A hypothesis in conversion rate optimisation is a specific assumption about what a change will cause and why. A usable hypothesis contains three parts: an observation backed by data or research, the proposed change, and the expected effect together with the metric on which it will show up. For example, session recordings show that visitors look for shipping information only at the cart stage, so it is added to the product page, and a reduction in cart abandonment is expected. A hypothesis formulated this way can be evaluated even when the change does not work, because it is clear what was being tested. Hypotheses are collected into a backlog and ranked by expected benefit and implementation effort – it is precisely this ranking that determines whether the optimisation delivers results or just minor tweaks.
See also: Website A/B Testing, Friction in the conversion process, On-site Survey.
