We find ourselves at crossroads or overwhelmed with choices more often than ever. With the world getting more social by the day, there are plenty of places where we can ask around before making a difficult choice.
There are already services like Yahoo Answers and Vark that help you tap the wisdom of the crowds when taking a decision. These apps are pretty great but they fall flat when you start asking iterative sub questions. Hunch claims that it can help you in such a predicament.
Hunch calls itself a decision engine and claims that it “gives customized recommendations and gets smarter the more you use it.” When answering a question, Hunch takes in to account your personality (with the help of few probing questions), the personality of Hunch users similar to you and its algorithm. Developed by a team of brilliant MIT graduates, Hunch claims to provide you with an answer in 10 questions or less. Let us jump straight into checking out that claim.
The sign up link is at an unusual position on the far right in between Home & Explore options. Besides that oddity the sign up form is very simple, with just four fields. The Hunch team explicitly states that “Spam is evil and you’ll never see any from us.” Few other web apps also carry similar messages and such finer details really go a long way in gaining user confidence.
Oh Wait! The sign up isn’t over yet. We are redirected to the profile page when we were expecting the dashboard. This step can be skipped but the details can be used to get better & localized results. The tags option can help better define your personality.
The third and final step is to find and connect with your friends from Facebook, Twitter or those who already have an account at Hunch. For Facebook and Twitter contacts import, the login credentials are not stored at Hunch as it happens via fbconnect and OAuth respectively.
Tough luck. None of my twitter buddies are aboard Hunch yet. May be they are smarter than me and can take decisions on their own (Not!).
User Interface and Dashboard
Unlike most of the web apps, Hunch has a lot of categories and links in its dashboard. However, the two column UI layout is so refined you will not be overwhelmed by the abundance of links and options. The entire Hunch network is condensed in to four categories : Topics, Activity feed, Browse and People.
On the right side column, Hunch has some interesting pictures for you to choose so that Hunch can understand you better. The idea and implementation are great & makes the user more curious. The images could be bit larger though. It took me a lot of squinting to make them out.
The questions range from interesting to intriguing. But there is absolutely no end in sight. The questions keep coming and I lost interest after twenty or so clicks.
There is an assessment box below the assessment box. Here all you have to do is select how well you like the picture and the associated question. Would be really nice if Hunch briefs how exactly these details can help in brief here at this page instead of hiding it in a fine print About page.
Hunch has a wide selection of topics and subtopics. You can either choose a topic that interests you or can check out the recent or recommended topics.
I chose the Books topic and the list of subtopics were listed on the left side. This page design and layout is very minimal and boring.
Now if you click on the magazine name you will be taken to profile page of the magazine that shows the popularity of it among the Hunch community, Pros & Cons report from everyone, Related magazines etc.
You can add your own pros and cons at the bottom of the page in 200 characters for each item.
From the Actions drop down menu, you can choose to edit the profile page of the magazine. The edits are performed like in Wikipedia: you can change the description, URL and add or delete images. You will have add a note as to why you have edited the page and you can also see the revision history of edits in this page. I added the URL for Wired magazine and was added immediately without any moderation. Sweet.
Fixing a Hunch
Based on the questions we have answered after signing up, Hunch prepares a data set of our likes and dislikes using its algorithm. To view how well the Hunch algorithm has “got you”, select Fix this Hunch option from the Actions drop down. Hunch will display a set of questions whose answers help train Hunch better understand you. If some questions were not answered by you, the percentage of how trained Hunch is to your likes comes down.
You can fix this by answering the untrained questions by clicking on the Untrained section. The answers marked in red indicate that Hunch has picked up your answers from the previous questions. All these answers you provide for the training questions and the answers given by like minded people enhances the efficiency of Hunch’s algorithm. Simply put, the more questions you answer, better are the recommendations from Hunch.
Asking a Question
Now let us check out the flagship feature: answering your questions based upon Hunch’s analysis of your answers. Hunch makes it very easy for you to ask the questions. No need to search for the relevant topics, sub topics and filter through the questions. You can start typing at the search bar on the top and Hunch will automatically start suggesting the question.
I selected the question “Should I buy an iPod Touch?”, and Hunch loads up a page with Yes and No answers. There are training questions relevant to the topic/question on the left hand side of the page.
This is where things start to get interesting. If you start answering the questions, Hunch will present to you the best choice based on your answers.
I chose that I wanted to browse, listen to music, watch video and play games. Based on my answer, Hunch suggested me to go ahead and buy one. You can end your search here or continue to answer the questions. The questions asked in here are not in an endless loop.
For the second question, I chose that I am willing to spend only less than $200 for the device. Now as you can see the percentage falls to 52% since the base model iPod Touch itself costs $199. Not the very perfect fit with the budget I have specified.
If you answer that you already own an iPhone, Hunch will emphatically say No since iPhone is an iPod Touch with a wireless radio and camera.
In the long and short, the more questions you answer, the better the answers you will get. Personally I found that the answers got better with every other question I answered and the final answer was right on the target. The questions and suggestions are brilliantly conceived in such a manner that you won’t feel like being advised by an algorithm.
Adding a question
Asking a question is as easy as selecting the option from the Actions drop down menu. The questions are more like a poll and that is good as people have very less attention span these days. You can add images to the question by toggling the Advanced option.
Hunch, being the training addicted app it is, will require you to train this question too.
Despite having a world class decision algorithm under the hood, Hunch also sports a brilliantly designed UI and community features. You can start following people whom you find interesting, just like in Twitter. Similarly, it is very easy to explore new and emerging topics. Hunch recognizes active contributors and participators with an excellent credits and badge system. There is also an excellent messaging system in place to communicate with your friends.
With all these great features, one huge problem with Hunch is that it looks complicated. Except for a one liner explaining the service, there is nothing prominent in the homepage to introduce the service to new users. The About and How it works pages are full of textual information that can put an insomniac to sleep. A video would really go a long way. That being said, Hunch is deserves a 9/10 by any scale.
Don’t Forget, Subscribe and Follow
- Hunchable: Decision Making Made Easy
- Ask Your Questions at Aardvark
- Create a Q&A Website in a Snap with Qhub
- Stack Overflow-like Open Source Q&A Systems for Download
- Become a Social Network Ninja with Brizzly