Meeting with Martin: Birth of a Startup
Here’s the tldr:
Summary of an awesome day - met up this morning with my mate Martin (@matrinrue), drank lots of tea & coffee and ended the day shaking hands with Twocial.com co-founder Martin.
Here’s the regular version:
So, I am sat here REALLY excited, I’ve had to rewrite this post a few times because I am typing too much. So we each took a day off work, we met up at the awesome North Tea Power in the Northern Quarter of Manchester – easy access to coffee and free WiFi.
After discussing all things Esperanto and a hundred other tangents, we both talked through how we’d both arrived at the same idea. To set the scene here, Martin is a seriously talented coder who always works on a number of concurrent projects. Martin has already launched BrandListen which already has 500+ users. I’ll let Martin explain in more detail, but what we hope to achieve together with Twocial is an extension of BrandListen. As an aside: see my next post for the decision making process behind the name Twocial.
The day centred on discussing what the product should be and who would use it.
Sentiment Analysis; We want to be able to measure the sentiment of a tweet, is it positive or negative? Not only that, but how can you use this to better engage with your social audience?
From a brand perspective:
- How many people have tweeted about your brand?
- How many times has it been retweeted, so what is the potential audience of that negative/ positive tweet?
- Of that potential audience, who are the most influential tweeters of your brand?
- Of that potential audience, who are your most vocal? Who tweets regularly and as such are your brand advocates/ early adopters?
Once the above questions are answered, that’s all very well but what does it tell you? How are you better informed to then engage with that audience? Well that’s what where our startup comes in.
How to measure Sentiment?
Martin with BrandListen has already designed an implementation of the Bayesian algorithm to measure the word associated with a tweet. Say for instance brand “ShafBrand”, any tweets with words with negative word association would be identified as such: “ShafBrand has terrible service”, “ShafBrand ruined my life” etc. The same goes for positive.
How this works is something that Martin tried to explain to me, I understood but not well enough to explain to you. Again, Martin is best placed to go through this in a separate post.
In short, I am very pleased to be working with an awesome co-founder in Martin.