figshare
Browse
Screen Shot 2013-07-24 at 1.01.14 PM.png (521.48 kB)

Neural Networks To Analyze Market Data

Download (0 kB)
figure
posted on 2013-07-24, 17:22 authored by OS BH-LabsOS BH-Labs

I have been working on a series of custom neural networks for use in various different projects, recently out of necessity I decided to apply machine learning to the analysis of market demographic data for Indie Review Magazine, which is produced by www.OnlyTheIndies.com. 

 

The goal was to increase facebook page fan likes as well as content enagement on page posts, and offpage locations such as the blog and the website. 

 

After analyzing various paremeters I developed a software package that not only succesfully identified our target demographic, but also clearly helped us determine what content of ours that they do and do not like. In addition, the software can now also makes suggestions about which aspects of our posts, marketing strategy, etc should be changed.

 

We rolled out the automated content suggestion module a few days ago, since then we have made the necesary modifications to the content we post on facebook, and we have mananged to grow the page likes from 10k to almost 15k in less than 48 hours.

 

Prior to that we had grown the page from under 100 likes to 10k likes in less than 2 months, and as the software improved we recently grew from 10k likes to 12k likes in 24 hours. 

 

As you can see from the attached figure, we have maintained that growth rate now for an additional 2 days, and our page engagement has also increased to over 40% according to stats in the admin panel. Those stats are phenomenal based on the performance of other pages with similar numbes of likes, and even when consdier the pages of major brands with significantly more likes and serious advertising budgets.

 

I should point out that we accomplished all of this on a shoestring budget. 

 

Our ad stats are as follows:

- the average CTR (click through rate) is 0.557% 

- the average CPC (cost per click) is 0.02 cents though we have recently created ads with CPC's as low as 0.01. 

We are planning to continue to make adjustments to this project so that we can not only accurately predict which content best suits a specific target demographic, but also how to keep that demographic engaged with said content.

History