Finding Your 80-20 Rule
Web metrics are not as confusing or difficult as many think. This post shares three of my favorite web analytics tips:
- % Total.
- Running Total.
These three tips sure to help you find your “Pareto Distribution” aka your site’s 80-20 Rule the web’s enduring and important “fractal”.
When looking at products by sales we like to add and save rank. How ranks change over time are fast ways to visualize trends. Here is an example from our Haiku Deck tracking Analysis (on Google docs):
Ranking comparisons create “one look” trend analysis. Anything to make trends visible quickly is a good idea. Trends often need deep dives to understand the what, when, where and why. “Snapshoting” data tells you what is moving fast, faster and fastest.
Spend analysis time on outliers. Well let’s amend that. Spend analysis time on outliers that are several standard deviations beyond your modeled expectations in either direction (good or bad). Once you identify outliers dive in to identify actions, yours or others, supporting and explaining the trend.
% of Total
Another “snapshot” metric to help see trends fast is % of total. When you divide a row’s value by its sum you see what has power. You see the patterns fast.
By dividing Brand One’s $300,000 into total sales of $1.7M we can visualize the distance between #1 and #2. When creating a % of total don’t forget to make the reference absolute by using $ within the total’s destination cell
As you copy the cell the formula’s absolute reference is set to the total. Easy to know when you forget the $ signs since the column has “divide by zero” references as the formula’s wrongly set relative reference moves down below the actual total.
Ecommerce sites can have thousands to millions of SKUs (Stock Keeping Units aka products). % of total helps you visualize relative importance. In the example above Brand One has almost 18% of total sales. The top 3 products control more than 40% of sales. Sometimes I graph the transitions too:
If you are thinking this graph looks like a “long tail” graph you are correct. Highly recommend reading Anderson’s Long Tail book.
And that’s one of the reasons I like to visualize the transition. You can learn a lot from the smoothness and/or steepness of the curve. We know, in this example, the transitions graph is going to be steep since the top three brands control more than 40% of total sales.
Over time, we use transitions graphs to find trends that may, with a little nurturing and support, move something up into the head. Doubling down on winners is one of the many lessons I learned from my Direct Marketing bosses during a seven-year Director of Ecommerce tenure.
One of the most interesting mathematical truths discovered analyzing web metrics was the fractal nature of the 80/20 Rule. We found the Pareto Distribution (20% of products = 80% of sales or 20% of links get 80% of clicks) over and over. We came to think of this phenomenon as a fractal because it was constant and ever-present.
Finding your 80/20 points is important and running totals is an easy way to find your breaks. Since it’s bad practice to add %s divide the running total into the total to create a “running %” like this:
Many websites may have thousands of products from #14 to #5,000 with tiny % of totals from a sale here or a sale there. When sales are spread more thinly than this example 80%/20% splits may take many more products. Once you have these values SAVE THEM so you can compare and see trends. Watch the long tail and make sure your time is allocated to your best sellers OR products trending fast. Double down on winners and your return will be greater than trying to shore up and fix laggards.