"Users loved it" isn't evidence. Neither is "the new design looks much cleaner". If you can't show what changed for users and for the business, design gets treated as decoration, and the next budget conversation goes badly.
Many designers avoid metrics because they feel like an analyst's job, or because the numbers aren't always available. But measurement isn't about becoming a data scientist. It's about deciding, before you design, what success looks like, and checking whether you got there.
My approach is "research first, pixels second", and that includes testing and refining until the data, not opinion, shows a design works. Here's how I think about UX metrics, which ones I use, and how to present them honestly.
Decide what success means before you design
Metrics chosen after launch tend to be the ones that look good. Choose them at the start, tied to the problem you're solving.
A simple way to do this:
- Name the problem. For Demerara Bank, customers couldn't bank digitally without hassle; routine tasks meant a trip to a branch.
- Name the behaviour you want. Customers complete routine tasks in the app without help.
- Pick metrics that show that behaviour. Task completion, support queries, and trust measured through research.
If you start with research, the metrics often suggest themselves. In Demerara's research, the biggest frustrations were going to a branch for routine tasks (34%), separate logins for accounts and cards (28%), the nearest branch being too far (22%) and untracked applications (16%). Each of those maps to something you can measure later.
Here's how I'd turn those findings into things to track:
| Research finding | Behaviour you want | Metric to track |
|---|---|---|
| Branch trips for routine tasks (34%) | Routine tasks done in the app | Task completion rate for those tasks |
| Separate logins for accounts and cards (28%) | One login for everything | Login errors and password reset requests |
| Nearest branch too far (22%) | Self-service instead of travel | Use of service requests and appointment booking |
| Untracked applications (16%) | People check status themselves | "Where is my application?" support queries |
The metrics I reach for
Behavioural metrics (what people do)
- Task completion rate: the share of people who finish a key task. The most direct measure of usability.
- Time on task: how long a task takes. Useful for routine flows; less useful where people want to linger, such as reading content.
- Error rate: wrong taps, failed form submissions, back-and-forth navigation.
- Drop-off by step: where people abandon a funnel. This points you to the exact screen to fix.
- Retention and repeat use: do people come back? Often the truest signal of value.
Attitudinal metrics (what people say)
- System Usability Scale (SUS): a standard ten-question questionnaire that gives a usability score you can compare over time.
- Satisfaction or confidence ratings: a single question after a task, such as "How confident are you that this worked?"
- App store ratings and reviews: public, unprompted and very honest.
- Interview themes: qualitative, but powerful when you track how often a theme appears.
Business and operational metrics
- Support queries and tickets: good UX reduces "how do I...?" requests.
- Conversion: sign-ups, bookings, purchases.
- Adoption: how many users try a new feature.
Behaviour tells you what happened; attitude tells you why. Use both.
Watch out
Faster isn't always better. A shorter time on task can mean people gave up or skimmed past something important, so always read it next to completion and error rate.
Frameworks that keep you honest
Google's HEART framework is a useful structure: Happiness, Engagement, Adoption, Retention and Task success. For each category you define a goal, the signals that show progress, and the metrics you'll track. You don't need all five every time; pick the ones that fit your problem.
For smaller projects, I use a lighter version:
- One north-star outcome (for example, "routine banking without a branch visit").
- Two or three supporting metrics (task completion, support queries, confidence).
- One guardrail metric that must not get worse (for example, error rate on transfers).
Tip
Write the north-star outcome, supporting metrics and guardrail on the first page of your design file. When a stakeholder asks for a change late in the project, check it against that page.
Measuring when you don't have analytics yet
Early products and redesigns often have no dashboards. You can still measure.
- Usability tests with tasks. Give five to eight people the same tasks and record completion, time and errors. Compare old and new designs.
- Prototype tests. I often deploy live prototypes on Vercel so stakeholders and testers can try real flows and give feedback faster.
- Design targets. Set a measurable goal in the design itself. On Pukaar.ai, the key 2 a.m. flow had a clear target: from cry to answer in under 90 seconds. Every screen decision was judged against that.
- Pre-launch community signals. Before launch, Pukaar.ai had 400+ community parents, 50+ queries a day, 12+ daily active users and a 154+ waitlist. Those numbers showed demand and gave a baseline.
Good interviews feed this too. My guide to running user interviews that don't lead the witness covers how to get honest answers you can count.
Real numbers from my projects
Here's how this looks in practice, with the numbers I can stand behind.
Demerara Bank (reported impact after the redesign):
- 60% increase in task completion
- 35% reduction in support queries
- 90% improvement in user trust and confidence
These connect directly to the goals set at the start: simplify the routine, reduce cognitive load and build visible trust.
Note
These are reported impact figures, so that's how I label them in the case study. Being clear about the source is what lets a reader trust the number.
Pukaar.ai (after launch):
- Live on Google Play since March 2026 and the App Store since April 2026
- 50K+ downloads on Google Play
- 4.3 stars from 170 ratings
- Reviews that echo the design goals, such as "less anxiety since I started using"
That last point matters. When parents describe less anxiety, and reducing anxiety was the whole point of the 2 a.m. design, qualitative feedback becomes evidence.
Presenting metrics without overselling
How you report numbers affects your credibility as much as the numbers themselves.
- Say where numbers came from. Client-reported, measured in testing, or from app store data. Don't blur them together.
- Show the baseline. "60% increase" means more when people know what it increased from, if you have that.
- Don't claim sole credit. Products ship because of whole teams. Pukaar.ai's team included the CEO, CPO, CTO, Head of AI and 10+ paediatricians.
- Pair numbers with a story. One quote next to one metric is more memorable than a wall of percentages.
- Admit what you'd change. For Pukaar.ai, I'd show the top two cry reasons with confidence, start listening immediately with a live waveform, and allow a guest trial. Honest next steps signal maturity.
If you're putting results into a portfolio, my guide on writing UX case studies recruiters actually read shows where metrics belong.
Work with us
Want a team to design it with you? Our UI/UX design agency for startups and SaaS takes products from research to developer handoff.
Key takeaways
- Define success before you design, tied directly to the problem and research findings.
- Combine behavioural, attitudinal and business metrics; each tells part of the story.
- Use HEART or a light north-star plus guardrail setup to stay focused.
- No analytics yet? Measure through task-based tests, prototypes and design targets.
- Report numbers honestly: source, baseline, team credit and what you'd improve.
See the results in context
Both case studies show how metrics connect back to research and design decisions. Read the Demerara Bank case study for the trust and task-completion story, or the Pukaar.ai case study for a product measured from day one. If you want help defining UX metrics for your own product, email me.
Need budget before you can measure anything? Start with my guide on how to build a business case for UX design, with a simple ROI model you can adapt.