The Herd Locker sends out thousands of reports to subscribers each week, but the good folks over at Mailchimp limit me to just 60 topics. So every now and again I have to evaluate the hashtags in play and sometimes replace them.
Hashtags can decline in different ways. The most obvious is when nobody tweets about them any more. Looking at you, #altFin: added June 2016, removed October 2020 after months of inactivity.
Other hashtags get removed because their content gets confusing. #Cognitive (June 2016 - October 2020) has plenty of posts, but from different communities and scarce overlap. When I added the hashtag the discussion was largely related to AI, but today it's used by Psychologists as much as Programmers and that makes for a messy report.
Hashtags can also be overwhelmed by spam. I'll write a bit more about spammers another time, but in this context the spam is what can kill the hashtag because it overruns the conversation. Sometimes tweeps share a hashtag in a copy-paste-footer in all their messages, and if these tweeps are popular enough to have a genuine/bot following of retweeters these messages move up the rankings even though their content is irrelevant. If it's one or two tweeps originating these messages, my algo can deal with it and filter it out, but when it becomes widespread there's nothing much I can do. A perfectly good topic like #reactJS (June 2016 - October 2020) was swamped by tweeps promoting affiliate links to computer books which have nothing to do with the subject, or more commonly, folks looking to promote any old thing by trying to latch onto as many trending tags as can fit into the character limit. For example:
Top 8 #MachineLearning Tools For #Cybersecurity .— Abhishek Yadav (@abhishek__AI) October 13, 2020
#MachineLearning #DeepLearning#ArtificialIntelligence #AI #Python #Java #RStats #reactjs #VR #AR#TensorFlow #IoT #IIoT #DataScience #programming#Coding #WomenWhoCode #ML#100DaysOfCode #TensorFlowhttps://t.co/WkvfKmhjmY
Sure I could remove this tweet as irrelevant to #reactJS (or #VR, #100DaysofCode...) or just block this guy's tweets wholesale, but there are too many like him. It's a shame, but sometiems there's too much noise to tune out.
And then there are the hashtags which have a combination of mixed interpretation and spam and way too much traffic to make it worth the bother. #fintech is tweeted multiple times per second and after lasting four years (May 2016 - October 2020) it wasn't the burden on the server that meant I had to kill it off.
If a hastag is a bit obscure and doesn't get much web traffic, that's ok. But if it's not seeing much site traffic and its email subscribers aren't opening the free weekly reports, that usually means the end. There are hashtags which are so quiet that some weeks there is barely a top ten report to send out, but whose following is keen enough to open the mail when there is. I'm happy to support them.
Some stop being interesting when the community changes. Sometimes a tag can have lots of content being written, hundreds of articles a week, but it's an empty void where nobody actually bothers to read any of it. You can usually see this when the number of retweets averages < 1 or the content is being churned out by the same handful of accounts, and/or retweeted by the same bots. Nobody cares, #devops.
Up until now I've ruthlessly removed any hashtag which no longer justifies one of those 60 mailing slots. But I think in future I might give hashtags a chance to redeem themselves. I'll take them out of email rotation to make way for something new but keep collecting the data on the website. A signup page with more than 60 topics to choose from looks a bit messy anyway.
So if you have a hashtag which you'd like to add to the site, let me know.