One MVP a Week. (week 4)

Hey sorry I’m late. This 2 weeks have been damn crazy.

Let’s talk about first. Last week I launched this platform which allows auto-saving in a kinda tricky way. It works thanks to the stripe crazy-amazing checkout button. Business part is tricky still but looking forward to improve it in an upcoming week or 2. (week 3)

The results were not that amazing with huge numbers but I got to learn a lot about product development.

bounce rate bounce rate

227 unique users, 518 pageviews in 4 or 5 days is not that bad. Let’s talk bounce rate: 35.92% seems amazingly crazy for me since that indicates my actual product was used by ~65% of the people that visited the page. That says a lot about ease of use and validation of the core concept, at least that’s my take on this.

I processed exactly 0 payments on stripe but added a nice alternative option to save on your own via typeform that just reminds you to save that got used 5 times. Which pushes me to actually invert the roles of payment and free options and make the free option the core business of the service perhaps.

questions for free saving

Will definitely revisit bank piggy in the mid future since it is my favorite MVP so far and loved what I have achieved here. Let’s talk about this week’s MVP. (week 4)

This week I wanted to create a little bot. I wanted something that would take complex data and analyze it mechanically over and over. Complex data? I said let’s do twitter then. Then I just borrowed some of the logic behind week 2 ( and try to guess how is people feeling right now because of what they are writing in Twitter.

I chose python for the project because I found a damn nice library that supports the Twitter Streaming API called Tweepy. Let’s talk dev:

I defined my workflow as follows:

  • Fetch tweets from a certain area
  • Analyze the tweets looking for specific words
  • Output some nice file.json indicating how is people feeling

Finally I could use that .json to feed a webpage that would show in real time the feelings of tweeters.

Here is the process in 3 pics so we can see how tweets are fetched, processed and searched for feelings. My algorithm is way too basic, search for phrases that sound related to the feeling.

fetching streaming tweets

I ended up with 3 different scripts that output 3 different texts files. I know that sounds that like a mess and I will try to improve it. But the magic part is that I got it to work and it feels amazing.

pure text from tweets

The actual UI of the webpage was built over the CSS and design of bankpiggy. From there I just changed the map and integrated markers from google maps. It was a fairly quick process but I have to admit I had to revisit javascript and learn some ajax logic thanks to my friend Sergio.

Launch time

There is the first take on the MVP: a UI that shows off emojis according to the feelings logged from streaming tweets. Check it out here.

Here’s the repo (minus my private keys for you to use) and definitely improve because I’m pretty positive the code is shit right now.

Actual count from tweets

That was week 4. Live feelings from people around Mexico City, New York and San Francisco based on the Streaming API from Twitter.

I’m open to any feedback and would love to discuss ideas for Week 5. I thank you for reading this as usual and hope you 💚 recommend this if you found it entertaining.

(Also please fork on github if you would like to help me improve the bot)

Show your support

Clapping shows how much you appreciated Martín Granados’s story.