[shayanjm]

Kudos on building something cool to solve a pain point, but unsure about the efficacy of this implementation. A list of negative-sentiment tweets about competitor products is certainly a good place to start, but is by no means a list of actionable leads. Still requires quite a bit of human interaction to figure out which tweets are actually solid leads, and is only truly useful if your competitors have only one product.

You also miss out on users asking for suggestions who aren't currently using a competitor product (which IMO is a more valuable segment).

A more interesting implementation is one that takes context into account, but that would require some homemade ML work and likely outside of the scope of quick & hacky solutions.

[farhadhf]

Yep, I agree that this is not a perfect solution, but it worked for us.

I'd mentioned in the blog post that out of 34 tweets that were added to the spreadsheet, only 6 of them were solid leads. But going through 34 tweets to find those 6 is a lot easier than going through hundreds of them over 8 hours.

[shayanjm]

Agreed. I think generating a more filtered list is possible - but would take significantly more time than it took to build the sentiment analysis tool.

The results of the more-filtered-list-tool would be quite interesting, though, as you'd essentially be modeling a set of "ideal leads" and determining how close/far a set of tweets are to those models. Just figuring out an "ideal lead" model for the segments you're targeting would be an interesting intellectual pursuit.

I think I might end up building this...

[farhadhf]

Ping me if you decide to do it, maybe we can join forces, I've started working on the basics over here, @farhad_hf on twitter.

[shayanjm]

@shayanjm on twitter.

I have a half-baked contextual analysis implementation which I could probably spin into a high-volume twitter analysis tool. Was doing NLP analysis on unstructured data (like news articles) and extracting topics + extrapolating commonalities between sets. Could be used to pick up topics from tweets and determine if two unrelated tweets are actually talking about the same thing (without necessarily replicating the same syntax).