Finding the best chocolate bars

📖 Background

A specialty foods import company wants to expand into gourmet chocolate bars. Let’s research this market to inform the initial approach to potential suppliers.

After finding valuable chocolate bar ratings online, we will need to explore if the chocolate bars with the highest ratings share any characteristics that could help narrow down the search for suppliers (e.g., cacao percentage, bean country of origin, etc.)

💾 The data

The data file contains the following information (source):

• “id” - id number of the review

• “manufacturer” - Name of the bar manufacturer

• “company_location” - Location of the manufacturer

• “year_reviewed” - From 2006 to 2021

• “bean_origin” - Country of origin of the cacao beans

• “bar_name” - Name of the chocolate bar

• “cocoa_percent” - Cocoa content of the bar (%)

• “num_ingredients” - Number of ingredients

• “ingredients” - B (Beans), S (Sugar), S* (Sweetener other than sugar or beet sugar), C (Cocoa Butter), (V) Vanilla, (L) Lecithin, (Sa) Salt

• “review” - Summary of most memorable characteristics of the chocolate bar

• “rating” - 1.0-1.9 Unpleasant, 2.0-2.9 Disappointing, 3.0-3.49 Recommended, 3.5-3.9 Highly Recommended, 4.0-5.0 Oustanding

Acknowledgments: Brady Brelinski, Manhattan Chocolate Society

💪 Challenge

This notebooks aims to report on the following questions:

1. What is the average rating by country of origin?

2. How many bars were reviewed for each of those countries?

3. Create plots to visualize findings for questions 1 and 2.

4. Is the cacao bean’s origin an indicator of quality?

5. [Optional] How does cocoa content relate to rating? What is the average cocoa content for bars with higher ratings (above 3.5)?

6. [Optional 2] Your research indicates that some consumers want to avoid bars with lecithin. Compare the average rating of bars with and without lecithin (L in the ingredients).

7. Summarize your findings.

Summary

• The average rating by country of origin ranges from 2.71 to 3.63. The country with the highest average rating is Tobago, whereas the lowest is Puerto Rico.

• The highest numbers of bars were reviewed for the country Venezulea with 253 bars reviewed.

• Cocoa bean’s country of origin is not an indicator of quality.

• Cocoa percent and rating is not correlated. The average cocao content for bars with higher ratings (above 3.5) is 71.108.

• The average rating of bars without Lecithin is higher than bars with the ingredient.

Imports

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from scipy.stats import pearsonr
sns.set(rc = {'figure.figsize':(20,5)})
plt.style.use('ggplot')

# Function to display value on seaborn barplot
def show_values_on_bars(axs):
    def _show_on_single_plot(ax):        
        for p in ax.patches:
            _x = p.get_x() + p.get_width() / 2
            _y = p.get_y() + p.get_height() + 0.05
            value = '{:.2f}'.format(p.get_height())
            ax.text(_x, _y, value, ha="center") 

    if isinstance(axs, np.ndarray):
        for idx, ax in np.ndenumerate(axs):
            _show_on_single_plot(ax)
    else:
        _show_on_single_plot(axs)

Load Data

# Reading in the data
df = pd.read_csv('data/chocolate_bars.csv')
# Set id as index
df.set_index('id', inplace=True)
# Take a look at the first datapoints
df.head()

	manufacturer	company_location	year_reviewed	bean_origin	bar_name	cocoa_percent	num_ingredients	ingredients	review	rating
id
2454	5150	U.S.A.	2019	Tanzania	Kokoa Kamili, batch 1	76.0	3.0	B,S,C	rich cocoa, fatty, bready	3.25
2458	5150	U.S.A.	2019	Dominican Republic	Zorzal, batch 1	76.0	3.0	B,S,C	cocoa, vegetal, savory	3.50
2454	5150	U.S.A.	2019	Madagascar	Bejofo Estate, batch 1	76.0	3.0	B,S,C	cocoa, blackberry, full body	3.75
2542	5150	U.S.A.	2021	Fiji	Matasawalevu, batch 1	68.0	3.0	B,S,C	chewy, off, rubbery	3.00
2546	5150	U.S.A.	2021	Venezuela	Sur del Lago, batch 1	72.0	3.0	B,S,C	fatty, earthy, moss, nutty,chalky	3.00

Data Size and Structure

• Dataset comprises of 2530 observations and 10 columns, of which some are categorical and numerical.

• num_ingredients and ingredients contains null values.

• There are no duplicated values.

print(f'Shape: {df.shape}')

Shape: (2530, 10)

df.info()

<class 'pandas.core.frame.DataFrame'>
Index: 2530 entries, 2454 to 2170
Data columns (total 10 columns):
 #   Column            Non-Null Count  Dtype  
---  ------            --------------  -----  
 0   manufacturer      2530 non-null   object 
 1   company_location  2530 non-null   object 
 2   year_reviewed     2530 non-null   int64  
 3   bean_origin       2530 non-null   object 
 4   bar_name          2530 non-null   object 
 5   cocoa_percent     2530 non-null   float64
 6   num_ingredients   2443 non-null   float64
 7   ingredients       2443 non-null   object 
 8   review            2530 non-null   object 
 9   rating            2530 non-null   float64
dtypes: float64(3), int64(1), object(6)
memory usage: 217.4+ KB

df.describe()

	year_reviewed	cocoa_percent	num_ingredients	rating
count	2530.000000	2530.000000	2443.000000	2530.000000
mean	2014.374308	71.639723	3.041343	3.196344
std	3.968267	5.616724	0.913728	0.445321
min	2006.000000	42.000000	1.000000	1.000000
25%	2012.000000	70.000000	2.000000	3.000000
50%	2015.000000	70.000000	3.000000	3.250000
75%	2018.000000	74.000000	4.000000	3.500000
max	2021.000000	100.000000	6.000000	4.000000

df.isnull().sum()

manufacturer         0
company_location     0
year_reviewed        0
bean_origin          0
bar_name             0
cocoa_percent        0
num_ingredients     87
ingredients         87
review               0
rating               0
dtype: int64

df.duplicated().sum()

0

Missing Data

For the 87 records that have missing information. If ingredients is NaN, then num_ingredients is also NaN. Therefore num_ingredients is MNAR (missing not at random).

df[df.num_ingredients.isnull() & df.ingredients.isnull()]

	manufacturer	company_location	year_reviewed	bean_origin	bar_name	cocoa_percent	num_ingredients	ingredients	review	rating
id
1964	Amazing Cacao	Russia	2017	Peru	Maranon	70.0	NaN	NaN	raisins, pungent	3.00
1964	Amazing Cacao	Russia	2017	Peru	Platan Bajo	70.0	NaN	NaN	few grits, blackcherry, licorice	3.50
486	Artisan du Chocolat	U.K.	2010	Venezuela	Venezuela	100.0	NaN	NaN	very nutty, very bitter	1.75
81	Bonnat	France	2006	Blend	One Hundred	100.0	NaN	NaN	acidic, bitter, dry	1.50
341	Bouga Cacao (Tulicorp)	Ecuador	2009	Ecuador	El Oro, Hacienda de Oro	100.0	NaN	NaN	cardboard, very bitter, floral	1.50
...	...	...	...	...	...	...	...	...	...	...
153	Vintage Plantations (Tulicorp)	U.S.A.	2007	Ecuador	Los Rios, Rancho Grande 2004/2007	100.0	NaN	NaN	bland, mild fruit, strong bitter	2.00
153	Vintage Plantations (Tulicorp)	U.S.A.	2007	Ecuador	Los Rios, Rancho Grande 2004/2007	90.0	NaN	NaN	some fruit, strong bitter	2.00
153	Vintage Plantations (Tulicorp)	U.S.A.	2007	Ecuador	Los Rios, Rancho Grande 2004/2007	75.0	NaN	NaN	cocoa,earthy,very astringent	3.00
733	Whittakers	New Zealand	2011	Ghana	Ghana	72.0	NaN	NaN	sticky, dry, vanilla dominates	2.50
1109	Willie's Cacao	U.K.	2013	Indonesia	Java, Indonesian Black	100.0	NaN	NaN	fruit, strong smoke, bitter	2.25

87 rows × 10 columns

1. What is the average rating by country of origin?

The average rating by country of origin ranges from 2.71 to 3.63.
The country with the highest average rating is Tobago, whereas the lowest is Puerto Rico.

avg_rating_by_origin = (
    df
    .groupby("bean_origin")
    .rating.mean()
    .reset_index()
    .sort_values(by="rating", ascending=False)
)
avg_rating_by_origin

	bean_origin	rating
54	Tobago	3.625000
7	China	3.500000
42	Sao Tome & Principe	3.500000
44	Solomon Islands	3.450000
9	Congo	3.318182
...	...	...
43	Sierra Leone	2.750000
47	St.Vincent-Grenadines	2.750000
38	Principe	2.750000
30	Martinique	2.750000
39	Puerto Rico	2.714286

62 rows × 2 columns

ax = sns.barplot(x="bean_origin", y="rating", data=avg_rating_by_origin[0:8])
show_values_on_bars(ax)
plt.title("Top 8 average rating by country of origin", fontsize=20)
plt.xlabel("Country of Origin", fontsize=16)
plt.ylabel("Average rating", fontsize=16)
plt.show()

2. How many bars were reviewed for each of those countries?

The highest numbers of bars were reviewed for the country Venezulea with 253 bars reviewed.

bar_count_by_origin = (df
                       .groupby("bean_origin")
                       .bar_name.count()
                       .reset_index()
                       .sort_values(by='bar_name', ascending=False)
                       .rename({"bar_name": "no. of bars reviewed"}, axis=1)
                      )
bar_count_by_origin

	bean_origin	no. of bars reviewed
60	Venezuela	253
36	Peru	244
13	Dominican Republic	226
14	Ecuador	219
28	Madagascar	177
...	...	...
50	Suriname	1
49	Sumatra	1
48	Sulawesi	1
47	St.Vincent-Grenadines	1
38	Principe	1

62 rows × 2 columns

ax = sns.barplot(x="bean_origin",y="no. of bars reviewed", data=bar_count_by_origin[0:8])
show_values_on_bars(ax)
plt.title("Number of bars reviewed by country of origin", fontsize=20)
plt.xlabel("Country of Origin", fontsize=16)
plt.ylabel("Number of bars reviewed", fontsize=16)
plt.show()

3. Is the cacao bean’s origin an indicator of quality?

Cocoa bean’s country of origin is not an indicator of quality.

sns.boxplot(y="rating", x="bean_origin", data=df)
plt.xticks(rotation=90)
plt.title("Rating of bean by country of origin", fontsize=20)
plt.xlabel("Country of Origin", fontsize=16)
plt.ylabel("Rating", fontsize=16)
plt.show()

4. How does cocoa content relate to rating?

cocoa_percent and rating is not correlated.

# calculate Pearson's correlation
corr, _ = pearsonr(df.rating, df.cocoa_percent)
print('Pearsons correlation: %.3f' % corr)

sns.scatterplot(y="rating", x="cocoa_percent", data=df)
plt.title("Rating of bean by cocao content", fontsize=20)
plt.xlabel("Cocao Content %", fontsize=16)
plt.ylabel("Rating", fontsize=16)
plt.show()

Pearsons correlation: -0.147

4.5 What is the average cocoa content for bars with higher ratings (above 3.5)?

np.round(df[df['rating']>=3.5].cocoa_percent.mean(), 3)

71.108

5. Compare the average rating of bars with and without lecithin (L in the ingredients).

Some consumers want to avoid bars with lecithin.

The average rating of bars without Lecithin is higher than bars with the ingredient.

df = df.dropna()
contains_L = df["ingredients"].str.contains("L")
avg_rating_with_L = round(df[contains_L].rating.mean(), 3)
avg_rating_without_L = round(df[~contains_L].rating.mean(), 3)
print(f"Average rating with lecithin: {avg_rating_with_L}")
print(f"Average rating without lecithin: {avg_rating_without_L}")

Average rating with lecithin: 3.151
Average rating without lecithin: 3.225