Most designers know the feeling: you’re asked to “measure UX impact,” but the metrics never quite fit. HEART sounds elegant until you apply it to an internal workflow tool and realise that “Happiness” won’t tell you if your colleagues can actually finish their work.
Internal tools live in a different world. They’re not about delight- they’re about efficiency, accuracy, and sanity. The question users silently ask is simple:
That’s where CASTLE comes in, a UX measurement framework designed for the messy, pragmatic reality of internal software.
Can I do my job faster, with fewer mistakes, and without asking for help?
Why internal tools need a different lens
CASTLE stands for Cognitive Load, Aesthetics, Satisfaction, Task Success, Learnability, and Engagement.
It’s a practical blend of human-centred and operational metrics — focused on how people think, learn, and work inside digital tools.
- Cognitive Load → how hard the interface makes users think.
- Aesthetics → visual clarity and polish that support trust and speed.
- Satisfaction → perceived confidence and comfort in getting things done.
- Task Success → completion rate, error frequency, and rework effort.
- Learnability → how quickly a new or infrequent user gets up to speed.
- Engagement → frequency and depth of productive use (not vanity clicks).
Together, these create a language that resonates with both design and operations. CASTLE asks: Is our product reducing friction at every step of the job?
We found Task Success was strong, but Learnability and Cognitive Load lagged
When HEART falls short
HEART is great for consumer experiences — where emotion, discovery, and delight are central. But inside an organisation, “Happiness” often plateaus while inefficiency quietly festers.
I once joined a project where the internal billing team rated the product 4.5/5 in satisfaction surveys — but still needed six clicks to finish a single transaction. They were “happy” because they’d learned to cope.
When we measured with CASTLE, we found Task Success was strong, but Learnability and Cognitive Load lagged. The product looked familiar, but it wasn’t intuitive. Within two releases, we simplified the workflow, clarified the hierarchy, and dropped time-to-competence by nearly 40%.
That’s the CASTLE difference, it catches the friction happiness can hide.
So what do you actually measure?
A minimal viable CASTLE scorecard
The secret to keeping CASTLE useful is focus. Start small: pick one core flow and define a simple scorecard.
| Metric | Method | Healthy Range |
|---|---|---|
| Cognitive Load | Short NASA-TLX after key flows | < 35/100 |
| Task Success | % first-time completion | > 85% |
| Learnability | Time-to-competence (TTC) for a novice | ≤ 2 sessions |
| Satisfaction | Post-task confidence (1–5 scale) | ≥ 4.2 |
| Aesthetics | 3-item semantic differential (e.g., cluttered ↔ clear) | Track movement |
| Engagement | Weekly active users vs. expected cadence |
Consistent over time |
Pro tip: Don’t over-instrument. Choose one primary metric per flow and one secondary that signals improvement.
The goal is a scorecard your product manager can read in 30 seconds — not a dashboard that needs a PhD.
Turning metrics into movement
Here’s a pragmatic rollout sequence that’s worked for me:
-
- Start with one workflow — something measurable with business value (like billing or case management).
- Define success criteria per CASTLE facet with numeric thresholds.
- Run a two-phase test:
- (a) expert internal users
- (b) unfamiliar users for learnability.
- Instrument your “TTC” baseline and track movement across four sprints.
- Publish a monthly one-pager linking CASTLE outcomes to business results (time saved, error reduction, cost avoidance).
Every metric movement becomes a story — a visible proof that design is improving operational performance.
What tends to go wrong
- Metric Salad: Too many KPIs dilute focus. Start with one flow.
- The Happiness Trap: High satisfaction ≠ high efficiency. Balance perception and behaviour.
- Ownership Gap: Assign a named PO and designer to each scorecard. Shared ownership = sustained attention.
A good CASTLE rollout isn’t perfect — it’s visible. People should see metrics evolve and understand why they matter.
Here’s how it played out in real life
During an off-cycle billing project, our first CASTLE pass showed Task Success jumping quickly but Learnability lagging for infrequent users. That led to simple UI changes: clear empty states, progressive hints, and better defaults. The next round of testing saw TTC drop by nearly half.
No rebrand, no overhaul — just focused measurement, iteration, and communication.
CASTLE gave the team confidence that design changes were genuinely helping people work better, not just look better.
The real lesson
Design for internal tools often feels invisible — but that’s exactly why measuring it matters. CASTLE gives teams a shared language for the invisible wins: reduced cognitive effort, faster learning, fewer errors.
And perhaps most importantly, it reminds designers that our impact isn’t measured by how much users love the product, but by how confidently they can use it.
In the end, CASTLE isn’t just a framework.
It’s a mirror, showing how well design serves
the people already holding the business together.
