Question = Design Twitter trending service to get top K trending hash-tags

My solution = I used count of each tags being used in tweets to get top trending tags. I used distributed redis cluster to store & increment the count. But at this the interviewer pointed out that its not always the hash-tag with max count that is trending.

For eg there's an tournament going on for 10 days which has 1M tweets, but then there occurs an earthquake & lot of people starts tweeting about it, so even theough the count(tournament) > count(earthquake) but since the rate of tweet for earthquake is more it will be at position-1.

On this I had to change my entire approach, Instead of count it should be the write/throughput. I was unable to come with a scalable approach which costed me my chance to get into adobe.

Can someone help me with an solution