Better web performance can earn you a generous payback from your customers.
How much? One big study demonstrated 7% better conversion for every second saved on load time.
The true answer is more complicated. Some pages are almost numb to speed effects. Others are so sensitive, a single second can add 50% more profit. On the same website.
There is no single answer because every site, product, and set of customers is different. Speed’s importance even differs between pages. Heck, it changes from month-to-month. The good news is Google Analytics can help.
Google Analytics has serious mojo
That unassuming little GA tag your IT guys dropped on the website is more powerful than it appears. Most companies use Google Analytics for building simple reports that are ignored by decision makers. But the genie in this lamp can handle more than scut work. GA delivers two big jobs for us:
- Real User Measurement Until recently, the industry could only measure time from the server side. “How long does it take to build and send a web page?”. But this only caught a tiny part of the real question, “How long for my users to see my pages?”. So we moved to “synthetic testing” and began simulating the user’s viewpoint from other servers.RUM is the gold standard. It checks the user’s browser to catch the complete answer, including files from other servers and transmit times. Google Analytics checks a sample of your customers with RUM and stores the answers on Google servers.GA’s biggest RUM flaw is the sampling rate. By default, GA only checks 1% of page loads. However, you can sometimes boost sampling with _gaq.push([‘_
- Conversion and spend Here’s where GA really shines. “How much does this customer spend after viewing that page?”
Consider how much work Google is performing to answer this simple question. GA helps you assign “spend” values to important goals (engagement, white paper downloads, etc) or link directly to your e-commerce system. Google then tracks each of your customer visits and links spend to each of the prior pages.
These features allow you to configure a continuously occurring “natural” experiment, instead of a “controlled” experiment with randomly assigned tests. While natural experiments are more difficult to analyze correctly, controlled experiments require you to deliberately slow down speed for some customers. You’d need to serve many slow pages just to prove the need for faster web. Worse, you’d need to rerun the experiment every time the site changes (ad campaigns, demographics, user interface, etc.).
Using statistical analysis to understand the GA data
Useful raw data is only the starting point. Even if the data looks clean, it can mislead.
Remember the truism, “Correlation is not causation”? Some seemingly obvious relationships are not what you’d think. So make sure you’re working the statistical basics when you build speed-profit formulas:
- Filter bad data
- Run cross-sectional analysis to isolate speed effects from other changes that affect spend on your site
- Handle confounding factors that distort the speed / value relationship
- Build comprehensive regressions that work across the entire site
- Check t-statistics and other measures to avoid statistically insignificant predictors
Building useful reports from statistical outputs
The final step is weaving sensitivity formulas together with dollar volume. Why bother with a speed-sensitive page if it doesn’t get traffic? In the end, you need to know two things about speed profits on your site: “How much?” and “Where?”.
Knowledge brings profits. And Google Analytics delivers the information you need.
So to review:
- Sign up and enable Google Analytics
- Create goals, tie to ecommerce, and capture all of your conversions
- Transform your data into useful reports. This is where new initiates can struggle and even seasoned GA veterans can use some help.
(originally posted on yottaa.com)