Stas Kulesh

PH Bench - Product Hunt posts sentiment analysis

by

PHBench is using machine emotional intelligence to measure products reception at ProductHunt. A bunch of strongly worded comments can bring a mammoth product down.

Add a comment

Replies

Best
Ryan Hoover
This is fascinating, @stas_kulesh. How do you detect "arrogant CEO's" and can you create an arrogance score? 😝
Stas Kulesh
@rrhoover Haha, thanks! Check out Intercom's product in here (high upvotes, poor reception): http://phben.ch/compare/know-you... One of the trusted community members challenged the CEO and...
Stas Kulesh
@rrhoover While we're talking here. One of the things that slow PH Bench down is the API requests limit. Can we increase it somehow? P.S.: I'd be delighted to work with PH team bringing similar functionality to the actual PH. Got a tech-savvy team of pros to feed over here. Thoughts?
Stas Kulesh
At first, I had an issue with Product Hunt hype. I come here to find the best product for my needs. Tons of products get a really good reception at the start. Everybody is, like, 'well done!', 'Good stuff!' A very supportive and positive community indeed. However, sometimes later people who actually used the products come back and share some REAL LIFE experience with the product, which is not always that positive and super-exciting. These comments are golden! One nugget like that means more than a hundred of upvotes or ten exclamations to me. When choosing from the list of related products I don't always have time to read all comments and see if of the hunters have anything against using the product in question. That's why Sliday.com team built http://phben.ch PH Bench 'reads' Product Hunt comments for me and based on everything that's known about the product gives a comprehensive 0-100 score. A bunch of strongly worded comments can bring a mammoth product down. Hype is easy to gain (buy in some cases), the launching comes naturally to many product owners. The real experience of the real customers is what matters the most. Compare products, share your findings. PH Bench finds coupons and discounts, detects scams, dark UX patterns and arrogant CEOs. A bunch of strongly worded comments can bring a mammoth product down. Powered by Product Hunt API, Google Cloud Natural Language API, somewhat tricky math, and common sense. PH Bench is free.
Stas Kulesh
PH Bench detects scams: http://phben.ch/compare/datememe...
Stas Kulesh
PH Bench detects dark UX patterns: http://phben.ch/compare/intro-6/...
Stas Kulesh
Similar quality products have similar scores: - http://phben.ch/compare/asana-bo... - http://phben.ch/compare/monument...
Stas Kulesh
Forgot to mention the good stuff. PH Bench can detect if one of the makers mentioned any specials, discounts or coupons in product posting comments.
Stas Kulesh
Try changing a topic (category) in the search field and check out top products in trending topics: - Books: http://phben.ch/compare/hacking-... - Games: http://phben.ch/compare/pokemon-... - Maps: http://phben.ch/compare/retro-pa... ...
Nick Abouzeid
Hey Stas! This is amazing. Well done! Quick question: how did you pick the emoji rating scale in the first category ("Reception")? It was slightly confusing at first - didn't know how to compare a 🙌 launch to a 😝 launch, as both have positive connotations to me. Maybe 😡 😐 😄 would be a better trio of emojis!
Brian Roach
@nickabouzeid I agree with your r proposed scale. I've seen that same scale in emoji satisfaction surveys and it definitely is clear what each means.
Stas Kulesh
@nickabouzeid should probably switch to something like this https://gist.github.com/drwasho/...
David Lux
This is a great and well executed idea! Not only for makers who are able to benchmark their product and compare it to others, but also for people who're researching useful tools. Some constructive feedback: - from an optimization perspective: I would get rid of the GIF logos. It sure is kind of 'fancy', but it also is kind of distracting and creates a feel of confusion. You could A/B test this to back it up with data. I'd bet removing it would decrease your bounce-rate! :) - It would be cool to see top comments (by votes). Both for negatively and positively connoted comments. - Maybe you could also add some meta-data like Facebook Page Likes, Instagram Followers etc. to incorporate some kind of social proof that complements PH - Reception shows now data for me when comparing products (Chrome Version 60.0.3112.101 4 Bit) Best of luck and congrats on the launch! 🎉
Stas Kulesh
@itsdavelux Thanks for your detailed response David. - from an optimization perspective: I would get rid of the GIF logos. It sure is kind of 'fancy', but it also is kind of distracting and creates a feel of confusion. Atm we use the imagery provided by PH API. We could copy the pics to our cloud, but that would to the overall complexity of the thing and increase the costs. :) Also, I like gifs. - It would be cool to see top comments (by votes). Both for negatively and positively connoted comments. Solid idea! Most positive, most negative. Not sure if we store the metrics though. Google API is not free, unfortunately. - Maybe you could also add some meta-data like Facebook Page Likes, Instagram Followers etc. to incorporate some kind of social proof that complements PH Unless we integrated with another API, I'm afraid, we don't have that info: no FB pages, no Instagram accounts are in PH API product object, we _just don't have the data_. - Reception shows now data for me when comparing products (Chrome Version 60.0.3112.101 4 Bit) Could you please elaborate on that?
Stas Kulesh
@itsdavelux Added the most positive/negative comments to the desc http://take.ms/7vsl7
Dre Durr💡
Great Idea! The idea behind Bench is solid. A great way to take a quick scan of everything. Dope 🚬🚬
David Kravitz
@dredurr I hope you will use it in future, thanks for your feedback!
I could swear I spoke to Ryan about this idea on Twitter a few months ago...
David Kravitz
@jgage718 great minds think alike
@david_kravitz thanks :) – wondering what you wrote the backend in and if you used something like TextBlob or you trained stuff yourself?
David Kravitz
@jgage718 backend is on Rails with Google natural language processing API and some in-house written algorithms
David Adamu
This is great and quite interesting because recently, I've observed certain products here having high reception but very low reception, as shown by later comments. Worse is, these products were not free. I've also seen a user advertising counterfeit currency using AskPH. Great to see a solution to ensuring quality on PH!
Stas Kulesh
@alternatedayve thanks for the feedback. The issue is real. Here is an example from the recent benchmarks: http://phben.ch/compare/shazam-f...
Stas Kulesh
@alternatedayve this is what PH Bench picked up in Songbird (paid app) comments:
Bo Wang
Do you have a reference to the math equation that you put in the "How it works" page?
Stas Kulesh
@bluemonk482 It's pretty much what's in there, apart from the logarithm that evens out the benchmark distribution. As usual, the coefficients are the main ingredient of the secret sauce.
Laef 🐶🐱
Neat! Out of interest, do you have a page showing how PH Bench has been scoring..? 😉
David Kravitz
@laef we will tomorrow! I'll post a link here
Krishna Dahal
looks awesome but would love to learn the math behind it. However my upvote goes to you here.
Marek Pasieczny
How long does it take to find your product o PH Bench since the first time is uploaded? We uploaded our product 30 hours ago and I can't still find in on phben.ch 😞
Stas Kulesh
@marek_pasieczny Please try now. Product Hunt has a strict limitation on the API requests.
Mohammed Rafy
No Shit! This thing is dope. Nice stuff, Stash. Whats with the last active? Is it something like a user who commented or upvoted that particular product since when? I don't know if this makes sense, you should even compare if the particular products are free or not.
Stas Kulesh
@rafyasarmatta 'last active' = last tracked activity in the post: a comment. Unfortunately, we don't know if the product free or not :) PH does not store such info.