Experiment - how much live money can bring website acceleration.
Given: an experimental online store, quite successful in its field, more than a thousand visitors daily.
Task: find out how the optimization of page loading will affect the revenue.
An invisible javascript counter was installed on the site, which recorded the time until the window.onload event occurred and some other parameters, statistics were collected within 5 working days and its analyzer was run in. After that, over the weekend, the following work was done to optimize the site:
- the total amount of graphics and styles was reduced (from 300 to 200 kilobytes for the main page) with virtually no loss of quality.
- the load order of external counters has been changed, their division into two pictures has been removed.
- affixed the correct caching headers for statics
- included compression for styles and javascript
Most of all the hassle of course was with the graphics, some background pictures had to be redrawn. We didn’t manage to do everything as we would like, since there was no “good” command for serious processing of website templates and background images.
So, after optimization, the counter began to write to a new log, and it became possible to compare what was “before” and “after”. First of all, I was interested in the percentage of views that “fit” in 2 and 4 seconds, and separately, to visit the site for the first time, and to view subsequent pages.
First pages: before optimization in 2s. 30% were met, after - 50%. In 4s. Filed 57% and 65% respectively.
Repeated views: Two seconds: before - 58%, after - 80%. 4 seconds: before - 79%, after - 88%.
And now for the fun part. If we take the number of orders received (including phone calls) and calculate the return on the site, taking into account the difference in attendance, we get an increase of at least 10%.
Yes, it was not revenue as such that was taken into account, namely, calls and attempts to place an order. The financial return depends too much on the quality of the work of managers so that the difference could be estimated “in rubles”.
In general, quite a lot of interesting statistics have accumulated, a little later I will analyze and post either a separate post or here.
Task: find out how the optimization of page loading will affect the revenue.
An invisible javascript counter was installed on the site, which recorded the time until the window.onload event occurred and some other parameters, statistics were collected within 5 working days and its analyzer was run in. After that, over the weekend, the following work was done to optimize the site:
- the total amount of graphics and styles was reduced (from 300 to 200 kilobytes for the main page) with virtually no loss of quality.
- the load order of external counters has been changed, their division into two pictures has been removed.
- affixed the correct caching headers for statics
- included compression for styles and javascript
Most of all the hassle of course was with the graphics, some background pictures had to be redrawn. We didn’t manage to do everything as we would like, since there was no “good” command for serious processing of website templates and background images.
So, after optimization, the counter began to write to a new log, and it became possible to compare what was “before” and “after”. First of all, I was interested in the percentage of views that “fit” in 2 and 4 seconds, and separately, to visit the site for the first time, and to view subsequent pages.
First pages: before optimization in 2s. 30% were met, after - 50%. In 4s. Filed 57% and 65% respectively.
Repeated views: Two seconds: before - 58%, after - 80%. 4 seconds: before - 79%, after - 88%.
And now for the fun part. If we take the number of orders received (including phone calls) and calculate the return on the site, taking into account the difference in attendance, we get an increase of at least 10%.
Yes, it was not revenue as such that was taken into account, namely, calls and attempts to place an order. The financial return depends too much on the quality of the work of managers so that the difference could be estimated “in rubles”.
In general, quite a lot of interesting statistics have accumulated, a little later I will analyze and post either a separate post or here.