCount"},
            {"name":"keyEvents"},
            {"name":"screenPageViews"},
            {"name":"sessions", "aggregationType": "UNIQUE"},
            {"name":"totalRevenue"}],
 "dateRanges":[{"startDate":"2024-04-02","endDate":"2024-04-03"}],
 "metricAggregations":["TOTAL"]}

注意：eventCount、screenPageViews这类事件计数指标不需要添加该聚合类型，保持默认求和逻辑即可。

方法2：拆分查询后本地合并 #

如果聚合类型调整不生效，可将查询拆分为两组，分别获取用户/会话数据和事件/页面数据，最后在本地按维度关联合并：

def run_ga4_report(client, property_id, report_request):
    """通用GA4报表查询函数"""
    dimensions = [Dimension(name=dim["name"]) for dim in report_request.get("dimensions", [])]
    metrics = []
    for metric in report_request.get("metrics", []):
        metric_obj = Metric(name=metric["name"])
        if metric.get("aggregationType") == "UNIQUE":
            metric_obj.aggregation_type = "UNIQUE"
        metrics.append(metric_obj)
    
    request = RunReportRequest(
        property=f"properties/{property_id}",
        date_ranges=[DateRange(start_date=dr["startDate"], end_date=dr["endDate"]) for dr in report_request["dateRanges"]],
        dimensions=dimensions,
        metrics=metrics,
        metric_aggregations=report_request.get("metricAggregations", [])
    )
    try:
        response = client.run_report(request)
        return response.to_dict()
    except (GoogleAPICallError, RetryError) as e:
        print(f"查询出错: {e}")
        return None

# 拆分后的用户/会话指标查询
user_session_request = {
    "dimensions": [{"name":"date"},{"name":"hostName"},{"name":"pagePath"}],
    "metrics": [{"name":"activeUsers", "aggregationType": "UNIQUE"}, {"name":"sessions", "aggregationType": "UNIQUE"}],
    "dateRanges":[{"startDate":"2024-04-02","endDate":"2024-04-03"}],
    "metricAggregations":["TOTAL"]
}

# 事件/页面指标查询
event_page_request = {
    "dimensions": [{"name":"date"},{"name":"eventName"},{"name":"hostName"},{"name":"pagePath"}],
    "metrics": [{"name":"eventCount"},{"name":"keyEvents"},{"name":"screenPageViews"},{"name":"totalRevenue"}],
    "dateRanges":[{"startDate":"2024-04-02","endDate":"2024-04-03"}],
    "metricAggregations":["TOTAL"]
}

# 执行查询后按date+hostName+pagePath关联合并数据
user_session_data = run_ga4_report(client, property_id, user_session_request)
event_page_data = run_ga4_report(client, property_id, event_page_request)
# 自行实现本地合并逻辑，比如用字典按维度键映射关联

方法3：基于会话ID二次聚合 #

如果需要单查询返回数据，可先加入sessionId维度做会话级分组，获取原始数据后在本地对activeUsers、sessions做去重统计（比如按用户ID/会话ID去重后计数），但这种方式会返回大量原始数据，需要注意性能。

内容的提问来源于stack exchange，提问作者Mrs. Silpa Das