UX Metrics That Matter Beyond Conversion Rate

Measure whether people can complete the tasks that matter (task success), how long and how hard it is for them (time on task, errors, a post-task ease rating), how usable the product feels overall (SUS), and whether the experience holds up in live use (Core Web Vitals, retention or adoption). Conversion rate is the business outcome; these measures tell you whether the experience is the reason it moved. Pick a handful tied to specific questions, measure them before and after a change, and use the same tasks each time.

Conversion rate is affected by traffic mix, pricing, seasonality, campaigns and stock as much as by design. A redesign can improve the experience and still see conversion flat because of a price change, or see conversion rise because of a promotion. UX metrics separate the two. This guide covers which metrics to use and how to tell if a change worked; running the studies themselves is covered in the usability testing guide.

Read across from the question you are trying to answer. The caveat column is the part most guides skip: many metrics are meaningless at the sample sizes small teams can manage, so it tells you what evidence each one needs.

Task success is the simplest and most useful UX metric. NN/g defines success rate as "the percentage of users who were able to complete a task in a study". If you want to give partial credit, use labelled levels (complete success, success with a minor issue, success with a major issue, failure) and report how many people fell into each. Nielsen and Budiu are explicit that turning those levels into numbers and averaging them is wrong, because they are labels on an ordinal scale. There is no universal benchmark for a "good" success rate; compare with your own baseline.

Time on task matters for anything people do often, such as searching a catalogue or processing an order. Time data is skewed by a few slow participants, so MeasuringU recommends the geometric mean for samples under about 25 and the median above that, with confidence intervals. Only compare times for the same task wording and starting point, and decide in advance how to treat failed attempts.

Errors include wrong paths, validation errors and actions that need undoing. In a live product you can count validation errors per form step, or support contacts tagged to a feature. A falling error count after a change is often more persuasive to a team than a satisfaction score.

The Single Ease Question asks, straight after each task, how difficult or easy it was on a 7-point scale. MeasuringU reports that the average across a large set of tasks "hovers between about 5.3 and 5.6". It is quick, which makes it practical in unmoderated tests, and a low score next to a successful task often points to friction people pushed through.

The System Usability Scale is a 10-item questionnaire with five response options, scored from 0 to 100. Jeff Sauro's analysis of about 500 studies puts the average at 68. SUS is useful for comparing versions or competitors over time, but a decent SUS score can hide a task that most people fail, so never report it alone.

Net Promoter Score asks "How likely are you to recommend us to a friend or colleague?" on a 0 to 10 scale. According to Bain, which owns the system, 9 to 10 are promoters, 7 to 8 passives and 0 to 6 detractors, and NPS is the percentage of promoters minus the percentage of detractors. Frederick Reichheld introduced it in Harvard Business Review in December 2003 as "the one number you need to grow".

That growth claim has been challenged. Keiningham and colleagues, writing in the Journal of Marketing in 2007, examined Norwegian customer data over time and did not find the clear superiority over other satisfaction and loyalty measures that had been claimed. MeasuringU's summary of the paper notes that its samples of firms were small, which limits conclusions in both directions.

For UX specifically, NN/g's 2024 article sets out the problems. NPS shows perception but not the reason for it. It is not valid from the small samples in a typical usability study. Grouping scores into three bins throws information away, so a user moving from 2 to 5 is still a detractor. Loyalty is affected by price, lack of alternatives and brand, not just usability. A single global score rarely reflects a specific design change. And it is easy to game, for example by surveying only happy customers. NN/g suggests pairing it with behavioural measures such as task success and with SUS.

Use NPS, if at all, as a slow relationship measure across the whole customer base. Do not use it to judge whether a new onboarding flow or checkout is better.

Frameworks help you pick a few metrics that map to goals, rather than tracking everything. HEART (Happiness, Engagement, Adoption, Retention, Task success) comes from Kerry Rodden, Hilary Hutchinson and Xin Fu at Google, published at CHI 2010. The paper pairs the categories with a process for mapping product goals to metrics. In practice: write down the goal, decide what user behaviour or attitude would show progress towards it, then choose one metric that tracks that signal. You do not need a metric in every category.

CASTLE (Cognitive load, Advanced feature usage, Satisfaction, Task efficiency, Learnability, Errors) is NN/g's framework for products that are compulsory for users, such as workplace and internal tools. NN/g's Page Laubheimer states that it "is not intended to replace or supersede HEART". The reason it exists: if staff must use an ordering portal, high retention tells you nothing about whether it is good.

Speed and stability are part of the experience. Google's Core Web Vitals are Largest Contentful Paint (loading; good is 2.5 seconds or less), Interaction to Next Paint (responsiveness; 200 milliseconds or less) and Cumulative Layout Shift (visual stability; 0.1 or less). web.dev recommends assessing them at the 75th percentile of page loads, separately for mobile and desktop, and says only field data from real users gives the complete picture. They are useful UX metrics because they come from real visits, not a lab.

GA4's engagement rate is the share of engaged sessions, where an engaged session lasts more than 10 seconds, has a key event or has two or more page views. Bounce rate is the inverse. These are cheap to watch but weak evidence of quality: a lost user clicking through three pages counts as engaged. Use engagement rate to find pages that people abandon at once, then investigate with task-based testing.

Most arguments about whether a redesign "worked" come from measuring different things before and after. A before/after measurement plan avoids that:

Hypothetical example: a B2B software company redesigns its account set-up flow. Its questions are: can new admins set up an account unaided, and does it feel easier? Before the redesign it runs five set-up tasks with 20 participants, recording success levels, time per task and SEQ, then a SUS questionnaire. In live data it tracks the share of new accounts completing set-up within seven days and support tickets tagged "set-up".

After launch it repeats the same five tasks with 20 similar participants. If success on the hardest task improves and the confidence intervals barely overlap, that is reasonable evidence. If SUS rises but success does not, the flow looks better but is not easier to complete. If live set-up completion rises in the same month as a sales push that brought in larger customers, the team notes the confounder rather than claiming the credit.

Frameworks: Rodden, Hutchinson and Fu, Measuring the user experience on a large scale (CHI 2010); NN/g, the CASTLE framework (Page Laubheimer, 2023).

Task and survey metrics: NN/g, success rate (Nielsen and Budiu); NN/g, how many test users; MeasuringU, the Single Ease Question; MeasuringU, measuring usability with SUS; MeasuringU, average task times in usability tests.

NPS: Bain, measuring your Net Promoter Score (vendor source; Bain owns NPS); Reichheld, The one number you need to grow (HBR, 2003); Keiningham et al., A longitudinal examination of Net Promoter and firm revenue growth (Journal of Marketing, 2007) (abstract) and MeasuringU's summary; NN/g, NPS and UX (Therese Fessenden, 2024).

Live metrics: web.dev, Web Vitals; Google Analytics Help, engagement rate and bounce rate.

Accessed 15 October 2026. Analytics definitions and thresholds can change; check the current pages.

Conversion rate tells you whether the business outcome moved; UX metrics tell you whether the experience is why. Choose a few metrics from the selection table based on the questions you need answered, measure a baseline with fixed tasks before you change anything, and compare like with like afterwards. Next steps: write down the two or three questions your next design change should answer, pick a metric for each, and schedule the baseline round before design work starts. For finding problems before you measure them, see what a UX audit covers.

Our UI/UX design work can include setting these success measures at the start of a project.