A Likert scale is a fixed ordinal rating scale (most often 1–5, 1–7, or 1–10) used to capture a subjective judgment in a consistent numeric form. Named after psychologist Rensis Likert, it is the standard mechanism for converting "how do you feel today" into a number that can be averaged, charted, and compared across weeks and across stack cycles.
Why it matters
Unstructured self-report is close to useless for stack evaluation. "I felt pretty good today" cannot be compared to "I felt really good last week" with any rigor. A Likert rating can. Every subjective proxy in Unfair — energy, focus, mood, sleep quality — depends on a Likert scale captured consistently at the same time of day, which is why Unfair's fast entry paths default to a single tap on a 1–10 row.
Scale length trade-offs
| Scale | Strengths | Weaknesses | Good default for |
|---|---|---|---|
| 1–5 | Fast, forces a clear category, low drift | Coarse; small real changes get rounded away | Beginner loggers, mood daily |
| 1–7 | The psychology-research standard; captures gradient cleanly | Less familiar to health-app users | Research-grade daily tracking |
| 1–10 | Familiar; wide dynamic range | High end compresses (few users ever log 9 or 10) | General-purpose energy or focus |
| 0–100 slider | Feels precise | False precision; noise swamps signal; hard to anchor | Rarely useful |
Unfair defaults to 1–10 because users expect it from other tools, but the research literature supports 1–5 or 1–7 as producing cleaner data, and advanced users are encouraged to switch.
Anchored 1–5 example (energy)
An unanchored scale drifts. A user who rates "energy" as 7 in January may rate the same felt state as 5 in July because the internal reference point shifted. Explicit anchors defend against drift.
| Rating | Anchor |
|---|---|
| 1 | Bed-bound tired. Cannot start anything. |
| 2 | Functional but slow. A single task takes most of the morning. |
| 3 | Average. Work gets done, no unusual lift. |
| 4 | Clearly above my norm. Two work blocks without a crash. |
| 5 | Top few days of the quarter. Would notice it even without tracking. |
Re-anchor every 4–8 weeks. Drift is real, and explicit anchors are the only reliable correction.
Common pitfalls
- Ceiling effects. Users treat "8" as the maximum on a 1–10 scale and never use 9 or 10, compressing the usable range. Explicit anchors help.
- Central tendency bias. Users pull toward the middle to avoid extreme ratings. A shorter scale helps, as does removing a mid-point option (forced-choice 1–4 or 1–6).
- Halo effects. A good mood raises energy and focus ratings in lockstep, which obscures whether any single compound moved any single outcome. Separating the prompts visually helps, and annotating with tag-based logging captures the shared cause.
- Time-of-day drift. A score logged after coffee is not comparable to a score logged before. Fix the logging time and the daily check-in timer holds you to it.
A numeric example
Fourteen-day ashwagandha pilot on a 1–10 scale:
- Baseline 14-day mean (pre-stack) = 5.9.
- Intervention 14-day mean = 6.7.
- Mean delta = +0.8.
- Standard deviation of daily scores = 1.4.
A 0.8-point lift against a 1.4 SD is a small effect (Cohen's d ≈ 0.57) — enough to keep testing, not enough to declare the compound a winner. On a 1–5 scale the same shift would have registered as roughly +0.4 out of 4 and pointed to the same decision. Either scale works; the logging discipline is what matters.
How this appears in Unfair
Every daily check-in metric uses a 1–10 Likert scale with user-configured high and low anchors captured during onboarding. The app prompts a scale review every 60 days and surfaces the anchors as a tooltip on the logging screen. All rolling averages inherit this scale, so ratings remain comparable across a stack cycle and the ranked output does not chase noise.
Clinical safety note
A Likert mood score sitting persistently at the low anchor is a signal for clinical support, not a stack-tuning move. Unfair's check-in flags those cases, surfaces crisis resources, and pauses stack optimization nudges rather than continuing to recommend changes.