Snack Club 2023 Wrapped Data Analysis

04 Jan 2024

Intro

png

Snack Club is four friends getting together ~once a week to rate snacks. Think of it like a book club, but for snacks. This is the data analysis of snack scores given in 2023, in the spirit of the “wrapped” trend started by Spotify.

This project may not make a whole lot of sense to outsiders, but I wanted to share it because (1) I utilized my data analysis skills and (2) snacks are fun, right?

Some further context info:

Members received a monthly box of snacks from the subscription service Snack Crate. Each box was themed around a country or culture.
Snacks were rated in four categories: packaging, presentation, taste, and texture.
The score range was 1-5. Some members strictly gave scores in 0.5 intervals, while others gave whatever score they felt like (e.g., 2.9).
Drinks are considered snacks (why the hell not).

Analysis

Below is the Python code and some commentary for the analysis. In short, I took the snack review data, cleaned/prepared it, analyzed it in various ways (by member, snack, box/country, etc.), and generated visualizations (tables and plots). The two Python packages I used heavily are pandas and matplotlib.

Note: this was done in a JupyterLab notebook. The HTML version is here, which might look better but has less polish and commentary.

Code is on GitHub here.

Import Data

The data was originally input into a shared Google sheets workbook. I manually moved the data around and saved them into CSV files. From there I processed them with Python and pandas. This included some simple quality checks for missing values and correct data types.

import glob
data_files = glob.glob("./input/*.csv")
data_files

['./input/2023 data pre-processing - J.csv',
 './input/2023 data pre-processing - E.csv',
 './input/2023 data pre-processing - C.csv',
 './input/2023 data pre-processing - A.csv']

import pandas as pd
pd.set_option('display.precision', 2)

df_list = []
for fn in data_files:
    df = pd.read_csv(fn)
    df['Member'] = fn[-5]
    df_list.append(df)
df = pd.concat(df_list)
df['Date'] = pd.to_datetime(df['Date'], format='%m/%d/%Y')
df = df.sort_values(['Date', 'Country'])
df = df.set_index(['Date','Country','Snack','Member'])

# Identify missing scores
import numpy as np
print('Any null values?')
display(df[df.isna().any(axis=1)])
print('Any non-numeric values?')
display(df[~df.applymap(np.isreal).any(axis=1)])

Any null values?

				Packaging	Presentation	Taste	Texture
Date	Country	Snack	Member

Any non-numeric values?

				Packaging	Presentation	Taste	Texture
Date	Country	Snack	Member

Stats

The “overall score” is a mean of the four score categories assigned to each snack (packaging, presentation, taste, texture).

# Overall Score
df['Overall Score'] = df.mean(axis=1)
df.to_csv('2023_snack_club_scores.csv')

Overall Statistics

Present a few simple overall metrics, like number of snacks rated and total scores given.

# Stats overall
print('2023 Snack Club Statistics!')
print('Snacks rated:', df.reset_index()['Snack'].value_counts().count())
print('Countries / snack crates #:', df.reset_index()['Country'].value_counts().count())
print('Countries / snack crates:', df.reset_index()['Country'].value_counts().index.values)
print('Snack scores, total given:', df.count().sum())

2023 Snack Club Statistics!
Snacks rated: 147
Countries / snack crates #: 14
Countries / snack crates: ['Thailand' 'Colombia' 'United Kingdom' 'Italy' 'Greece' 'Belgium'
 'Ireland' 'India' 'China' 'Korea' 'Spain' 'Canada'
 'America-Vegas-Special' 'Caribbean']
Snack scores, total given: 2940

By Member

Looking at the scores by member. The types of questions trying to answer here include “who gave the highest/lowest scores” and “who was the harshest critic in the taste category”.

# display(df.agg(['idxmax','idxmin']))
idx = pd.IndexSlice
# Stats Members
display('*************************************')
display('Stats by Member')
df_stats_members = df.groupby(['Member']).agg(['min','max','mean','median','std','idxmax','idxmin'])
display(df_stats_members.loc[:, idx[:, ['min','max']]])
display(df_stats_members.loc[:, idx[:, ['mean','median','std']]])
# display(df_stats_members.loc[:, idx[:, ['idxmax','idxmin']]])
display(df_stats_members.loc[:, idx[:, ['mean']]].agg(['idxmax','idxmin']))
display(df_stats_members.loc[:, idx[:, ['std']]].agg(['idxmax','idxmin']))

'*************************************'

'Stats by Member'

	Packaging		Presentation		Taste		Texture		Overall Score
	min	max	min	max	min	max	min	max	min	max
Member
A	1.75	5.0	1.5	4.8	1.5	4.75	1.0	5.0	1.75	4.55
C	1.50	5.0	1.5	5.0	1.0	4.50	1.0	5.0	1.88	4.50
E	1.50	5.0	1.0	5.0	1.2	5.00	1.0	4.7	1.62	4.45
J	1.50	5.0	1.0	5.0	1.0	5.00	1.0	5.0	1.50	4.62

	Packaging			Presentation			Taste			Texture			Overall Score
	mean	median	std	mean	median	std	mean	median	std	mean	median	std	mean	median	std
Member
A	3.32	3.4	0.79	3.16	3.0	0.73	3.55	3.75	0.74	3.38	3.5	0.82	3.35	3.38	0.53
C	3.47	3.5	0.73	3.31	3.5	0.73	3.48	3.50	0.78	3.52	3.5	0.78	3.44	3.50	0.50
E	3.26	3.3	0.73	2.96	3.0	0.88	3.37	3.50	0.83	3.24	3.4	0.84	3.21	3.25	0.56
J	3.18	3.0	0.84	3.12	3.0	0.79	3.14	3.00	0.92	3.28	3.5	0.81	3.18	3.25	0.56

	Packaging	Presentation	Taste	Texture	Overall Score
	mean	mean	mean	mean	mean
idxmax	C	C	A	C	C
idxmin	J	E	J	E	J

	Packaging	Presentation	Taste	Texture	Overall Score
	std	std	std	std	std
idxmax	J	E	J	E	J
idxmin	E	C	A	C	C

By Snack

Moving onto looking at snack scores by… snack. This is to answer the all important question of which are the best and worst snacks. This is done with a simple mean of the member score categories and the overall score. Looking at the standard deviation, we can also get a sense of which snacks are the most “controversial”. For example, Keogh’s Shamrock and Sour Cream Chips was the best tasting snack across all members, while the Big Roll Seaweed had the highest standard deviation, a reflection of how it was either liked a lot or not at all.

Top snacks by category, overall

# # Stats Snacks
display('Stats by Snack')
df_stats_snacks = df.groupby(['Snack']).agg(['min','max','mean','median','std'])
idx = pd.IndexSlice
df_ss_mean = df_stats_snacks.loc[:, idx[:, ['mean']]].droplevel(level=1,axis='columns').sort_values(by=['Taste','Overall Score','Texture'],ascending=False)
display(df_ss_mean)
display('*************************************')
df_ss_mm = df_stats_snacks.loc[:, idx[:, ['mean','median','std']]]
display(df_ss_mm.agg(['idxmax','idxmin'])) # only returns first instance

'Stats by Snack'

	Packaging	Presentation	Taste	Texture	Overall Score
Snack
Keogh's Shamrock and Sour Cream Chips	4.69	4.20	4.62	4.05	4.39
Prele-G Gold Biscuits	2.25	4.38	4.62	4.00	3.81
Hi-Chew Grapefruit	4.40	3.70	4.50	4.14	4.18
Salati Preziosi Parika chips	3.92	3.25	4.50	3.50	3.79
Smiths Scampi Flavor fries	2.75	4.06	4.42	4.00	3.81
...	...	...	...	...	...
Lotte Choco Pie	2.42	2.75	2.08	2.05	2.33
Choclitos Limon chips	2.75	3.20	2.05	3.38	2.84
Kerr's Toffee candy	3.00	1.62	2.00	1.82	2.11
Haitai Apple French Pie	3.12	1.88	1.64	1.50	2.03
Sherbert fountain	3.00	1.25	1.38	1.12	1.69

147 rows × 5 columns

'*************************************'

	Packaging			Presentation			Taste			Texture			Overall Score
	mean	median	std	mean	median	std	mean	median	std	mean	median	std	mean	median	std
idxmax	Hard MtnbDew Baja Blast Zero Sugar	Hard MtnbDew Baja Blast Zero Sugar	Chips Ahoy Cranberry	Soejun Yang Jelly "cat paws"	Soejun Yang Jelly "cat paws"	Tronky Nocciola	Keogh's Shamrock and Sour Cream Chips	Keogh's Shamrock and Sour Cream Chips	Big Roll Seaweed	Cote D'OR chocolate	Cote D'OR chocolate	Millions strawberry fruit candy	Soejun Yang Jelly "cat paws"	Soejun Yang Jelly "cat paws"	Pororo Straw Type Jelly
idxmin	Hula Hoops Puft Grilled Beef	Hula Hoops Puft Grilled Beef	Caffrey's Big Time Chocolate Caramel Bar	Sherbert fountain	Sherbert fountain	5050 Maska Chaska	Sherbert fountain	Sherbert fountain	Hi-Chew Grapefruit	Lotte Custard Cake	Lotte Custard Cake	Bear Paws Pattes d'ours	Sherbert fountain	Sherbert fountain	Koh Kae Coconut Cream Peanuts

Plots of Best Snacks by Average Member Score & Category

Plot the best/worst 5 snacks by the categories.

import matplotlib.pyplot as plt

df_mean = df_stats_snacks.loc[:, idx[:, ['mean']]]
df_mean.columns = df_mean.columns.droplevel(1)

fig, axs = plt.subplots(5,1,figsize=(3,18))
keys = [x for x in df_mean.columns] # preserves order
targets = zip(keys, axs.flatten())
fig.suptitle('Snack Club 2023 Best & Worst Snacks', fontsize=24)
fig.supylabel('Snacks',x=-1.4,ha='left', fontsize=18)
fig.supxlabel('Average Score', fontsize=18)
colors = {'Packaging':'blue',
          'Presentation':'green',
          'Taste':'red',
          'Texture':'purple',
         'Overall Score': 'cyan'}
by_label = {}
for i, (key, ax) in enumerate(targets):
    df_topfive = df_mean[key].sort_values(ascending=False)[0:5].sort_values()
    df_lowfive = df_mean[key].sort_values()[0:5].sort_values()
    df_low_top = pd.concat([df_topfive, pd.Series(['~',0]), df_lowfive])
    
    ax.set_title(f"{key}", fontsize=14,pad=10)
        
    snacks = df_lowfive.index.values
    scores = df_lowfive.values
    color = 'red'
    ax.barh(snacks, scores,
           color=color,label=key, 
           alpha=.5,height=0.5)
    
    ax.barh(['~~~'], [0],height=0.0)
    
    snacks = df_topfive.index.values
    scores = df_topfive.values
    color = 'green'
    ax.barh(snacks, scores,
           color=color,label=key, 
           alpha=.5,height=0.5)
fig.align_labels() 
fig.subplots_adjust(top=0.93,bottom=0.05,hspace = 0.3, wspace = 0.1)
fig.savefig(f'plots/sc_bestworst_snack_avg.png',bbox_inches='tight')

png

Plots of Snack Standard Deviation Scores by Member Score & Category

Plot the best/worst 5 snacks by standard deviation, again a measure of how “controversial” or not the snack was. Note how in multiple cases all the same scores were assigned for a snack and so the standard deviation was 0. This is especially seen in the texture category.

import matplotlib.pyplot as plt

df_std = df_stats_snacks.loc[:, idx[:, ['std']]]
df_std.columns = df_std.columns.droplevel(1)

fig, axs = plt.subplots(5,1,figsize=(3,18))
keys = [x for x in df_std.columns] # preserves order
targets = zip(keys, axs.flatten())
fig.suptitle('Snack Club 2023 Lest and Most Controversial Snacks', fontsize=24)
fig.supylabel('Snacks',x=-1.4,ha='left', fontsize=18)
fig.supxlabel('Score Standard Deviation', fontsize=18)
colors = {'Packaging':'blue',
          'Presentation':'green',
          'Taste':'red',
          'Texture':'purple',
         'Overall Score': 'cyan'}
by_label = {}
for i, (key, ax) in enumerate(targets):
    df_topfive = df_std[key].sort_values(ascending=False)[0:10].sort_values()
    df_lowfive = df_std[key].sort_values()[0:10].sort_values()
    df_low_top = pd.concat([df_topfive, pd.Series(['~',0]), df_lowfive])
    
    ax.set_title(f"{key}", fontsize=14,pad=10)
        
    snacks = df_lowfive.index.values
    scores = df_lowfive.values
    color = 'red'
    ax.barh(snacks, scores,
           color=color,label=key, 
           alpha=.5,height=0.5)
    
    ax.barh(['~~~'], [0],height=0.0)
    
    snacks = df_topfive.index.values
    scores = df_topfive.values
    color = 'green'
    ax.barh(snacks, scores,
           color=color,label=key, 
           alpha=.5,height=0.5)
fig.align_labels() 
fig.subplots_adjust(top=0.93,bottom=0.05,hspace = 0.3, wspace = 0.1)
fig.savefig(f'plots/sc_bestworst_snack_std.png',bbox_inches='tight')

png

Plots, Best Snacks by Member & Category

For each member, create a plot of their best/worst snacks in each category. Save the plots as pictures because they are so large.

df_stacked = df.stack().reset_index().rename(columns={'level_4':'Category',
                                        0 : 'Score'})
df_stacked = df_stacked.set_index(['Date','Country','Snack',
                                  'Member','Category'])

grouped = df_stacked.groupby(['Member','Category'],sort=False)
keys = [x for (x,y) in grouped] # preserves order
members = [k[0] for k in keys]
# categories = [k[1] for k in keys]
categories = ['Overall Score','Taste','Texture','Presentation','Packaging']

for i, memb in enumerate(members):
    fig, axs = plt.subplots(5,1,figsize=(4,24))
    targets = zip(categories, axs.flatten())
    fig.suptitle(f'Snack Club 2023 Best & Worst Snacks - Member {memb}', fontsize=24)
    fig.supylabel('Snacks',x=-0.3,ha='left', fontsize=18)
    fig.supxlabel('Score', fontsize=18)
    by_label = {}
    for j, (cat,ax) in enumerate(targets):
        idx = pd.IndexSlice
        df_f = df_stacked.loc[idx[:, :, :, memb, cat]]
        df_top = df_f['Score'].sort_values(ascending=False)[0:10].sort_values()
        df_low = df_f['Score'].sort_values()[0:10].sort_values()
        # df_low_top = pd.concat([df_top, pd.Series(['~',0]), df_low])
        # display(df_low_top)
        ax.set_title(f"{memb} - {cat}", fontsize=14,pad=10)
        snacks = [x[2] for x in df_low.index.values]
        scores = df_low.values
        color = 'red'
        ax.barh(snacks, scores,
               color=color,label=cat, 
               alpha=.5,height=0.6)

        ax.barh(['~~~'], [0],height=0.0)

        snacks = [x[2] for x in df_top.index.values]
        scores = df_top.values
        color = 'green'
        ax.barh(snacks, scores,
               color=color,label=cat, 
               alpha=.5,height=0.5)
        ax.set_xticks([1,2,3,4,5])
    fig.subplots_adjust(top=0.93,bottom=0.05,hspace = 0.3, wspace = 0.2)
    fig.savefig(f'plots/sc_bestworst_snack_{memb}.png',bbox_inches='tight')
    plt.close(fig)

By Country/Crate

Every snack is associated with a country/culture theme per Snack Crate. In addition, there’s a bonus “Las Vegas” category for snacks rated while on a trip to Las Vegas. Look at snack scores by these countries/cultures, see which ones had best/worst scores, etc.

# # Stats Country

display('*************************************')
display('Stats by Country')
df_stats_country = df.groupby(['Country']).agg(['min','max','mean','median','std'])
idx = pd.IndexSlice
# display(df_stacked.droplevel(['Member','Category','Date']).reset_index().drop(columns='Score').set_index())
df_sc = df_stats_country.loc[:, idx[:, ['mean','std']]].copy()
# better way to get count of countries? lol
c_count = df_stacked.reset_index()[['Country','Snack']].drop_duplicates().set_index(['Country']).groupby(['Country']).agg(['count'])
# df_sc[('','count')] = c_count
df_sc = pd.concat([df_sc,c_count], axis=1)

display(df_sc.sort_values(by=('Taste','mean'),ascending=False))
display(df_sc.agg(['idxmax','idxmin']))

'*************************************'

'Stats by Country'

	Packaging		Presentation		Taste		Texture		Overall Score		Snack
	mean	std	mean	std	mean	std	mean	std	mean	std	count
Country
Caribbean	3.18	0.73	2.82	0.42	3.73	0.56	3.72	0.58	3.36	0.33	4
Thailand	3.60	0.75	2.86	0.59	3.56	0.89	3.42	0.70	3.36	0.46	16
Belgium	2.89	0.75	3.42	0.58	3.54	0.80	3.62	0.81	3.37	0.46	11
China	3.89	0.57	3.29	0.85	3.51	0.76	3.40	0.79	3.52	0.54	11
America-Vegas-Special	3.81	0.85	3.75	0.48	3.46	0.80	3.59	0.38	3.65	0.30	4
Ireland	3.45	0.78	3.15	0.79	3.45	0.89	3.35	0.78	3.35	0.61	11
Italy	3.12	0.59	3.25	0.96	3.43	0.69	3.45	0.85	3.31	0.64	12
Spain	2.83	0.78	3.07	0.68	3.39	0.75	3.49	0.68	3.19	0.52	10
India	3.11	0.79	2.94	0.81	3.38	0.91	3.31	0.39	3.19	0.38	11
Canada	3.33	0.72	3.39	0.90	3.36	0.93	3.22	0.87	3.33	0.64	9
Colombia	3.26	0.60	2.96	0.56	3.34	0.69	3.35	0.78	3.23	0.40	14
Greece	3.18	0.77	3.46	0.56	3.26	0.84	3.42	0.67	3.33	0.41	12
Korea	3.40	0.89	3.02	1.01	3.22	0.86	2.83	1.11	3.12	0.78	10
United Kingdom	3.33	0.77	2.91	1.00	2.94	0.95	3.07	1.15	3.06	0.71	12

	Packaging		Presentation		Taste		Texture		Overall Score		Snack
	mean	std	mean	std	mean	std	mean	std	mean	std	count
idxmax	China	Korea	America-Vegas-Special	Korea	Caribbean	United Kingdom	Caribbean	United Kingdom	America-Vegas-Special	Korea	Thailand
idxmin	Spain	China	Caribbean	Caribbean	United Kingdom	Caribbean	Korea	America-Vegas-Special	United Kingdom	America-Vegas-Special	America-Vegas-Special

Best/West Snack in Country/Crate

df_stats_cs = df.groupby(['Country','Snack']).agg(['mean'])
idx = pd.IndexSlice
df_scs = df_stats_cs.loc[:, idx[:, ['mean']]]
display('Best Snacks By Country (Average Score)')
display(df_scs.groupby('Country').agg(['idxmax']).applymap(lambda x: x[1]).droplevel(axis='columns',level=[2,1]))
display('Worst Snacks By Country (Average Score)')
display(df_scs.groupby('Country').agg(['idxmin']).applymap(lambda x: x[1]).droplevel(axis='columns',level=[2,1]))

'Best Snacks By Country (Average Score)'

	Packaging	Presentation	Taste	Texture	Overall Score
Country
America-Vegas-Special	Hard MtnbDew Baja Blast Zero Sugar	Hard MtnbDew Baja Blast Zero Sugar	Rap Snacks Nikki Minaj Salt & Vinegar & Trufflesl	Rap Snacks Lily Baby "All In" Flavor Salt & Vi...	Hard MtnbDew Baja Blast Zero Sugar
Belgium	Lotus Biscoff with Belgium Choclate	Cote D'OR chocolate	Cote D'OR chocolate	Cote D'OR chocolate	Cote D'OR chocolate
Canada	Old Dutch Ketchup Flavored Potatoe Chips	KitKat Chunky (with french on it)	Humpty Dumpty Ringolos BBQ Flavor	Ruffles "All Dressed"	Humpty Dumpty Ringolos BBQ Flavor
Caribbean	Ole Xtreme Rranch Tortilla Chips	Ole Xtreme Rranch Tortilla Chips	Tiki Gold Hazelnut	Tiki Gold Hazelnut	Tiki Gold Hazelnut
China	Orion Peach Jelly Gummies	WantWant Honey Kissed Crunchy Rice Balls (bee ...	Hi-Chew Grapefruit	Chips Ahoy Cranberry	Hi-Chew Grapefruit
Colombia	Jumbo Flow	Jumbo Rosca	Supercoco Choco Snacks	Chokis Chocolatoasas y Deliciosass	Jumbo Flow
Greece	Derby Caramel Bar	Jumbo Vrikoles, cheese and tomoato	Kiss Strawberry milk chocolate strawberry bar	Kiss Strawberry milk chocolate strawberry bar	Derby Caramel Bar
India	Waffy vanilla wafers	Prele-G Gold Biscuits	Prele-G Gold Biscuits	Prele-G Gold Biscuits	Prele-G Gold Biscuits
Ireland	Keogh's Shamrock and Sour Cream Chips	Keogh's Shamrock and Sour Cream Chips	Keogh's Shamrock and Sour Cream Chips	Keogh's Shamrock and Sour Cream Chips	Keogh's Shamrock and Sour Cream Chips
Italy	Salati Preziosi Parika chips	Giovanni Cova & C Crostatina albicocca (aprica...	Salati Preziosi Parika chips	Giovanni Cova & C Crostatina albicocca (aprica...	Giovanni Cova & C Crostatina albicocca (aprica...
Korea	Soejun Yang Jelly "cat paws"	Soejun Yang Jelly "cat paws"	Soejun Yang Jelly "cat paws"	Soejun Yang Jelly "cat paws"	Soejun Yang Jelly "cat paws"
Spain	el Valle Sabor Huevo Frito Fried Egg Taste	Dulcesol palmeritas	Dulcesol palmeritas	Dulcesol palmeritas	el Valle Sabor Huevo Frito Fried Egg Taste
Thailand	Pocky Sumer Paradise Peach Lychee & Elderflower	Jack & Jill Cocont Crackers	Manora Fried Pumpkin Chips	Jack & Jill Cocont Crackers	Jack & Jill Cocont Crackers
United Kingdom	Jaffa Cakes "The Original"	Rowntrees Random fruit snacks	Space Raiders Spicy	Cadbury crunchie chocolate bar	Space Raiders Spicy

'Worst Snacks By Country (Average Score)'

	Packaging	Presentation	Taste	Texture	Overall Score
Country
America-Vegas-Special	Rap Snacks Nikki Minaj Salt & Vinegar & Trufflesl	Rap Snacks Lily Baby "All In" Flavor Salt & Vi...	Rap Snacks Lily Baby "All In" Flavor Salt & Vi...	Hard MtnbDew Baja Blast Zero Sugar	Rap Snacks Lily Baby "All In" Flavor Salt & Vi...
Belgium	Gicopa violet	Grefin(?) Eclair	DeliChoc Choco'Bar	DeliChoc Choco'Bar	DeliChoc Choco'Bar
Canada	Coffee Crisp Double Double Chocolate Bar	Kerr's Toffee candy	Kerr's Toffee candy	Kerr's Toffee candy	Kerr's Toffee candy
Caribbean	Tiki Gold Hazelnut	Jamaican Choice Spice Bun	Ole Xtreme Rranch Tortilla Chips	Cherry Berries	Jamaican Choice Spice Bun
China	Oishi Orange Candy	Aji Oolong Peach Jelly	Oreo Oolong + Peach	Aji Oolong Peach Jelly	Aji Oolong Peach Jelly
Colombia	Chokis Chocolatoasas y Deliciosass	Chokis Chocobase cookies	Choclitos Limon chips	Colombina Coffee Delight chewy candy	Choclitos Limon chips
Greece	Kiss Strawberry milk chocolate strawberry bar	Derby Caramel Bar	Keraso Pitaki cherry pie	Keraso Pitaki cherry pie	Keraso Pitaki cherry pie
India	5050 Maska Chaska	Rola-a-cola "The solid cola"	Treat Burst Choco Fills	Cheetos Masala Balls	Treat Burst Choco Fills
Ireland	Cadbury Snack! shortcake	Burts Guinness Thick Cut Hand Cooked Potato Chips	Caffrey's Big Time Chocolate Caramel Bar	Cleeves Original Toffee	Cleeves Original Toffee
Italy	Buondi Motta Cioccolato cake	Buondi Motta Cioccolato cake	Buondi Motta Cioccolato cake	Buondi Motta Cioccolato cake	Buondi Motta Cioccolato cake
Korea	Lotte Custard Cake	Haitai Apple French Pie	Haitai Apple French Pie	Lotte Custard Cake	Haitai Apple French Pie
Spain	nocilla sin chocoleche dip sticks	Palotes damel sabor fresa	Fini Gourmet Red liqurice srawberry	Fini Gourmet Red liqurice srawberry	Palotes damel sabor fresa
Thailand	Arnotts Stikko fingers chocolate wafer sticks	Big Roll Seaweed	Black Pink "venom cream" (Strawberry) Oreos	Big Roll Seaweed	Lush chocolate chewy candy
United Kingdom	Hula Hoops Puft Grilled Beef	Sherbert fountain	Sherbert fountain	Sherbert fountain	Sherbert fountain

Score Spread, Histograms

Plot histograms to get a sense of the spread of scores given by each members for the four categories. The histogram bin sizes are 0.5.

df_stacked = df.stack().reset_index().rename(columns={'level_4':'Category',
                                        0 : 'Score'})
df_stacked = df_stacked.set_index(['Date','Country','Snack',
                                  'Member','Category'])
# df_stacked = df_stacked.reindex(
#     labels=['Packaging','Presentation','Taste',
#             'Texture','Overall Score'], level='Category')


df_stacked = df_stacked.drop(index='Overall Score',level='Category')
fig, axs = plt.subplots(4,4,figsize=(10,10))
groups1 = ['Member','Category']
groups2 = ['Category']
grouped = df_stacked.groupby(groups1,sort=False)
keys = [x for (x,y) in grouped] # preserves order
targets = zip(keys, axs.flatten())
fig.suptitle('Snack Club 2023 Score Spreads', fontsize=24)
fig.text(0.5,0.92,'Histogram, bin size = 0.5', va='center',ha='center', fontsize=18)
fig.supxlabel('Score (1.0-5.0)', fontsize=18)
fig.supylabel('Frequency', fontsize=18)
colors = {'Packaging':'blue',
          'Presentation':'green',
          'Taste':'red',
          'Texture':'purple'}
by_label = {}
for i, (key, ax) in enumerate(targets):
    j = i%4
    member=key[0]
    category = key[1]
    if i < 4:
        ax.set_title(f"{category}", fontsize=18,
                    pad=10)
    if j == 0:
        ax.set_ylabel(f"{member}", rotation=0, fontsize=18,
                     labelpad=10)
    color = colors[category]
    dfg = grouped.get_group(key)
    dfg = dfg.reset_index(drop=True)
    dfg = dfg.reset_index()
    # display(dfCategory)
    avg = dfg.mean()[1]
    counts, bins = np.histogram(dfg['Score'],range=(0,5.0))
    ax.hist(bins[:-1], bins, weights=counts,
               color=color,label=category, 
               alpha=.5)
    ax.axvline(x=avg, color='maroon', linestyle='--', 
               label="Average",
               linewidth=1.5)
    ax.set_ylim([0,55])
    ax.set_xlim([0.9,5.1])
    ax.text(avg-1, 44, f'{avg:.2f}', ha='left',color='maroon')
    ax.set_xticks([1,2,3,4,5])
    handles, labels = ax.get_legend_handles_labels()
    by_label.update(dict(zip(labels, handles))) # removes duplicate legends
order = [0,2,3,4,1]
lvals = [x for x in by_label.values()]
lvals = [lvals[idx] for idx in order]
kvals = [x for x in by_label.keys()]
kvals = [kvals[idx] for idx in order]
fig.legend(lvals,kvals , loc='upper right')
fig.subplots_adjust(top=0.85,bottom=0.1, hspace = 0.3, wspace = 0.3)
fig.savefig('plots/sc_2023_histograms.png')

png

Score Spread, Bar Chart Value Counts

A plot to get a different perspective on score spreads. Instead of a histogram, plot bar charts plotting unique scores counts. This will especially illustrate how some members gave scores in scrict 0.5 intervals and others did not.

import matplotlib.pyplot as plt


df_stacked = df.stack().reset_index().rename(columns={'level_4':'Category',
                                        0 : 'Score'})
df_stacked = df_stacked.set_index(['Date','Country','Snack',
                                  'Member','Category'])
df_stacked = df_stacked.reindex(
    labels=['Packaging','Presentation','Taste',
            'Texture','Overall Score'], level='Category')


df_stacked = df_stacked.drop(index='Overall Score',level='Category')
fig, axs = plt.subplots(4,4,figsize=(10,10))
groups1 = ['Member','Category']
groups2 = ['Category']
grouped = df_stacked.groupby(groups1,sort=False)
keys = [x for (x,y) in grouped] # preserves order
targets = zip(keys, axs.flatten())
fig.suptitle('Snack Club 2023 Score Spreads', fontsize=24)
fig.text(0.5,0.92,'Value Counts', va='center',ha='center', fontsize=18)
fig.supxlabel('Score (1.0-5.0)', fontsize=18)
fig.supylabel('Count', fontsize=18)
colors = {'Packaging':'blue',
          'Presentation':'green',
          'Taste':'red',
          'Texture':'purple'}
by_label = {}
for i, (key, ax) in enumerate(targets):
    j = i%4
    member=key[0]
    category = key[1]
    if i < 4:
        ax.set_title(f"{category}", fontsize=18,
                    pad=10)
    if j == 0:
        ax.set_ylabel(f"{member}", rotation=0, fontsize=18,
                     labelpad=10)
    dfg = grouped.get_group(key)
    dfg = dfg.reset_index(drop=True)
    avg = dfg['Score'].mean()
    dfg = dfg.value_counts().reset_index()
    dfg = dfg.rename(columns={0:'count'})
    # display(dfCategory)
    color = colors[category]
    ax.bar(dfg['Score'],dfg['count'],
               color=color,label=category, 
               alpha=.5,width=0.4)
    ax.axvline(x=avg, color='maroon', linestyle='--', 
               label="Average",
               linewidth=1.5)
    ax.set_ylim([0,50])
    ax.set_xlim([0.5,5.5])
    ax.set_xticks([1,2,3,4,5])
    ax.text(avg-1, 45, f'{avg:.2f}', ha='left',color='maroon')
    handles, labels = ax.get_legend_handles_labels()
    by_label.update(dict(zip(labels, handles))) # removes duplicate legends
order = [1,2,3,4,0]
lvals = [x for x in by_label.values()]
lvals = [lvals[idx] for idx in order]
kvals = [x for x in by_label.keys()]
kvals = [kvals[idx] for idx in order]
fig.legend(lvals,kvals , loc='upper right')
fig.subplots_adjust(top=0.85,bottom=0.1, hspace = 0.3, wspace = 0.3)  
fig.savefig('plots/sc_2023_bar_vcs.png')

png

Number of unique scores

After seeing the last plots, now interested to see number of unique scores given by member.

display('# Unique scores')
display(df_stacked.groupby('Member').apply(lambda x: x.value_counts().shape[0]).sort_values())

'# Unique scores'



Member
C     9
J     9
A    33
E    33
dtype: int64

Ones, Fives, Below 3

Trying to dig into score spreads some more. Looking at the number of 1 and 5 scores given, the lowest and highest score possible. Also, looking at number of scores which were less than 3. Either of these metrics could indicate some how critical or not a member is, e.g., a low number of ~5 scores given means a rater who is rarely impressed.

def custom_round(x, base=5):
    return round(base * round(float(x)/base),2)

df_thresh = pd.DataFrame()
df_thresh['Score'] = df_stacked['Score']
df_thresh['Score Round 0.5'] = df_stacked['Score'].apply(lambda x: custom_round(x,base=0.5))
df_thresh['Score Rounded == 1'] = df_thresh[['Score Round 0.5']] == 1
# display(df_thresh.loc[df_thresh['Score Rounded == 1'] == True])
df_thresh['Score Rounded == 5'] = df_thresh[['Score Round 0.5']] == 5

df_thresh['Score Rounded < 3'] = df_thresh[['Score Round 0.5']] < 3
# display(df_thresh.loc[df_thresh['Score Rounded == 5'] == True])
# df_thresh['Score, around 2.5'] = (df_stacked[['Score']] > 2.25) & (df_stacked[['Score']] < 2.75)
# display(df_thresh)

for gs in [['Member'],['Category'],['Category','Member']]:
    df_thresh_mc = pd.DataFrame()
    df_thresh_mc['# Score'] = df_thresh['Score Rounded == 1'].groupby(gs).apply(lambda x: x.count())
    df_thresh_mc['# Score Rounded == 1'] = df_thresh['Score Rounded == 1'].groupby(gs).apply(sum)
    df_thresh_mc['# Score Rounded == 5'] = df_thresh['Score Rounded == 5'].groupby(gs).apply(sum)
    df_thresh_mc['# Score Rounded < 3'] = df_thresh['Score Rounded < 3'].groupby(gs).apply(sum)
    display(df_thresh_mc)

	# Score	# Score Rounded == 1	# Score Rounded == 5	# Score Rounded < 3
Member
A	588	3	11	131
C	588	3	8	108
E	588	11	6	151
J	588	6	10	168

	# Score	# Score Rounded == 1	# Score Rounded == 5	# Score Rounded < 3
Category
Packaging	588	0	17	145
Presentation	588	7	5	181
Taste	588	6	4	122
Texture	588	10	9	110

		# Score	# Score Rounded == 1	# Score Rounded == 5	# Score Rounded < 3
Category	Member
Packaging	A	147	0	5	36
	C	147	0	2	26
	E	147	0	3	33
	J	147	0	7	50
Presentation	A	147	0	2	45
	C	147	0	1	30
	E	147	6	1	56
	J	147	1	1	50
Taste	A	147	0	1	21
	C	147	1	0	28
	E	147	1	2	32
	J	147	4	1	41
Texture	A	147	3	3	29
	C	147	2	5	24
	E	147	4	0	30
	J	147	1	1	27

Score Similarity

This is looking at how similar member scores are. Basically, for ever combination of two members, calculate the difference between the scores and average them by category. For example, the highest average difference in the table directly below is 0.74 for the taste category between members C and J. This means these two members, relatively speaking at least, had the most opposing tastes in snacks. Note that I could maybe make this better by rounding all scores to nearest 0.5 interval.

import itertools
df
df_stacked = df_stacked.drop(index='Overall Score',level='Category')
df_m = df_stacked.unstack(level=-2)
df_m.columns = df_m.columns.droplevel()
# is there an easier way to do this? lol
combo = list(itertools.combinations(df_m.columns, 2))
for c in combo:
    c_str = c[0] + c[1]
    df_m[c_str] = abs(df_m[c[0]] - df_m[c[1]])

df_dif = df_m.drop(columns=['A','C','E','J'])    

# by category
df_dif_mean = df_dif.groupby('Category').mean()

# All scores
df_dif_mean_all = df_dif.mean().to_frame().T
df_dif_mean_all = df_dif_mean_all.rename(columns={0:'Average Difference, All Scores'})
df_dif_mean_all = pd.concat([df_dif_mean, df_dif_mean_all]) 
df_dif_mean_all = df_dif_mean_all.rename({0 : 'All Scores'}) 
df_dif_mean_all = df_dif_mean_all.T
df_dif_mean_all

	Packaging	Presentation	Taste	Texture	All Scores
Member
AC	0.43	0.33	0.49	0.39	0.41
AE	0.45	0.41	0.56	0.39	0.45
AJ	0.44	0.30	0.60	0.35	0.42
CE	0.35	0.48	0.56	0.44	0.46
CJ	0.47	0.41	0.74	0.45	0.52
EJ	0.41	0.41	0.68	0.40	0.47

Now looking at member score differences by snack. Just looking at a sample of three high/low differences, because there could be many that are the same. “We really agreed on these snacks” vs “We really disagreed on these snacks”

# by snack
idx = pd.IndexSlice
df_dif_taste = df_dif.unstack('Category').loc[:, idx[:, ['Taste']]].droplevel(axis='columns',level=1)
df_dif_taste_max = df_dif_taste.agg(['idxmax']).applymap(lambda x: x[2])
d_dif_taste_m = df_dif_taste.T.droplevel(axis='columns',level=[0,1]).unstack().to_frame()
d_dif_taste_m = d_dif_taste_m.reset_index().set_index(['Member','Snack']).sort_values(by=[0],ascending=False)
display(d_dif_taste_m.reorder_levels([1,0]).groupby('Member').apply(lambda x: x[-4:-1]).droplevel(axis='rows',level=2).rename(columns={0:'Taste Score Least Diff (Sample 3)'}))
display(d_dif_taste_m.reorder_levels([1,0]).groupby('Member').apply(lambda x: x[0:3]).droplevel(axis='rows',level=2).rename(columns={0:'Taste Score Most Diff (Sample 3)'}))

		Taste Score Least Diff (Sample 3)
Member	Snack
AC	Hard MtnbDew Baja Blast Zero Sugar	0.0
	Oishi Orange Candy	0.0
	Croccantelle susto Pizza	0.0
AE	Chokis Chocobase cookies	0.0
	Yoyo Cola Jelly Pieces	0.0
	Loacker Classic Cocoa & Milk Wafer	0.0
AJ	Swizzlers Refreshers Lemon Flavor	0.0
	Hard MtnbDew Baja Blast Zero Sugar	0.0
	Cleeves Original Toffee	0.0
CE	Buondi Motta Cioccolato cake	0.0
	Cabury BOOST	0.0
	Manora Fried Pumpkin Chips	0.0
CJ	Cleeves Original Toffee	0.0
	Rap Snacks Rick Ross Sweet Chilli Lemon Pepper	0.0
	Rap Snacks Lily Baby "All In" Flavor Salt & Vinegar, BBQ, Onion, garlic, and more	0.0
EJ	Buondi Motta Cioccolato cake	0.0
	Cadbury Snack! shortcake	0.0
	Rap Snacks Lily Baby "All In" Flavor Salt & Vinegar, BBQ, Onion, garlic, and more	0.0

		Taste Score Most Diff (Sample 3)
Member	Snack
AC	Parle kismi candy	2.00
	Pocky Sumer Paradise Peach Lychee & Elderflower	2.00
	Jaffa Cakes "The Original"	2.00
AE	Big Roll Seaweed	2.90
	Bear Paws Pattes d'ours	2.50
	Big Sheet Seaweed Spicy	2.50
AJ	Cadbury crunchie chocolate bar	2.50
	Chokis Chocolatoasas y Deliciosass	2.25
	KitKat Chunky (with french on it)	2.00
CE	Cheetos Masala Balls	2.50
	Big Sheet Seaweed Spicy	2.50
	Ruffles "All Dressed"	2.20
CJ	Parle kismi candy	3.00
	Lacta Oreo Bar	3.00
	Cadbury crunchie chocolate bar	3.00
EJ	Cabury Fudge	3.00
	Big Roll Seaweed	3.00
	Cucumber Lays	2.80

This plot visualizes the results described above. Especially interesting is how the taste scores are mostly higher than the other categories, an indicator perhaps that we give taste the most critical and independent thought. The table at the bottom looks at this by snack and shows the ten which could be considered the most “looked better than it tasted”.

fig, axs = plt.subplots(1,5,figsize=(10,3))
fig.suptitle('Snack Club 2023 Score Similarity', fontsize=24)
fig.supxlabel('Member Pair', fontsize=18)
fig.supylabel('Average Score Differences', fontsize=18)

keys = df_dif_mean_all.columns # preserves order
targets = zip(keys, axs.flatten())
colors = ['blue','green','red','purple','cyan']
for i, (key, ax) in enumerate(targets):
    dfg = df_dif_mean_all[[key]]
    dfg = dfg.reset_index().reset_index()
    color = colors[i]
    dfg.columns = ['index','Member Pair', 'Average Difference']
    ax.scatter(dfg['Member Pair'],dfg['Average Difference'],
               color=color,
               label=key, 
               alpha=.5)
    ax.set_ylim([0,0.8])
    ax.set_title(f"{key}", fontsize=18,
                    pad=10)
fig.subplots_adjust(top=0.7,bottom=0.2, hspace = 0.3, wspace = 0.3)
fig.savefig('plots/sc_2023_score_sim.png')

png

Tasted Better/Worst Than Looked

Trying to get a sense of which members tended to think a snack tasted better/worst than it looked.

df_tp = pd.DataFrame()
df_tp['Taste - Pres.'] = df['Taste'] - df['Presentation']
df_tp['Taste < Pres.'] = df['Taste'] < df['Presentation']
display(df_tp['Taste - Pres.'].groupby('Member').mean().rename('Taste and Presentation Difference, Average').to_frame())
display('****************************************')
display(df_tp['Taste < Pres.'].groupby('Member').apply(lambda x: f"{x.sum()/x.size*100:.1f}" +'%').rename('Tasted Worst Than Looked (%)').to_frame())
display('****************************************')
display(df_tp['Taste - Pres.'].groupby('Snack').mean().sort_values().rename('Taste and Presentation Difference (avg by memb.), Top 10')[0:10].to_frame())

	Taste and Presentation Difference, Average
Member
A	0.38
C	0.17
E	0.41
J	0.02

'****************************************'

	Tasted Worst Than Looked (%)
Member
A	22.4%
C	29.9%
E	30.6%
J	35.4%

'****************************************'

	Taste and Presentation Difference (avg by memb.), Top 10
Snack
Oreo Oolong + Peach	-1.70
Keraso Pitaki cherry pie	-1.64
Rowntrees Random fruit snacks	-1.44
Old Dutch Ketchup Flavored Potatoe Chips	-1.36
Croky Bicky flavor Chips	-1.32
Rap Snacks Lily Baby "All In" Flavor Salt & Vinegar, BBQ, Onion, garlic, and more	-1.25
Choclitos Limon chips	-1.15
Lacta Oreo Bar	-1.06
DeliChoc Choco'Bar	-1.05
Treat Burst Choco Fills	-1.00