JAR + Penalty Analysis

1 Introduksi

1.1 Latar Belakang

Just-About-Right (JAR) scale adalah consumer test yang menanyakan deviasi dari level ideal untuk intensitas atribut (sweetness, saltiness, color intensity, dll.) — skala 5 titik. Berbeda dari hedonic 9-point (yang mengukur liking), JAR mengukur perceived intensity vs ideal. Penalty analysis (Mean Drop/Mean Penalty) menjawab: "Atribut mana yang menurunkan overall liking ketika tidak JAR?" Logika: hitung mean overall liking saat panelis check JAR vs non-JAR; selisihnya = penalty. Kombinasikan dengan % panelis non-JAR untuk total mean drop. Atribut dengan total mean drop tinggi adalah driver dissatisfaction utama dan menjadi target reformulasi (Plaehn, 2012; Pagès et al., 2014). Referensi: Rothman & Parker, 2009; Schraidt, 2009.

1.2 Tujuan Modul

Membaca tabel: panelis × atribut JAR + overall liking. Menghitung distribusi JAR. Menerapkan penalty analysis. Memberikan prioritization plot: penalty vs % impacted. Audiens: praktisi R&D consumer testing, sensory scientist, marketing.

1.3 Posisi di Antara Alternatif

Pilih JAR + Penalty Analysis untuk dissatisfaction drivers. Untuk hedonic ranking: Single Factor Rate That Apply / Quick Sensory Check. Untuk biner kualitatif: CATA + Cochran Q. Untuk trained intensity profile (QDA): QDA / Descriptive Profiling.

2 Metode

2.1 Dasar Teoretis

JAR 5-point scale:

Skor	Label	Coding
1	Much too little	−2
2	Too little	−1
3	Just-about-right	0
4	Too much	+1
5	Much too much	+2

Distribusi JAR (formula display):

$$\%_{\text{JAR}} = \frac{N_{\text{JAR}}}{N_{\text{total}}} \times 100$$

Target: $\%_{\text{JAR}} \geq 70\%$ (Schraidt, 2009).

Classical penalty analysis (Mean Drop):

$$\text{Penalty}_{j,k} = \bar{Y}_{\text{JAR}} - \bar{Y}_{j,k}$$

dengan $\bar{Y}_{\text{JAR}}$ mean overall liking saat panelis check JAR untuk atribut $j$, dan $\bar{Y}_{j,k}$ mean liking saat panelis check level $k$ atribut $j$.

Total mean drop — formula kunci:

$$\boxed{\, \text{Total drop}_{j,k} = \frac{N_{j,k}}{N_{\text{total}}} \times \text{Penalty}_{j,k} \,}$$

Combined ends mean drop (formula display):

$$\text{Total drop}_{j,\text{deficient}} = \frac{N_{\text{TL}} + N_{\text{MTL}}}{N_{\text{total}}} \times \frac{N_{\text{TL}} \cdot P_{\text{TL}} + N_{\text{MTL}} \cdot P_{\text{MTL}}}{N_{\text{TL}} + N_{\text{MTL}}}$$

Kriteria prioritisasi:

Total drop	Action
> 1.0 unit	High priority reformulation
0.5–1.0	Medium priority
< 0.5	Minor; mungkin random noise

Signifikansi statistik penalty: paired t-test atau Wilcoxon signed-rank.

2.2 Persamaan Inti

%JAR: $N_{\text{JAR}} / N_{\text{total}} \times 100$
Penalty per level: $\bar{Y}_{\text{JAR}} - \bar{Y}_{\text{level}}$
Total mean drop: $(N_{\text{level}} / N_{\text{total}}) \times \text{Penalty}$
Combined ends drop: weighted average untuk deficient atau excess.

2.3 Asumsi & Batas Validitas

Asumsi	Konsekuensi jika dilanggar	Cara cek di SQalytics
$n \geq 80$ panelis konsumen	Statistik bias	Modul flag $n < 50$
JAR scale dipahami panelis (briefing)	Bias coding	Pre-test instruction
Overall liking + JAR di-rate same session	Inconsistency	SOP sequence
Multi-atribut tidak terlalu banyak ($\leq 8$)	Fatigue	Catat order
20% minimum non-JAR untuk penalty valid	Penalty tidak meaningful	Modul flag

3 Cara Kerja

3.1 Step-by-Step di SQalytics

Buka JAR + Penalty Analysis dari domain Sensori dan Riset Konsumen.
Muat tabel: Panelist, Product, Overall_Liking (1–9 hedonic), + multi-atribut JAR (1–5 each).
(Opsional) Filter per product bila multi-product test.
Pilih combined ends mode (rekomendasi) atau per-level.
Atur significance test: paired t / Wilcoxon.
Klik Run Penalty Analysis.
Tinjau hasil: Tab JAR Distribution, Tab Penalty Table, Tab Prioritization Plot, Tab Significance.

Step 5 opsional — Publication Graph Studio (PGS). JAR + Penalty Analysis sekarang menampilkan tombol 📰 Open in Publication Graph Studio. Anda dapat memilih Penalty Bubble atau Penalty Heatmap; SQalytics akan mengirim figure terpilih, ChartSpec, dan tabel penalti yang sudah dibersihkan ke PGS.

3.2 Template Tabel Input + Contoh Data Sintetis

Kolom	Tipe	Wajib	Catatan
`Panelist`	category	✓	ID konsumen
`Product`	category	✓	Nama formulasi
`Overall_Liking`	numeric (1–9)	✓	Hedonic
`Sweetness_JAR`	numeric (1–5)	✓	JAR sweetness
`Saltiness_JAR`	numeric (1–5)	✓	JAR saltiness
`Color_JAR`	numeric (1–5)	◯	Opsional

SYNTHETIC Data sintetik di bawah adalah contoh ilustratif untuk consumer test Soup-A (n = 100 panelis). Nilai sengaja dirancang untuk menunjukkan pola khas JAR + penalty: Saltiness terlalu tinggi (too much dominan), Sweetness agak kurang, Color mendekati ideal. CSV setara: docs/assets/example-data/id/sensory/template_sensory_jar_penalty.csv.

Panelist	Product	Liking	Sweet_JAR	Salt_JAR	Color_JAR
P01	Soup-A	6	3 (JAR)	4 (too much)	3 (JAR)
P02	Soup-A	5	2 (too little)	4 (too much)	3 (JAR)
P03	Soup-A	7	3 (JAR)	3 (JAR)	3 (JAR)
P04	Soup-A	4	2 (too little)	5 (much too much)	3 (JAR)
P05	Soup-A	6	3 (JAR)	4 (too much)	4 (too much)
… (n = 100 panelis total)

3.3 Contoh Luaran + Figure

Distribusi JAR per atribut (n = 100):

Atribut	Too little (%)	JAR (%)	Too much (%)
Sweetness	28	65	7
Saltiness	5	50	45
Color	10	78	12

Penalty table (combined ends):

Atribut	Level	$N$	$\bar{Y}_{\text{level}}$	$\bar{Y}_{\text{JAR}}$	Penalty	Total drop
Sweetness	Too little	28	5.4	7.2	1.8	0.50
Sweetness	Too much	7	6.8	7.2	0.4	0.03
Saltiness	Too little	5	6.5	7.5	1.0	0.05
Saltiness	Too much	45	5.8	7.5	1.7	0.77
Color	Too little	10	6.8	6.9	0.1	0.01
Color	Too much	12	6.7	6.9	0.2	0.02

Prioritization table:

Rank	Atribut + Level	Total drop	Action
1	Saltiness too much (45%)	0.77	HIGH priority — reduce salt 15–20%
2	Sweetness too little (28%)	0.50	MEDIUM — increase sugar slightly
3	Color too much	0.02	Minor — no action

Distribusi JAR per atribut dan prioritization plot penalty vs % impacted untuk Soup-A — Gambar 1. **(a)** Distribusi JAR per atribut — Saltiness hanya 50% JAR (target 70%), indikator utama masalah. **(b)** Prioritization plot: Saltiness too much (45% impacted, total drop 0.77) di zona HIGH priority; Sweetness too little (28%, drop 0.50) di zona MEDIUM; *Color* tidak prioritas.

Kesimpulan ringkas: Soup-A instant menunjukkan dua driver dissatisfaction: Saltiness too much (45% panelis flag, penalty 1.7 unit liking) memberi total mean drop 0.77 — high priority. Sweetness too little (28% panelis flag, penalty 1.8) memberi drop 0.50 — medium. Color OK (78% JAR). Recommendation: reformulasi dengan reduksi garam 15–20% dan tingkat sweet sedikit. Confirmatory test dengan target %JAR Saltiness ≥ 70%. Lanjut ke Single Factor Rate That Apply untuk uji overall liking reformulasi vs original.

4 Kesimpulan

4.1 Relevansi Real-World

Reformulasi reduksi gula/garam/lemak; renovasi produk; concept testing; cross-cultural product launching; healthy product development.

4.2 Where to Go from Here

Single Factor Rate That Apply — uji liking reformulasi.
CATA + Cochran Q — descriptor diagnostics.
Quick Sensory Check — multi-attribute hedonic.
Rothman & Parker (2009) ASTM untuk standar JAR.

⚙ Troubleshooting Cepat

%JAR < 50% di semua atribut. Produk jauh dari ideal; major reformulation needed.

Penalty kecil semua (< 0.3). Atribut bukan driver liking; cari atribut lain via descriptive.

Total drop tinggi dengan $N$ kecil. Statistik tidak reliable; pakai $n \geq 80$ konsumen.

i Riwayat Revisi

Tanggal	Revisi	Penulis
2026-05-12	Draft v2 publikasi (KaTeX JAR 5-pt + penalty/mean drop + APA Rothman-Parker/Schraidt/Plaehn)	Claude

4 Referensi

Rothman, L., & Parker, M. J. (Eds.). (2009). Just-about-right (JAR) scales: Design, usage, benefits, and risks (ASTM MNL63). ASTM International. https://doi.org/10.1520/MNL63-EB
Schraidt, M. (2009). Penalty analysis or mean drop analysis. In Just-about-right (JAR) scales (pp. 50–61). ASTM International.
Plaehn, D. (2012). CATA penalty/reward. Food Quality and Preference, 24(1), 141–152. https://doi.org/10.1016/j.foodqual.2011.10.008
Pagès, J., Berthelo, S., Brossier, M., & Gourret, D. (2014). Statistical penalty analysis. Food Quality and Preference, 32, 16–23. https://doi.org/10.1016/j.foodqual.2013.07.008
Lawless, H. T., & Heymann, H. (2010). Sensory evaluation of food: Principles and practices (2nd ed.). Springer. https://doi.org/10.1007/978-1-4419-6488-5
ASTM E2299-13. (2018). Standard guide for sensory evaluation of products by the in-home consumer. ASTM International.