Nuhdge

1300 123 456            CONTACT
What We Do   How We Do It   Who For  

Split Testing

Split testing, otherwise known as A/B Testing, is where you test a change in marketing tactics in order to statistically validate your results (rather than assume). The concept is to perform a controlled experiment where you test a new tactic (treatment condition) against an unchanged environment (control condition) and compare to see if there is any impact to your results (statistical significance).

Examples of split testing

  • Enquiries and sign-ups
  • Time on page and content engagement
  • Landing page redesign
  • Headlines and slogans
  • Banners and Display advertisements
  • Navigation and goal paths
  • Action buttons
  • Product and subscription plan pricing
  • Email open rates and click-throughs
  • Check-out purchases

Where are you at with your split testing?

  • You’ve never performed split-testing before and not sure where to start?
  • You’re frustrated by lack of results in your split test programme?
  • You’re not confident in your split test statistics?

We can assist with setting up or optimising your split testing programmes. Getting value out of split testing depends on having the right measurement in place, and applying correct analysis to the data.

[CONTACT US]

Key to effective split testing

  1. Develop a sound testing plan - frame your ideas for tests into a precise hypothesis and a test programme that will provide the right answers to the questions you’re asking
  2. Define the correct measurements - different experiments require different metrics. What matters for you: conversion, revenue, engagement?
  3. Setup your split testing system - preferably engage a certified consultant who understands the ins and outs of your selected system
  4. Integrate your split testing system with your existing analytics systems - this is often an overlooked item, and is vital to understanding effects outside of your experiment environment
  5. Use proper statistical analysis models and approaches - understand how to extract signal from noise, and how to get useful answers where you don’t have enough data to create ‘statistical analysis’ (yes, it’s possible)