在子查询中 s 下面计算了用户和操作系统的总计数。由于使用了分析函数，因此行数与源数据集中的行数相同。这就是为什么您需要按消费者id和操作系统进行聚合。使用 max 或者 min :

select --group by consumer_id and operating_system
           customer_id,
           operating_system,
           max(operating_system_cnt)                    operating_system_cnt,
           max(total_cnt)                               total_cnt,
           max(operating_system_cnt)*100/max(total_cnt) operating_system_percent
    from
    ( 
    select   --calculate total count and operating_system_count  
    customer_id, 
    operating_system,
    count(transaction_id) over(partition by customer_id, operating_system) operating_system_cnt,
    count(transaction_id) over(partition by customer_id) total_cnt
    from your_table
    where --your filter conditions here for mobile/tablet and last 360 days
    )s
group by
        customer_id, 
        operating_system