iSideWith is a lean startup. And by lean, I mean it's just two of us and this isn't even our full-time job. Yet we are profitable with over 8 million users and are growing month-over-month because we maximize our time spent constantly adding new innovative features to keep users engaged in politics. Here are a few tips to make your startup more profitable by saving you hours per day on tasks that are time consuming and basically not fun.
Amplify your Crowdsourcing
One of the challenges of staying on top of the political landscape is knowing which issues are most important to Americans. Especially local issues. At the end of our survey, we ask users to suggest an important local issue that we did not cover. Instead of reading every suggestion (over 1,500 submitted per day), I wrote an SQL query that organizes all the suggestions by city and state, then shows a list of the most frequently used phrases. This let's us see that "asian carp" was the most frequent phrase suggested by Michigan voters, and we then add that as a local state question. The 4hr daily task of reading thousands of crowdsourced suggestions now turns into a 3 second task.
Moderate your Moderators
Im a huge fan of Mechanical Turk. MTurk allows you to programmatically send human intelligence micro tasks to workers who are willing to perform them for a very small fee. We are using it to moderate comments in our soon-to-launch discussion forum to prevent profanity or off topic posts from showing up on the site. When we first started using MTurk, we found a few moderators were approving everything and anything, plowing through jobs so they could quickly collect their fees. We needed a way to moderate the moderators, so I tweaked our MTurk script to penalize or reward moderators based on their history of accuracy. To better explain, let me list the steps involved. When an iSideWith user posts a new comment to our discussion forum, it automatically gets sent to MTurk for moderation (and keeps getting sent to moderation until we have a reliable consensus). The first worker to moderate the comment has never moderated comments for us and therefore has a low reputation of 1%. The next worker has a good history of moderating comments for us and therefore has a high reputation of 70%. The next worker has a decent history of moderating comments for us and has a reputation of 20%. This keeps going until the collective moderation score reaches 100% and we then use the majority consensus to approve or reject the comment, rewarding workers who were apart of the majority consensus and penalize workers who went against consensus. Moderators are now moderated. Although the cost of having many moderators approve one comment is costly up front, over time you save big as you end up with many moderators with 100% reputations who can approve a comment in one job.
Let Data Give You a Headstart
With the local 2014 mid-term elections coming up, we are adding dozens of local citywide ballot issues (and issues collecting signatures) per week to our survey. Our goal is to cover every issue up and down the 2014 ballot. In the first tip, I provided insight into how we determine which new issues to cover. But for every new issue we add, we have to match all sides of that issue to the political parties. This gets challenging on hyper local issues with little legislative history. To give us a head start on our research, we will launch these issues without any matching listed (you'll see these listed as "Democrats have not given a stance on this yet"). Then we watch how users start answering the question. From the other 48 questions they answered, we can accurately determine their political affiliation and watch political affiliations gravitate towards specific stances on the new issue. This insight helps Taylor know where to start his research to confirm the parties' positions. This cross-referencing method also alerts us if there is an error in our stance matching algorithm, or there is just strong dissent within a party on an issue. Find out how you can use patterns in your data or users usage that can alert you when things go wrong.
We hope these tips inspire you to save time and money in your startup journey. If you ever come across a task that feels boring, expensive, or time consuming, hit the pause button and think of a way to automate most or all of it. There's always a way.