What it is
Choice architecture is the design of how choices are presented. A nudge is a change to that design that steers decisions without removing options or changing incentives much. The tools include (Johnson et al., 2012):
- Defaults: what’s selected if you do nothing
- Number and order of options
- Grouping and partitioning: how options are split into categories
- Attribute translation: how numbers are expressed (per month vs per year)
- Effort: how many steps each choice takes
Every page has choice architecture. The only question is whether it’s designed with care.
How strong is the evidence?
Contested. A meta-analysis of more than 200 studies found an average small-to-medium effect, with decision-structure changes (like defaults) working better than information-based nudges (Mertens et al., 2022). A re-analysis of the same data argued that, after adjusting for publication bias, there’s little evidence for an average effect, though some kinds of nudge may still work (Maier et al., 2022).
For CRO, the lesson is practical: don’t expect a nudge to work because a study says nudges work. Test it on your audience.
Examples
Ecommerce
- Helpful: grouping delivery options by speed, with the cheapest and fastest clearly labelled.
- Helpful: showing the per-unit price so pack sizes are easy to compare.
- Manipulative: making the “no insurance” option a tiny grey link.
SaaS
- Helpful: a recommended plan for a stated use case, with the others equally visible.
- Helpful: a short “which plan fits you?” question before the pricing table.
- Manipulative: cancellation hidden five steps deep while sign-up takes one click.
Ethics
Medium risk. The same tools that help people choose can push them into choices they wouldn’t make. The UK CMA describes how online choice architecture can harm consumers, including through defaults, drip pricing and false urgency (CMA, 2022). Design for the customer’s interest; make every option easy to choose and easy to undo.
How to test it
- Hypothesis: “Because recordings show visitors switching between plans repeatedly, we believe adding a one-question plan finder above the pricing table will increase paid sign-ups, because it reduces the effort of comparing.”
- Primary metric: completed sign-ups or orders.
- Guardrails: downgrades, refunds, cancellations and support tickets, which show whether people were helped or pushed.
Related principles: the default effect, framing and choice overload.