Website optimization is all about the numbers. Whether improving sign-up rates or form completions, increasing average order value, or achieving faster page load times, every A/B test is trying to maximize (or minimize) at least one specific behavior about a user, a visit, or page performance. These are metrics we call key-performance indicators (KPIs). With each A/B test using SiteSpect, we almost always include multiple metrics in addition to the basic KPIs. These other metrics help to support interpretation of the A/B test hypothesis, and also sometimes reveal unexpected results or consequences of the A/B test. While SiteSpect has many default metrics that streamline A/B test setup, our product also offers the ability to create custom variables, to include custom calculations.
A calculation is a combination of metrics that uses standard SQL operators to join its components. These combinations result in several different types of calculations. Some common types rely on mathematical operators:
- Counts – discrete tallies expressed as integers (…,-1, 0, 1, 2, 3,..)
- Sums/differences – addition or subtraction between two numerical metrics
- Rates – summarize the relationship between two metrics; often an average
- Proportions – a ratio of a part to the whole, usually expressed as a percentage
- Statistics – arithmetic or geometric mean, variance, skewness, or kurtosis
- Duration – the amount of time between two events
In addition to calculations that rely on arithmetic operators, SiteSpect offers the ability to build calculations with logical operators. Some logical operators we use are:
- AND – returns a value of 1 if both operands are satisfied
- OR – like AND, but returns 1 if either operand is satisfied
- NOT – returns 1 if the operand is evaluated as 0, 1 if nonzero, and NULL if the operand is NULL
- IF – returns a parameterized value then or else
- BETWEEN – returns 1 or 0 if a value is within a parameterized range
In the use of either mathematical or logical operators, (or more complex calculations that involve both), SiteSpect makes it easy to point to components such as other calculated or custom metrics, visit number, user agent, URLs, pageviews, visit length, bounces, or many other values. These calculations can then be added to campaigns to automatically compute and populate at the time an analytics report is requested. By using custom calculations to monitor many different aspects of a campaign, a more insightful interpretation of the A/B test is yielded. For customers of SiteSpect, you can develop these calculations on your own, or you can work with your optimization consultant to build calculations, apply them to A/B tests, and interpret the results.
To learn more about SiteSpect, visit our website.
About Tim Hudson
Tim Hudson is an optimization consultant with SiteSpect. Hudson is responsible for ensuring that customers receive the business intelligence and insights they're seeking by successfully leading them through the processes of optimization strategies and analysis.