UX OPTIMISATION FRAMEWORK
After enough projects, the surface stops mattering. A checkout, a care service, a foldable phone — different constraints, same three questions underneath.
This is the lens I bring before I design anything.
01 · WHY THIS EXISTS
I didn't set out to build a framework. I just kept noticing the same thing across retail, social care and consumer electronics — the same interface failing different people for completely different reasons.
After the third project where "add trust signals" turned out to answer a completely different question, it stopped being a coincidence. This is just the write-up of what I was already doing.
The pattern, three times
Three industries, three problems — one diagnostic path.
Key takeaway
The same interface fails different people for different reasons. Until you know which reason, you're guessing at the fix.
In practiceAt CeX the barrier was condition uncertainty. At GoToHealthcare it was institutional trust. Same symptom — people leaving — completely different causes.
FRAMEWORK AT A GLANCE
02 · THE FRAMEWORK IN BRIEF
This isn't a replacement for established UX practice — the methods are the same ones everyone uses. The difference is what the process is organised around, and what it is required to produce.
03 · UNDERSTANDING PEOPLE
Not who they are — what they're doing right now. The same person is patient browsing and impatient checking out.
This reframed the whole WiseLaundry concept →
Before any design decision, identify the dominant behavioural mode. The same user can shift between modes depending on context, familiarity, and stakes — the interface needs to serve the mode, not just the person.
Goal-driven
→ Speed & completion
Knows what they want. Friction is the enemy. Optimise for the shortest path to completion — reduce steps, surface CTAs early, eliminate decision load.
Hesitant
→ Trust & reassurance
Ready to act but held back by uncertainty. Needs visible trust signals — warranties, reviews, return policies — surfaced at the point of decision, not buried.
Explorer
→ Comparison & evaluation
Comparing options and building a shortlist. Wants difference made obvious — not more information, better contrast.
Habitual
→ Efficiency & shortcuts
Has done this before and expects it to work the same way. Novelty is friction; recognition is speed.
Price-sensitive
→ Value focus
Value is the deciding variable, but value is not the same as cheap. Needs the price justified, not just displayed.
Key takeaway
Start with the mode, not the persona. The same person behaves differently depending on what's at stake.
In practiceWiseLaundry: the user isn't a "laundry enthusiast" — they're a hesitant user facing an interface that won't explain itself. That single reframe made error prevention the product, not a feature.
04 · UNDERSTANDING CONTEXT
Every category teaches its users what "trustworthy" looks like. Ignore that and your reassurance lands on deaf ears.
In social care it was accreditation — see GoToHealthcare →
Behavioural modes don't exist in a vacuum — industry context shapes expectations, risk tolerance, and the cognitive load a user arrives with. The same hesitant behaviour looks different in fintech than it does in e-commerce.
E-commerce
→ Comparison + impulse
Users oscillate between evaluation and impulse. Trust, value communication, and checkout friction are the primary conversion levers.
SaaS / B2B
→ Multi-step decisions
Long consideration cycles with multiple stakeholders. Onboarding friction, time-to-value, and feature adoption are the core UX problems.
Fintech
→ Trust + risk sensitivity
Regulatory weight and financial exposure raise the cost of an error. Users read carefully and abandon quickly.
Media
→ Attention loops
Attention is the currency and it is contested continuously. Sessions are short, repeat visits are the goal.
Healthcare
→ Reassurance-led UX
Users often arrive under stress, sometimes on behalf of someone else. Reassurance carries more weight than persuasion.
Key takeaway
Industry decides which signals carry weight. The same reassurance lands differently depending on what the category has trained people to expect.
In practiceGoToHealthcare: OFSTED registration did more work than any persuasive copy could — because in social care, accreditation is the argument.
05 · TURNING INSIGHT INTO DECISIONS
All three are real studies on this site. The framework produced the hypothesis. The case study shows what happened next.
High cart abandonment at payment step
HYPOTHESIS
"Surfacing payment options and a security badge directly above the CTA will reduce drop-off by addressing the trust gap at the highest-intent moment."
Users with clear intent leaving before making contact
HYPOTHESIS
"Surfacing accreditation above the fold and rewriting service copy in plain language will reduce drop-off by resolving the trust question before the point of contact."
Persuasive content sitting below where users stop scrolling
HYPOTHESIS
"Reducing hero height so the next section breaks the fold will increase the proportion of users reaching 50% scroll depth — testing whether the constraint is visibility rather than content."
Key takeaway
When you know the mode, the context and the friction point, the hypothesis is no longer a creative act — it's the obvious next sentence.
In practiceSamsung: explorer mode + high-consideration category + a false bottom above the fold produced four isolated experiments, each with its own win criterion.
06 · THE WORKFLOW
Once behaviour, context and friction are understood, the hypothesis almost writes itself. Each step leaves an artefact behind.
Determine which behavioural mode the user is operating in before any design decision is made.
session recordings
user interviews
surveys
ProducesBehaviour-mode map, with the evidence behind each classification
Layer in the industry environment to understand which friction patterns and signals carry the most weight.
competitor analysis
benchmarking
UX audit
ProducesCompetitive benchmark and heuristic audit scorecard
Locate the exact points in the product journey where behaviour type meets industry context and creates drop-off.
heatmaps
funnel analysis
journey mapping
ProducesAnnotated journey map, friction points ranked by impact
Write a specific, testable statement that predicts what will happen if a defined friction point is addressed.
affinity mapping
opportunity definition
hypothesis framing
ProducesHypothesis brief — If / Then / Because, with win criteria
Test the hypothesis using quantitative and qualitative methods. Every recommendation arrives with evidence already attached.
A/B testing
analytics
VOC
ProducesTest result and decision record — scale, iterate or kill
Key takeaway
Every stage produces an artefact, not just a conversation. If a step doesn't leave something behind, it hasn't finished.
In practiceGoToHealthcare: the audit produced a scorecard, the testing produced a findings map, and the second round produced a decision record — ↓11% drop-off, measured.
07 · WHERE AI FITS
I integrate AI tools into my research workflow not to replace judgement, but to accelerate the mechanical parts — so more time goes on interpretation, synthesis, and decision-making.
GOTOHEALTHCARE · IN PRACTICE
"During qualitative synthesis at GoToHealthcare, I used LLM tools to identify patterns across interview transcripts in hours rather than days — then stress-tested those patterns against the data myself. The hypotheses arrived faster; the rigour stayed the same."
Synthesis acceleration
LLM tools process large volumes of qualitative data — interview transcripts, survey responses, usability notes — surfacing themes and candidate patterns for researcher review and validation.
Hypothesis generation
Using AI to generate candidate hypotheses from raw insight data — then evaluating, pruning, and pressure-testing each one against behavioural evidence before any experiment is designed.
Where I don't use it
AI surfaces patterns. It can't tell you which ones matter.
Every candidate finding gets checked against the raw data before it earns the word — and I've thrown some away.
BEFORE YOU ASK
No framework removes uncertainty — it reduces it. Sometimes the research disproves the hypothesis it started from. That's not a failed project; that's the framework doing its job before the build cost was spent.
08 · WHERE IT NEEDS ADAPTING
A framework that works everywhere usually isn't doing much. These are the conditions where I'd adapt it, or not reach for it at all.
It assumes you can observe behaviour
In low-traffic products, analytics are too thin to identify a dominant mode and qualitative work has to carry the whole diagnosis. The framework still applies — it just runs slower and with wider error bars.
It assumes one dominant behaviour mode
That breaks down in multi-stakeholder purchases where the researcher, the budget holder and the end user are three different people with three different modes. There, Layer 1 needs running once per role.
It optimises within an existing product
It will tell you why a flow is failing. It will not tell you whether the product should exist — that's a discovery and strategy question, and this is the wrong tool for it.
HOW I THINK
Analytics tells me where to look. Research tells me what I'm seeing.
You getA prioritised opportunity backlog
If I can't say what I'm testing and how I'll know it worked, it isn't ready to build.
You getA tested hypothesis with a win criterion
Success is behaviour change and business impact — not deliverables shipped.
You getA measured result and a decision
THE FORMULA
Every case study in this portfolio runs the same system. Different industries, different users, different constraints — the framework stays the same because the questions do.
Chapter five
None of this came from a textbook. It came from years of watching how people behave in products, communities and games — long before any of it had a name.