Dashboard — 10 Ideas · Bristlenose infoviz mockups

01 — hero chart

The study's pulse

SUPERSEDED → 02Emotion arrives quantised — sentence to sentence, gear to gear as topics shift. The continuous ridge interpolates feelings between quotes that nobody expressed. The tape (02) shows the same story without fabricating the in-betweens. Kept for reference.

Lineage: Joy Division's "Unknown Pleasures" cover · Wilke's ridgeline / joyplot · seismogram.

Emotional pulse across four interviews

y = quote intensity (1–3) · colour = sentiment · x = session-normalised time

s1
p1 · 4:22

s3
p3 · 4:06

Two of four sessions shown — enough to record why it lost: the smooth curve implies p1 passed through moderate feeling between the satisfaction plateau and the frustration ridge. They didn't. They changed topic.

02 — hero strip

Sentiment tape, with section boundaries

PROMOTED — the heroDiscrete ticks honour quantised emotion. Now with topic boundaries from topic_boundaries.json (real pipeline output, confidence 0.72–0.95).

Lineage: Bloomberg / Reuters trading tape · Wall-Street sentiment tickers · tube-map interchange bands.

Each quote is a tick: colour = sentiment, height = intensity. New in v2: section boundaries show where the participant moved — homepage to checkout, pre-surgery to aftercare. Two alignment treatments below solve the "boundaries vs ticks" overlap worry three ways: labels get their own lane (never in the tick field); hairlines render behind ticks, dashed and dim; narrow zones lose labels before they lose their lines.

Variant A — label lane above · Bristlenose · Project IKEA

s1

Pre-surgery info Aftercare Notifications

▼ aftercare-frustration → △ delight

s2

Homepage Top nav Beds category

▼ beds nav — confidence lost

s3

Homepage Browse PDP Browse Search Login Bag

◇ mixed, resolves positive

s4

Intro Duvets & bedding Beds & sofa beds Checkout

▼ late frustration at checkout

Variant B — topic band below (s3 shown for comparison)

s3

Homepage

Browse

PDP

Browse

Search

Login

Bag

tube-map band — same data

Own lane, never the field Labels live above (A) or below (B) the ticks — the tick field stays pure signal. No collision is possible by construction.

Hairlines behind, dashed, dim A tick landing exactly on a boundary is the normal case — boundaries are derived from speech, so the first quote of a zone sits on its line. Rendered behind the tick it reads as "first word on the new screen", which is true.

Narrow zones: line yes, label no s3's checkout zone is seconds wide. It keeps its hairline, loses its label (hover reveals in product). Lines are cheap; text is not.

Which variant? A for the dashboard hero — the strip stays clean and the label lane doubles as a skim-read of the journey. B for compact contexts (session cards, sidebar previews) — it's the same idiom as the existing coverage bar, and the band survives shrinking better than floating labels do.

03 — comparative panel

Kudos & kvetch — the order book

SHORTLISTCandidate replacement for the featured-quotes row: one kudos, one kvetch, maybe one neutral. Selection logic already exists — pick_featured_quotes() already scores and diversifies by polarity; this is a re-hang of the same picks, not new plumbing.

Lineage: Financial order books (bids left, asks right, spread in the middle). Recast as what worked vs what broke, sorted by intensity.

The single question every stakeholder asks — "is this working or not?" — has a shape. This makes the answer scannable in five seconds, both sides at their real weight.

KUDOS · what worked (11)

spread

KVETCH · what broke (13)

"I really like — you should be able to walk around with some pain, but take care not to lift. That is nice to actually say that."

s1 · pre-surgery notification · delight

±3

"There was no aftercare, the aftercare was quite poor."

s1 · post-surgery · frustration

"It works well for me. It's lovely. It's straightforward."

s1 · pre-surgery info · satisfaction

±2

"I rang the secretary asking to make an appointment, they said no. I went back to A&E about four times."

s1 · post-surgery · frustration

"Beds and mattresses. That's easy. That's in the top navigation."

s2 · top nav · confidence

±1

"Beds & mattresses. Beds… Has this gone slow because it's doing screen recording?"

s2 · beds category · confusion

"179. That's quite cheap."

s3 · beds browsing · surprise

±1

"Delivery. Hmm. Why is it £30? That seems a lot."

s4 · checkout · frustration

04

Ambivalence portraits

PARKEDAmbivalence is everywhere at fine grain — you hate the checkout pricing but love the checkout button, and different users voice each differently. Signal cards are designed to draw those distinctions out; a two-cell portrait over-aggregates what is really per-facet nuance. Theory stays interesting (Scharp 2021); the widget doesn't earn dashboard space.

Lineage: Scharp 2021, Thematic Co-occurrence Analysis (Journal of Communication).

Participant 1

Pre-surgery info vs post-surgery aftercare · s1 · 4:22

Praised (2× satisfaction · 1× delight)

"It works well for me. It's lovely. It's straightforward."

00:12 · pre-surgery info

Criticised (3× frustration)

"There was no aftercare. I went back to A&E about four times."

02:08 · post-surgery aftercare

One portrait kept as the record of why it parked: the split reads clean here only because s1's two topics happen to be far apart. At real grain the split multiplies per facet.

05 — rigorous grid · plus 5b, the compact remix

Thematic co-occurrence matrix

DISCUSS05 answers "where do participants agree?" 06 answers "which sections carry the strongest signal?" — see the comparison after 5b. One matrix on the dashboard seems right; the GitHub-grid remix below is the dashboard-scale candidate.

Lineage: Scharp (2021) — participants × themes matrix, cells encode × (present), + (emphatic exemplar), colour (positive/negative/both). Read along three continua: Sporadic ↔ Pervasive (columns), Unilateral ↔ Bilateral (cells), Unbalanced ↔ Balanced (rows).

participant · session	Pre-surgery info	Post-surgery aftercare	Post-surgery notifications	Top navigation	Beds category	Search results	Login & app	Shopping bag	Duvets browsing	Checkout & delivery
p1 · s1	×⁺	×⁺	×⁺	—	—	—	—	—	—	—
p2 · s2	—	—	—	×	×	—	—	—	—	—
p3 · s3	—	—	—	—	×	×	×	×	—	×
p4 · s4	—	—	—	—	×	—	—	—	×⁺	×

Sporadic ↔ Pervasive

column · how many participants touch this theme

Beds category is pervasive (3 of 4). Post-surgery aftercare is sporadic (1 of 4) — but strong.

Unilateral ↔ Bilateral

cell · one sentiment or both

× Bilateral cells (top nav, beds cat, login) — same person, both feelings.

Unbalanced ↔ Balanced

row · does this participant see mostly one thing?

p1's row is unbalanced by topic. p4's is balanced. p3 is balanced with high variance.

Emphasis marker (⁺) = intensity-3 quote in that cell. Method after Scharp, K. (2021), Journal of Communication 71(4).

5b — the same axes, GitHub-contribution clothes

Pre-surgery

Aftercare

Notifications

Top nav

Beds browsing

Duvets

Product page

Search

Login / app

Bag

Checkout

p1

p2

p3

p4

Rows = participants, columns = sections, cell hue = dominant sentiment, cell depth = quote count. 11 of 16 sections shown.

The 05 / 06 / signal-cards triangle. These three matrices carry different questions:

05 asks "where do participants agree?" — consensus and coverage. Its empty cells are information (who didn't hit checkout?). Nothing in the product answers this today.

06 asks "where is the negative signal densest?" — which is adjacent to what signal cards already compute in the analysis lens (concentration per label|sentiment cell). Shipping 06 on the dashboard would duplicate analysis.

So the dashboard slot goes to 05's axes, at 5b's scale. The GitHub grid reads Scharp's continua at a glance: a filled column is pervasive (Beds browsing — 3 of 4 rows, three different hues: everyone went, nobody agrees); a lone saturated cell is sporadic-but-strong (Aftercare, one row, full red); a split column is live disagreement (Checkout: p3 green, p4 red). Click through to the analysis lens for the full-fat version. Compare docs/mockups/mini-heatmaps-dashboard.html — the earlier GitHub-grid experiment this remixes.

06 — where the pain is

Friction heatmap

DISCUSSPairs with 05 — see the triangle discussion there. Strong chart, but its question ("where is negative signal densest?") is signal-cards territory in the analysis lens. If it ships, it ships there — not as a second dashboard matrix.

Lineage: Sports analytics — soccer xG heatmaps, shot-density charts. Topics × sessions, cell heat = negative-sentiment density.

Topic / screen

s1 healthcare

s2 beds

s3 beds

s4 duvets

Post-surgery aftercare

31 str.

—

Top navigation

—

1

—

Beds category page

—

2

0

Login / app

—

1

—

Shopping bag

—

1

—

Checkout & delivery

—

0

1

Cell = count of frustration + confusion + doubt quotes on that topic in that session. "str." marks an intensity-3 quote. Reading: beds category is spread across three sessions (product problem); aftercare is one participant, one session, maximum heat (single-voice, still a finding).

07 — dense

Session small multiples

OPENNo verdict yet. The chronology column overlaps with 02's tape; if 02 ships, this row likely slims to donut + intensity + topics.

Lineage: Tufte, The Visual Display of Quantitative Information. Bertin, Semiology of Graphics. Same thumbnail per session, compared at a glance.

SESSION

SENTIMENT MIX

INTENSITY LADDER

CHRONOLOGY

TOPIC SPREAD

s1 p1 · 4:22 10 quotes · healthcare

2● 3● 1● + 4○

peak intensity 3× (frust ×2, delight)

pos → neg → pos arc

3 topics · aftercare heavy

s3 p3 · 4:06 14 quotes · IKEA beds

3● 4● 2● 1● + 4○

shallow but many · 1 spike

confusion → satisfaction

7 topics · widest spread

Two of four rows shown (s1, s3 — the two long sessions).

08 — the front page

Hero quote — "quote of the study"

SHORTLISTThe poster frame — open the study and go "oh yeah, that one." Selection already ships: pick_featured_quotes() (export_core.py) filters to 12–33 words, scores intensity + sentiment + researcher-context, then diversifies across participants and polarities. This idea is the presentation upgrade: rank 1 gets the editorial treatment, runners-up stay small.

Lineage: New Yorker pull quote, FT Weekend leader. One quote, big serif type, air around it. Designed to be screenshotted into a stakeholder deck.

Quote of the study — p1 · s1 · 03:43

That is nice to actually say that, to send that to someone. I think that's really nice.

Participant 1 · post-surgery notifications · intensity ●●● delight · s1 @ 03:43

RUNNER-UP · frustration ●●●

"There was no aftercare, the aftercare was quite poor."

RUNNER-UP · surprise ●●

"179. That's quite cheap."

RUNNER-UP · confidence ●

"Beds and mattresses. That's easy. That's in the top navigation."

09 — the philologist's index

Study fingerprint + KWIC concordance

PURSUEThe word-strip is now the lead: real frequencies from the 42 quotes, each chip carrying a sentiment-mix underline. Twelve words in, you know what kind of study this is — that's the fingerprint. Click a chip, get the concordance.

Lineage: Key-Word-In-Context — philology's standard tool since the 1950s (Voyant, AntConc); classical concordances. Word clouds are decorative; concordances are readable evidence.

Every count below is real. The payoff word is back — 7 uses, frustration-leaning: "went back to my GP", "back to A&E", "referred back". The retry word. A researcher sees the red-tinted underline and clicks; the concordance shows the loop the participant was stuck in. Type size follows frequency; the underline splits by the sentiment mix of quotes containing the word.

okay15 bed12 shopping10 ikea8 beds8 bag8 surgery7 back7 read5 enough5 see5 easy2

s1·p1

…I felt that it wasn't healing … I went

back

to my GP and then they referred me back to North Mid…

aftercare

s1·p1

…and I said no, can I go

back

to another hospital? So I did.

aftercare

s1·p1

…they must be better than North London Hospital. So I just went

back

there.

aftercare

s1·p1

I went

back

to A&E and I showed them my stomach… I went to A&E about four times.

aftercare

Reading the pattern: four of seven uses of "back" are one participant re-entering a system that discharged them. The word itself is the journey map. This is what a concordance does that a word cloud cannot: it keeps the grammar, and the grammar is the finding.

10 — the methods reflex

Saturation curves — topics vs themes

DEVELOPNow two curves. Raw topic labels barely repeat across sessions (every topic in this corpus is session-unique), so a topic-level curve over-reads novelty. The honest curve saturates on clustered themes. Verdict logic: 0 new themes across the 2 most recent sessions → saturated.

Lineage: Glaser & Strauss (1967), grounded theory. Guest, Bunce & Johnson (2006), How many interviews are enough? The chart every methods reviewer wants and no commercial competitor ships.

raw topic labels (naive novelty) clustered themes (honest saturation)

Themes saturated; topics still novel.

No new theme since s3 — the pattern has stabilised. New sessions are adding colour, not categories.

Recommendation: 1–2 more sessions to confirm; expect theme count to hold at 4.

Verdict states: Not yet (curve steepening) · Approaching (last session added ≤1) · Saturated (0 new across last 2 sessions).

Topic counts are ground truth (3 → 5 → 12 → 16 cumulative). Theme-per-session attribution is approximated for the mockup; the real implementation keys on theme-cluster membership, which the pipeline already computes at s11.

Ten ways to show what people actually said.