PaperG recently launched Polly.IM, a new Twitter application built on top of PaperG’s PlaceLocal semantic technology. Polly is a social media manager that can organize customer tweets relating to your local business and suggest responses. 
 

In this case, we were able to make use of the data exhaust spit out by our existing semantic/sentiment engine that has been analyzing and grading thousands of local reviews for usage in ads. Even though it’s an app and not a whole new company, I think Polly.IM is a good study for people interested in launching lean startups quickly.

Idea (1 week)

 Small business owners have very limited time to do marketing/advertising and generally lack access to the same resources large firms have. We looked at a few different applications of our platform’s internal APIs to level the playing field. We realized that business owners could use a solution that responds for them on social media. This idea was our original thesis to approach this market.

Market Research (1 week)

 We looked for similar solutions out in the market to determine whether there was even a need already being address and how well it was. Some obvious candidates were LocalResponse (great for check-in analytics but not as helpful for other possible tweets), SocialOomph (feature heavy but poorly designed), and Twitter clients like Hootsuite (designed for agencies, not local). We played with them all and found some obvious pain points still unsolved. We decided there was a clear need but no one was solving it well for local businesses in particular.

To further test our thesis, we contacted small business owners who have used our other products. We got some good initial feedback on the idea that we took to the drawing board. We learned very quickly that business owners would rather approve suggested messages to customers rather than completely automate the process. This finding significantly improved our product before it even was released.

Design (4 weeks)

We realized design was paramount to making a simple, easy to use solution that addresses the pain point of the market. I really liked the design work of Brizzly and its cute little bear. Quora was very effective in helping us track down the actual designer — which normally leave his/her name on the site.

We knew exactly what we wanted stylistically and made it easy for the designer to work with us. When you work with someone you are a fan of and have a sample project to reference, you waste little time on revisions and trying to explain something in your head. 

Once we had the logo done, we threw up a splash page (from LaunchRock) which helped us gather beta signups and encourage people to spread the initial word.

Coding (4 weeks)

As one of his side projects, one of our engineers built out the back end functionality, creating a minimally viable product. Our first version literally only let you record a message which gets sent out automatically based on specific triggers. There was no option to turn it off or even see what it was responding to. It was a little embarrassing, but I do believe in the mantra that if you aren’t embarrassed by your first version, you’ve launched too late.

Private Beta (2 weeks)

We released the product to our beta cutomers discreetly to get some initial feedback. There were some great suggestions like keeping suggested tweets on manual but allowing automatic promotions to be sent to new followers. By the time we flipped on the switch for anyone to try, we had a very usable product because of some simple changes made during the private beta.

Flipping the Switch (1 week)

We felt strongly that it was important to get the product in the wild even if it was missing a number of features. The best case was we get incredibly valuable feedback from the first 50 users who spread the word to the next 100 in their area. The worst case was the first 50 users never come back cause they hate the product so much. However, there are more than 3 million local SMBs in the US alone so chances are there would still be a big market out there for the finished product.

Of course, flipping on the switch doesn’t result normally in thunderous applause. It’s more like crickets for most startups without any existing users. We were able to get some initial users for free (the best way to test it) by doing something completely simple — emailing businesses who were on Twitter. The trick was identifying a scalable way of distinguishing which Twitter accounts were small businesses, which were individuals, and which were big brands. 

We figured out a way to use our existing PlaceLocal technology that crawls multi-pages for geographic and business details to identify likely small business accounts. This sweep could yield us email accounts to ping about our product. You’d be surprised how well an authentic email from a founder asking for users and feedback can do.

Post-Launch

We’re now helping organize and suggest hundreds of tweets a day which is saving countless time for business owners. We have new versions coming out practically every other day to address new usage and patterns emerging. Only time will tell if this project is in the right market at the right time, but I’m quite proud of the initial work and the team behind it.