Bellwether Updates: Linear Regression Breakthroughs and Brands put on point!

I did some exciting additions to the Bellwether system including a trendline on our graph using linear regression and a ranking page by license showing top and bottom 10 on Amazon

Trendline Graph – The new trendline seen on the dynamic graphs generated at the item level is the result of some extra variable setting and some eighth grade linear algebra. With this, we can now see a possible forecast of rank position and visually witness an items velocity of popularity based on the slope. I only have it for the last 14 days as my chart was made for that time frame initially, but I may widen the window to get a better view of longer term trend.

License Ranking – this is a tricky one. Essentially, all I did was take the earliest rank and the latest rank of each item in a brand, found the difference and calculated the % change, then averaged the sum of all the % changes of all the items in the brand collection. I feel there’s some truth to it as I’m only showing licenses that have more than 10 skus, and the more skus there are the better overall weighting the average will hold, but I don’t take into consideration any explosive or deflating ranking values that may have occurred in the middle. Ideally, I’d like to include this as it would lead to overall brand trending, but its interesting to see how accurate the current equation is turning out to be. Check out the brand ranking here.

Posted in Blog and tagged , , , , .