DCHS Insights CX Human-Centered Design @ DCHS PROTOTYPE no real data Rapid high-volume feedback (kiosks)
What it is
A one- or two-tap rating at the point of exit — smiley terminals or tablets, one fixed question, four buttons, no neutral midpoint. Volume and trend detection, not explanation.
"How is satisfaction trending — by location, day, and hour?"
What it tells you and what it cannot
| Tells you | Cannot tell you |
|---|---|
| Huge response volume, continuously, at trivial per-response cost | Why — ever. A dip is a flag, never a diagnosis |
| When and where satisfaction dips (time/location tagging) | Who responded, or anything about complex needs |
| Whether a change moved the trend | Anything nuanced — one tap is one bit of signal |
How it runs
| Who does the work | Participants | Elapsed time | Cost |
|---|---|---|---|
| Almost none day-to-day; one owner for the dashboard and alerts | Self-forming, hundreds+ per week | Days to deploy; runs continuously | $–$$ — hardware/subscription |
Strengths, watch-outs and quality signals
frictionless; genuinely high volume; pinpoints hotspots by hour and site; anonymity can raise candor.
response bias — the very happy and very angry over-tap, and children mash buttons; no "why" attached to any number; placement bias (a kiosk by the happy exit measures the happy exit).
stable high response counts; dips that correlate with real events; and — the real test — every sustained dip triggering a deeper-inquiry method rather than a guess.
Equity & consent
Minimal consent burden (anonymous single tap), but physical and language accessibility decide who is counted: mounting height and reach for wheelchair users, screen contrast for low vision, and the prompt's language. Some clients fear a negative tap could be seen and held against them — placement away from staff sightlines matters.
When to choose it
as the always-on monitoring layer that tells you where to look.
HappyOrNot's four-smiley convention — deployed across 100+ countries with cumulative responses in the billions — deliberately drops the neutral midpoint to force a lean.
Exit-kiosk feedback — benefits lobby weekly summary
What this can and cannot tell us
- Can tell us: volume, direction, and where/when — one terminal's taps, trended by day and hour.
- Cannot tell us: why any number moved, who tapped, or anything nuanced.
- Everything below is a flag; every "why" routes to "Trigger deeper inquiry."
The question & scale
One temporary HappyOrNot-style terminal at the benefits-lobby exit, separate from the building wayfinding kiosk, Monday–Friday, August 3–7, 2026, asking one question:
"How was your visit today?"
Four smiley buttons — very happy (dark green), happy (light green), unhappy (light red), very unhappy (dark red) — with no neutral midpoint; the even scale forces a lean (real background: HappyOrNot's four-button convention; one billion cumulative presses across 120 countries). A negative tap offers one optional follow-up layer: five reason categories, no open text. Taps are anonymous — no identity, no segments.
Headline metrics
Week of August 3–7: 340 taps, 76% positive, down 3 points on the prior week.
Headline tiles: 340 taps, 76 percent positive, down three points, 8 percent very unhappy.
Weekly trend
Positive share fell every day, 82% to 69% — a 13-point slide. The kiosk does not tell us why.
Positive share falls from 82 percent Monday to 69 percent Friday.
Positive share by day
57 of 81 negative taps chose a reason; "Wait time" leads. Categories, not accounts.
Ranked bars of follow-up reasons: wait time leads with 24 taps.
Hotspots
Flags only; the kiosk sees where and when, never why. All cells:
- Weekday 12:30–1:30 pm — 54% positive (33 of 61 taps) against the 76% weekly average: the lunch hour. The kiosk can see the dip; it cannot see a single reason for it.
- Friday — 69% positive, the lowest day of a five-day slide.
- Before 10:00 am — 84% positive, the highest window — contrast, not explanation.
What the kiosk did not see
One tap, reconstructed as a composite vignette. Marisol Reyes came in on an unpaid lunch break to fix a SNAP renewal flagged after the portal rejected the same document twice. One door, an 18-minute wait, one pay stub. It worked — the benefit continues. 52 minutes on site, out by 1:40 pm; at 1:38 pm she taps light-green.
Persona tile for Marisol Reyes, composite client: one document fixed, three needs unseen.
Marisol Reyes
Works full time. Came to fix one document. Left with needs we would not have seen.
Goal: To not have to think about any of this
The tap was honest. It was also the whole record. Needs surfaced: 1 of 4 — leaving with her: a withdrawn 13-year-old and waitlists, an evening childcare gap, about $380 of electric arrears. Nobody asked past the document mismatch; an open lobby is not where she would volunteer more. She stood in the week's worst-scoring hour and pressed green.
Client quote: Fine. That is the problem.
Fine. That is the problem.
Tapped light-green at 1:38 pm, out by 1:40 pm
Left open, in the kiosk's own terms: of the week's 259 green taps, how many walked out the way she did — presenting task fixed, nothing further asked? The kiosk cannot count that; this report does not guess.
Response bias & equity caveats
- Self-selection. About 610 lobby check-ins vs 340 taps; roughly half of exits, and we do not know how the silent half differs.
- Button-mashing. Repeat-press bursts (a bored child) inflate counts; left in the totals, not guessed out.
- Reach. Adult standing height; the concept should be implemented in English, Spanish, Amharic, Arabic, and Dari/Farsi. Shorter, seated, low-vision, and low-literacy visitors may still be undercounted.
- Perceived exposure. Visible from the counter; anonymous in fact, but a visitor mid-case cannot know that, and a red tap can feel risky.
Trigger deeper inquiry
The kiosk points; these methods ask.
Findings table routing each kiosk flag to exit intercepts or shadowing with an owner and a follow-up metric.
| Finding | Recommendation | Owner | Evidence · metric |
|---|---|---|---|
| Positive share falls to 54% in the weekday 12:30-1:30 pm window | Exit intercepts across two lunch hours | CX research lead | 33 of 61 taps Completed intercepts per lunch hour |
| Positive share slid 13 points from Monday to Friday | Shadow two Friday afternoon sessions, door to exit | Lobby operations lead | 82% Mon to 69% Fri Timed waits and handoffs per session |
| 'Could not finish my task today' ranks second among reasons | Add a fixed intercept question on what preceded the visit | Digital services manager | 13 of 57 reason taps Share reporting a failed portal attempt first |
| A light-green tap can exit with needs never surfaced | Counter shadowing to observe what is asked beyond the presenting task | Benefits program manager | Composite vignette Questions beyond the presenting task per visit |
Limitations & quality self-check
- Trends, not causes. What changed, where, when — never why; no causal claims.
- Anonymous, so no segments. No demographics, no repeat-visitor linkage; dip cells cannot be decomposed.
- Small cells. The lunch-hour cell holds 61 taps; a handful of presses moves it several points.
- Synthetic throughout. All counts, percentages, and timestamps are fabricated; only the four-button convention and vendor volumes are real background, cited above.
- Unknowns preserved, verbatim and unresolved: "Whether she would accept a referral from someone she just met"; "Whether Ethan would go"; "Whether Rosa is someone DCHS has any way of reaching"; "How many people with similar circumstances are seen every week and never asked"; "Whether the portal failure is common or rare". The kiosk answers none of them; neither does this report.
- Continuity check. The vignette matches the persona record: 1 door, 18-minute wait, 52 minutes on site, out by 1:40 pm, needs surfaced 1 of 4, light-green tap at 1:38 pm.