my name is Basti, and I would like to showcase our project Buzzngo with you. Our system is running with RethinkDB under the hood and although the company behind it faded away we still think it's a nice tool to power real-time statistical analysis.
Our tests in Frankfurt, Berlin and Hamburg (Germany) were great collecting thousands of votes during a sunny day. Feel free to sign up if you want to start public polls in Germany/Europe in the near future.
Yes, using our web platform you are able to create campaigns and attach stations to it. Everything is automatic and just informs the team that a new station need to be placed on point X for Y duration. Once placed and activated the stations need to further watching.
Yes, Raspberry + WiFi + Batteries. :)
We are experimenting with a lot of things. Hitting, high-fiveing, jumping on the buzzer and many more. They all have there pro and cons. Haha
Best of luck! We still use RethinkDB under a lot of our tech stack, and love it. Great platform and I hope it continues to improve under open source care.
Yep, totally agree! We soon upload some polished media footage of actual buzzer stations placed in cities so that people can get a better understanding of it.
Backup and restore is not an everyday DB operation, I wouldn't think?!?
I've got a small side project [0] (actually, am HN aggregation and reading tool) which uses RethinkDB [1]. Currently has 4.2 million entries (7.7 GB) in the HN posts table and really no issues that I know of.
The platform is build using Vue for front end Node with Express as Backend and Socketcluster inbetween for real-time comm. And of course RethinkDB as database.
The landing page is build with Nuxt which reuses some components from the platform. So when you like the look you will love the member area. :)
Two of the example polls in the site screenshot are political (us 2020, cannabis).
I'm wondering how you get background demo info on poll participants?
I think it could be interesting to think about using visual cues from video (gender, age) but what about info like is registered to vote or other demographic info.
I don't see example of the hardware but I'm guessing this is like those smiley face poll stations in that it just collects a simple response and doesn't collect say first, last, address, birthday etc?
Based on the location we get some first demographic information (like banking sector, shopping mall, university campus,..)
In case of political questions it is not enough to place only one station in one city but instead up to 20 stations in more than 5 cities. A typical question would be “Would you vote for Person A or Person B?”
The system then periodically makes a pattern analysis and informs you about insights (“People with an academic background tend to vote for person A whereas people in north Germany tend to vote for B”)
The technology is most powerful when many stations are activated and orchestrated at the same time at different locations.
Just some un-prompted feedback. I work in US politics. I'm not a full time researcher but my baseline insider knowledge is probably pretty high.
I don't think that would be good enough for a poll used by most US political campaigns. Even the tiny amount of news funded surveys care about xtabs with meaningful voter data. I have seen a few huge response rate online surveys for a simple presidential head to head but it's a tiny market without more voter data attached and only for giant races with news tracking.
I think knowing what census tract one of these machines is in wouldn't be enough data for political surveys, same w example of being near a university.
We actually field online surveys but we use voter file data to target and we ask for respondent info to match back to voter file. We mostly only use them to measure our digital ad's recall and persuasion.
I wonder if your product would be a better fit for say a brand advertiser; everyone is a potential customer versus in the US few people vote. Plus it's easy to measure 1-10 do you like this CPG or have you seen an ad recently.
I also think there are some potential cool use cases using ML to infer gender, age bracket, race. For instance in live person polls often times the interviewer makes a note of the voice's likely gender. Again more valuable for a brand advertiser to know those demos for simple CPG attitude surveys. Really the best value for my work would be if could match a persons photo to like FB's db or maybe in China lol
Maybe its so much text in one space, or maybe the cool animation in the background, or maybe just my tired eyes, but I'm having a bit of trouble focusing on what to read in the top section. I am not immediately faced with the value prop because there is so much content!
That's a good question. I remember some stats professor tried to popularize the phrase "voluntary response data are worthless" (see also: 'tokenadult on that point, https://news.ycombinator.com/item?id=2322237). Point being, there's a strong self-selection happening with voluntary polls - people who are opinionated enough to be bothered to answer a poll are not a good sample of the population.
Another thing I wonder about: since this is a kiosk that's going to stand out there for some time, what happens if a group of interest decides to game a poll, by telling its members to go there and vote (possibly multiple times a day)?
Right. Even worse at an unguarded physical location. What's to stop me from calling my 30 friends from a student political group for some ballot-stuffing?
You can strategically install the polling station.
I would imagine something like this at a music festival, business or academic convention, etc...
That way you already have an idea of the respondents, who could be your customers/users, and then account for the bias they share because of their location.
As an example: organizers of an electronic festival use this to know if the 2018 festival-goers prefer Trance or Drum'n'Bass (or between artists).
With the results, they can book artists for 2019 accordingly. And it doesn't matter much if the results are not representative of the population. The bias of the respondents / non-respondents inside the festival can be mitigated by advertising the polls.
Maybe you should use that on your site instead of examples like "Presidential Election 2020" or "place the station near a university to get insights from academic people", which are clearly not going to yield sound, actionable results.
Ok, but what is your answer to the obvious selection bias with physical polling? Also, how do you deal with coordinated gaming of polls by organized groups? Would love to hear your thoughts. Great project though!!!
Currently when someone hits the button 20 times a sec or hits it every second for 2min our algorithm finds out and highlights the suspicious area. The user can decide to exclude the data or not. Not 100% reliable but we’re experimenting a lot. So I call it beta feature. :)
my name is Basti, and I would like to showcase our project Buzzngo with you. Our system is running with RethinkDB under the hood and although the company behind it faded away we still think it's a nice tool to power real-time statistical analysis.
Our tests in Frankfurt, Berlin and Hamburg (Germany) were great collecting thousands of votes during a sunny day. Feel free to sign up if you want to start public polls in Germany/Europe in the near future.
Happy to hear your opinion!
Greetings Basti