One of the most valuable metrics when looking at a sites performance is the Time to Interactive metric (TTI) . TTI is similar to the DOM Content Loaded metric - but historically is more dependable. At Edgemesh we find TTI to be consistently reliable proxy to key business drivers such as Bounce rate and Conversion rate.
TTI is basically, how long does it take until enough of the website has loaded to be useful?
TTI and Bounce Rate
Below is an example report showing the relationship between TTI and Bounce Rate based on 1 month of real user interactions.
But Time to Interactive really shines when looking at the impact on sales. The relationship between TTI and Conversion rate is so strong, we've actually built a tool to model it. With 200+ reporting into Edgemesh the impact of TTI and Conversion rate we've found this model to be accurate within +/-15%.
An interactive version is available at https://tti.edgemesh.com .
Modeling Conversion Rate Impact
To begin, simply enter your current Time to Interactive. If you don't have Edgemesh or a similar RUM tool to provide this, a synthetic estimate will do. You can get a guesstimate using Page Speed Insights from Google. The Time to Interactive is shown under "Lab Results".
Once you have an estimated TTI, head over to the Edgemesh tool and plug it in. If you're evaluating Edgemesh, it's a safe assumption to model a 30% reduction in TTI.
From there, Edgemesh will fit these inputs to a model based on real customer performance experiences. A summary of the impact is shown below:
If your time to interactive was 5.5 seconds and you decreased your time to interactive to 3.85 seconds, then your bounce rate would drop ~10% and your conversion rates would improve ~57% .
Next to the model there are graphs displaying other estimates for different TTI values. It's worth nothing that as TTI drops below 3.5s, the impact gets a bump. It turns out, 3.5 seconds is a "line in the sand" for many users where anything above this value results in high frustration. So get those interactive times below 3.5 seconds :)
