[THIS POST IS AN APRIL FOOLS JOKE]
Hello, and welcome to the first edition of the Round Sky monthly newsletter.
Our hope for the newsletter is that it will provide our clients with a little bit of insight into what we’re up to here at Round Sky.
In this edition of the monthly newsletter we’d like to talk about what we do a lot of here at Round Sky… data mining! With large data sets, patterns can be found, and those patterns can be used to predict behavior.
Many companies in our industry do data mining, but a lot of it is a bit limited. And those who do datamining tend to stick to the simple stuff like:
- Google looks at past and current browsing behavior to predict future behavior.For example if you’re searching for mountain bikes, over the next 24-48 hours Google will show you ads for mountain bikes on many of the future sites that you visit.…They are able to identify your interests by studying your browsing history, what you have searched for and clicked on.
- YouTube looks at what you’re viewing now to suggest what you would like to view next.
- Amazon suggests things for you to buy that are relatively similar to what you’re buying right now.
Here’s where we changed things up a little:
When someone applies for a payday loan, we know a lot about them:
- What zip code they live in.
- If they rent or own their home
- What their monthly income is
- Who they bank with
Our ever eager team found a way to put this data to AMAZING use. It’s common knowledge that a customer who applies for one payday loan is likely to look for another loan sometime in the future. The value in this datamining situation is knowing WHEN the customer will apply for a loan again, and if they will be a GOOD customer.
Here’s where we started thinking outside the box:
First, we experimented with adding non-standard fields to the payday loan application, such as:
- What is your favorite color?We used application data and loan results combined with the knowledge our CEO gained from skimming a few psychology books “back in college”, to determine that the customer’s favorite color matters a lot. .For example, we determined that the choice of “white” as a favorite color was indicative of whether the customer preferred vanilla ice cream or chocolate ice cream. And since our CEO likes chocolate an executive decision was made to show all the applicants who choose white as their favorite color, a list of bullet points as to why chocolate is better.
- Draw a picture of what your mood will be tomorrow.If the applicant were to draw rainbows and butterflies, we would filter them out. Satisfied people tend to rely less on material possessions. On the other hand, if the applicant were to draw a vampire they are probably young, and in need of money to buy the latest Twilight merchandise (We’re team Edward by the way).With the increase of the popular DrawSomething app on the iTunes store, customers really responded to this portion and it absolutely befuddled Spam bots. As a result our fraud rate dropped to below 1% and we were able to place 99% of those incoming leads with buyers.One of our satisfied lenders has opted to switch from sub-id reports to focus on picture categories (Ex. “Bananas, car pictures and wizards are all converting very well, so if you can continue sending those to us we’re willing to give you better than first look pricing at our top tiers for those leads”). Our next step with this technology is to build a front end to let lenders dynamically bid on picture tags, and when this is in place picture boxes will be mandatory on all lead forms that post in to us.
- How would you describe the difference between a bicycle and a train using a 7 syllable word?
Customers who replied with “wtf” or “whaaaatttt” were clearly poor quality customers.Where as customers who used exactly a 7 syllable word as asked for we deemed intelligent; thus this consumer won’t be fooled and is usually a responsible borrower.
- In essay form, describe your favorite kind of weather using numbers.If the customer leaves the section blank or puts zero, they demonstrate a fear of numbers and therefore lack the ability to keep track of money. A better customer would be able to formulate a pattern, demonstrating the ability to pay a lender back on time.
Next, we took these answers and began cross analysis:
- If the consumer lives in a zip code with high insurance rates (data publicly provided by the US Department of Insurance)
- and resides in an apartment (if the consumer rents his home, 70% of the time this denotes they reside in an apartment complex),
- and banks with Chase,
Then we know what they are living in a high density area, as this is where most Chase ATM machines are located, and as such are subject to peer pressure (from their neighbors) to party and drink A LOT of beer (high insurance rates), and therefore will come back multiple times for a loan.
- If the consumer banks with Bank of America, who over charges for everything
- and is over the age of 40,
Then we can extrapolate that the consumer is not cost conscience (as they’re with bank of America) and therefore they can get approved for a higher amount loan, and pay a higher interest rate.
- If the consumer has a dot (.) or underscore (_) in their email address, we were able to deduce that they would typically be a late adopter of technology, as their desired email address without the dot or the underscore was already taken, therefore they probably do not own a cell phone. We then were also able to deduce that they have right now more than likely just entered a fake email address, because they probably do not possess an email address to begin with.
By combing thru the data by hand and with the help of lots of caffeine we have determined many significant patterns. We have taken our new found knowledge and programmed it into a single complex algorithm.
While our minds are capable of following only one thread at a time, our server farm is not as limited.
In phase one of our algorithm, it will establish the “quality” of the consumer based on the answer to the Essay question above; from that, it will have the ability to predict the time and place where the consumer will apply for a loan again (example output: Uncle Joe’s backyard, using a wireless laptop, on June 13th at 6:03PM, because Uncle Joe needs his money back right now).
In phase two our algorithm, will be able to predict its own behavior in predicting consumer behavior, and will start writing code for itself, so that it can optimize predictive capabilities. As this algorithm reproduces itself over our network with each iteration, we’ve decided to call it RoundSkyNet.
In phase three, the algorithm will start predicting key numbers from larger data sets. These should allow us to quickly and accurately predict lottery numbers, horse and dog race outcomes, and professional sporting event results. Our mobile app should be launched around this time which will allow quick predictions of poker, black jack and roulette outcomes. Positive results in the testing phase convinced us to move our office to Las Vegas. Once we have cleaned out the casinos with our winning streaks, we will then cease publishing this newsletter, as we will be living in mansions by the sea.
As we conclude this newsletter we would like to invite you to give us feedback and share your excitement over our new revolutionary technology. Contact Round Sky for Payday or Insurance leads that convert.
And happy first of April.