Is Your Mobile Site Fast Enough? How to Run a Regression Analysis to Find Out

Shares

So what’s the big deal about mobile? Like almost anything that gives us pleasure, it’s the instant gratification aspect of it that gets us. With just one click we can order movie tickets, check the weather, check our email, our Facebook… you get the point.

But what’s the one thing that really irks us when we’re trying to get stuff done on our phone? Let’s all say it together:

“WHEN THE PAGE TAKES TOO LONG TO LOAD!”

If your page takes a long time to load, your customers will get frustrated and impatient and change sites without even thinking twice. Ooops! There goes any potential revenue! Think we’re exaggerating? Think again.

We decided to see how strong the relationship (and if there even was) was between the time it takes a mobile site to load and the revenue from the corresponding e-commerce site.

You may be saying to yourself, “yeah, yeah, I get it. But HOW can I find these relationships without just making blank assumptions?”

The answer is simple. A regression analysis.

Let’s start at the basics. What in the world IS a regression analysis?

A regression analysis is used in statistics to figure out if there is a relationship, or correlation, between variables.

Even though today we will be focusing on finding a relationship between mobile e-commerce sales and the loading time of a mobile site, you can use regression analysis to measure the relationship of almost anything all you need is one independent variable (the time it takes the page to load), a dependent variable (e-commerce sales), and a decent sample size.

Note- the dependent variable is named this because its value completely depends on the value and behavior of the independent variable.[1]

To determine if there is a relationship between the variables, there are a few things you’ve got to look out for:

1)      The best-fitting line

2)      Correlation coefficient (r)

Best-fitting line

There are three types of lines you need to know about in order to understand the trend of your data.

Linear: The linear regression consists of a straight line through the data points. In the figure below, the black line is the regression line and shows the predicted scores of Y for every possible value for X. The vertical lines going from the regression line to the points is called the error of prediction. When the distance between the data point and the line are small, that means the error of prediction is small. If your data follows a linear line, it means you are growing at a steady pace.

figure 1

Exponential: The exponential model shows you if your data is growing at an exponential rate. This means that as the value of x increases, so does the value of f(x), making it an increasing function.

figure 2

Logarithmic: The logarithmic line is a curve that plateaus over time. You are increasing at a slower and slower rate until you stop completely and reach a plateau where you wouldn’t expect to grow anymore. This function is the opposite of the exponential function, although it is still an increasing function. However, as the x increases, the increase values get slower and slower.

figure 3

 

Correlation Coefficient (R)

The correlation coefficient, or R, is used to measure the strength of the linear relationship between two variables. The scale ranges from -1 to 1, with -1 being a perfect negative relationship and 1 being a perfect positive relationship. The closer the number is to 0, the less likely there is a strong relationship between the two variables.

Finding out if there is a correlation between time it takes a mobile site to load and e-commerce sales

In order to find out if there is a relationship between the two variables, you need to run a regression analysis with the aforementioned lines and then compare their R values. The graph with the R closest to 1 will be your best fit. If you find that none of them are close to 1, then maybe there is no relationship. You can go and gather more data and run it again, but sometimes you can’t show what’s not there!

Alright, let’s begin.

Here’s how you can run a regression analysis on Excel

Step 1: Get your data in Excel

We collected the data of the top mobile e-commerce companies and their loading page time and simply put the values in our excel sheet.[2]

figure 4

 

Step 2: Highlight the data and create a scatter plot

Simply to go insert, and choose the scatter plot from the graphs!

 

figure 6

Step 3: Insert a line

On the right side of the graph there is a plus sign where you can add chart elements. Click on it and choose trendline. Because there is a seemingly linear relationship for this specific data, the linear line is automatically placed as the trend line.

figure 7

 

Step 4: Testing the other best-fitting lines

In order to test the exponential and logarithmic lines, simply click on the plus sign again and scroll down to the trendline. There should be an arrow pointing to additional options. Go ahead and follow that until a format trendline option pops up on the right side of the excel sheet.

From there, you can choose what type of trendline you want to show in the graph. Additionally, if you scroll all the way to the bottom, you will see an option for the R value to be posted on the chart. Definitely click this.

Linear:

figure 8

Exponential:

figure 9

Logarithmic:

figure 10

We see that the logarithmic line is the strongest with an R value of .9674, showing an extremely high correlation between the two variable. This means that the rate of change in the data is decreasing quickly and then levels out. We would most likely be able to see this if we provided additional data that focused on the sites that took a long time to load.

What can we conclude?

We see that there is in fact a strong relationship between mobile e-commerce sales and the time it takes a mobile site to load. As the site takes longer to load, the smaller your sales will be!

What about you? What are you trying to figure out?

 

About AdClarity

AdClarity is a Marketing Intelligence tool which provides online marketers with actionable insights about their competitors’ advertising activities. Driven by big data and proprietary behavioral content discovery technology, AdClarity unveils brands’ campaigns, ad creatives, impressions, and spend data across multiple channels, including Display, Mobile Web, Mobile Apps and Video. Data is collected across 20 geographies and covers over 50M URLs daily while discovering over 40K new campaigns every day. The AdClarity product suite is used by over 7,000 media and advertising professionals globally in Fortune 500 Brands, Agencies, Ad Networks, and Publishers.

 

Request a guided tour of AdClarity.

 

[1] http://www.sosmath.com/algebra/logs/log4/log42/log42.html

[2]  The data should be normalized or standardized to bring all of the variables into proportion with one another.

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *