Segmentation

Overview

Customer segmentation involves categorizing your customers into smaller groups based on shared characteristics such as buying habits or demographics. This helps retailers gain deeper insights into their customer base, enabling them to tailor their marketing strategies and products to meet each group's specific needs.

By employing customer segmentation, companies can enhance customer satisfaction, improve targeting accuracy, and optimize resource allocation, ultimately leading to increased profitability and competitiveness in the market.

In this notebook, we'll explore a few basic customer segmentations.

Setup¶

Here we'll load some simulated data to demonstrate the code. You do not need to know this to perform segmentations.

In [ ]:

import matplotlib.pyplot as plt
import pandas as pd

df = pd.read_parquet("../../data/transactions.parquet")
df.head()

import matplotlib.pyplot as plt import pandas as pd df = pd.read_parquet("../../data/transactions.parquet") df.head()

Out[ ]:

	transaction_id	transaction_date	transaction_time	customer_id	product_id	product_name	category_0_name	category_0_id	category_1_name	category_1_id	brand_name	brand_id	unit_quantity	unit_cost	unit_spend	store_id
0	16050	2023-01-12	17:44:29	1	15	Spawn Figure	Toys	1	Action Figures	1	McFarlane Toys	3	2	36.10	55.98	6
1	16050	2023-01-12	17:44:29	1	1317	Gone Girl	Books	8	Mystery & Thrillers	53	Alfred A. Knopf	264	1	6.98	10.49	6
2	20090	2023-02-05	09:31:42	1	509	Ryzen 3 3300X	Electronics	3	Computer Components	21	AMD	102	3	200.61	360.00	4
3	20090	2023-02-05	09:31:42	1	735	Linden Wood Paneled Mirror	Home	5	Home Decor	30	Pottery Barn	147	1	379.83	599.00	4
4	20090	2023-02-05	09:31:42	1	1107	Pro-V Daily Moisture Renewal Conditioner	Beauty	7	Hair Care	45	Pantene	222	1	3.32	4.99	4

Some details about the data

In [ ]:

print(f"Number of unique customers: {df['customer_id'].nunique()}")
print(f"Number of unique transactions: {df['transaction_id'].nunique()}")

print(f"Number of unique customers: {df['customer_id'].nunique()}") print(f"Number of unique transactions: {df['transaction_id'].nunique()}")

Number of unique customers: 4250
Number of unique transactions: 25490

Heavy / Medium / Light¶

Heavy, Medium, Light (HML) is a segmentation that places customers into groups based on the their percentile of spend or the number of products they bought. Heavy customers are the top 20% of customers, medium are the next 30%, and light are the bottom 50% of customers. These values are chosen based on the proportions of the Pareto distribution. Often, purchase behaviour follows this distribution. This is typified by the expression 20% of your customers generate 80% of your sales.

You would use a HML segmentation to answer question such as these,

• How much more are your best customers worth?
• How much more could I spend acquiring my best customers?
• What is the concentration of sales with my top (heavy) customers?

Using PyRetailScience it takes only one line of code to get HML segments.

In [ ]:

from pyretailscience.segmentation.hml import HMLSegmentation

seg = HMLSegmentation(df, zero_value_customers="include_with_light")
seg.df.head()

from pyretailscience.segmentation.hml import HMLSegmentation seg = HMLSegmentation(df, zero_value_customers="include_with_light") seg.df.head()

/home/sys15/Documents/pyretailscience/pyretailscience/segmentation/threshold.py:86: FutureWarning: `case` is deprecated as of v10.0.0, removed in v11.0; use ibis.cases()
  case = ibis.case()

Out[ ]:

	unit_spend	segment_name
customer_id
4221	2.99	Light
1406	3.49	Light
4094	6.98	Light
3499	8.49	Light
3720	8.50	Light

We can then attach those segments back to the original dataframe. You can see these segments to the very right of the dataframe.

In [ ]:

df_with_segments = seg.add_segment(df)
df_with_segments.head()

df_with_segments = seg.add_segment(df) df_with_segments.head()

Out[ ]:

	transaction_id	transaction_date	transaction_time	customer_id	product_id	product_name	category_0_name	category_0_id	category_1_name	category_1_id	brand_name	brand_id	unit_quantity	unit_cost	unit_spend	store_id	segment_name
0	16050	2023-01-12	17:44:29	1	15	Spawn Figure	Toys	1	Action Figures	1	McFarlane Toys	3	2	36.10	55.98	6	Light
1	16050	2023-01-12	17:44:29	1	1317	Gone Girl	Books	8	Mystery & Thrillers	53	Alfred A. Knopf	264	1	6.98	10.49	6	Light
2	20090	2023-02-05	09:31:42	1	509	Ryzen 3 3300X	Electronics	3	Computer Components	21	AMD	102	3	200.61	360.00	4	Light
3	20090	2023-02-05	09:31:42	1	735	Linden Wood Paneled Mirror	Home	5	Home Decor	30	Pottery Barn	147	1	379.83	599.00	4	Light
4	20090	2023-02-05	09:31:42	1	1107	Pro-V Daily Moisture Renewal Conditioner	Beauty	7	Hair Care	45	Pantene	222	1	3.32	4.99	4	Light

Now it's only one line of code to get standard KPIs on each customer segment. This makes segment comparison easy.

In [ ]:

from pyretailscience.segmentation.segstats import SegTransactionStats

seg_stats = SegTransactionStats(df_with_segments, segment_col="segment_name")
seg_stats.df

from pyretailscience.segmentation.segstats import SegTransactionStats seg_stats = SegTransactionStats(df_with_segments, segment_col="segment_name") seg_stats.df

Out[ ]:

	segment_name	spend	transactions	customers	units	spend_per_customer	spend_per_transaction	transactions_per_customer	price_per_unit	units_per_transaction	customers_pct
0	Light	1.344912e+07	7054	2125	56523	6328.995732	1906.594263	3.319529	237.940589	8.012900	0.5
1	Heavy	4.518138e+07	8572	850	80558	53154.567635	5270.809903	10.084706	560.855315	9.397807	0.2
2	Medium	3.045768e+07	9864	1275	86869	23888.378800	3087.761858	7.736471	350.616249	8.806671	0.3
3	Total	8.908818e+07	25490	4250	223950	20961.925033	3495.024770	5.997647	397.803891	8.785798	1.0

We can quickly get a bar plot to compare any of the values using the plot function. With a few extra lines of code, we can dress up the plot with a callout. Now the chart highlights how much more valuable Heavy customers are compared to the Light ones.

In [ ]:

from pyretailscience.style.graph_utils import GraphStyles

ax = seg_stats.plot(
    figsize=(10, 5),
    value_col="spend",
    source_text="Source: Transaction data financial year 2023",
    sort_order="descending",
    title="What's the value of a Heavy customer?",
    rot=0,
)

# Dress up the plot with an arrow calling out the important point
heavy_rev = seg_stats.df[seg_stats.df["segment_name"] == "Heavy"]["spend"].iat[0]
light_rev = seg_stats.df[seg_stats.df["segment_name"] == "Light"]["spend"].iat[0]
h_vs_l_rev = heavy_rev / light_rev

ax.annotate(
    f"The top 20% of customers\ngenerate {h_vs_l_rev:.1f}x MORE REVENUE\nthan the bottom 50%!",
    xy=(0.24, heavy_rev * 0.8),
    xytext=(0.65, heavy_rev * 0.8),
    fontsize=18,
    fontproperties=GraphStyles.POPPINS_REG,
    ha="left",
    va="center",
    arrowprops={
        "facecolor": "white",
        "arrowstyle": "->",
        "connectionstyle": "arc3,rad=0.1",
    },
    color="black",
    bbox={
        "facecolor": "white",
        "edgecolor": "white",
        "boxstyle": "round,rounding_size=0.75",
        "pad": 0.75,
    },
    linespacing=1.5,
)

plt.savefig("monthly_active_customers.svg")
plt.show()

from pyretailscience.style.graph_utils import GraphStyles ax = seg_stats.plot( figsize=(10, 5), value_col="spend", source_text="Source: Transaction data financial year 2023", sort_order="descending", title="What's the value of a Heavy customer?", rot=0, ) # Dress up the plot with an arrow calling out the important point heavy_rev = seg_stats.df[seg_stats.df["segment_name"] == "Heavy"]["spend"].iat[0] light_rev = seg_stats.df[seg_stats.df["segment_name"] == "Light"]["spend"].iat[0] h_vs_l_rev = heavy_rev / light_rev ax.annotate( f"The top 20% of customers\ngenerate {h_vs_l_rev:.1f}x MORE REVENUE\nthan the bottom 50%!", xy=(0.24, heavy_rev * 0.8), xytext=(0.65, heavy_rev * 0.8), fontsize=18, fontproperties=GraphStyles.POPPINS_REG, ha="left", va="center", arrowprops={ "facecolor": "white", "arrowstyle": "->", "connectionstyle": "arc3,rad=0.1", }, color="black", bbox={ "facecolor": "white", "edgecolor": "white", "boxstyle": "round,rounding_size=0.75", "pad": 0.75, }, linespacing=1.5, ) plt.savefig("monthly_active_customers.svg") plt.show()

Getting a plot of the segment's activity over time is also now a snap.

In [ ]:

from pyretailscience.plots.time import plot

df["transaction_date"] = pd.to_datetime(df["transaction_date"])

df_with_groups = seg.add_segment(df)
plot(
    df_with_groups,
    figsize=(10, 5),
    period="M",
    group_col="segment_name",
    value_col="customer_id",
    agg_func="nunique",
    title="Monthly Active Customers",
    y_label="Number of Customers",
    source_text="Source: PyRetailScience",
)
plt.show()

from pyretailscience.plots.time import plot df["transaction_date"] = pd.to_datetime(df["transaction_date"]) df_with_groups = seg.add_segment(df) plot( df_with_groups, figsize=(10, 5), period="M", group_col="segment_name", value_col="customer_id", agg_func="nunique", title="Monthly Active Customers", y_label="Number of Customers", source_text="Source: PyRetailScience", ) plt.show()

Activating¶

Once you have your segments, you will often want to activate them. It is easy to export them to a CSV for activation in your CRM or with platforms.

In [ ]:

filename = "heavy_customers.csv"
seg.df[seg.df["segment_name"] == "Heavy"].index.to_series().to_csv(filename, index=False)

# Display the CSV
with open(filename) as f:
    print("".join(f.readlines()[:5]))

filename = "heavy_customers.csv" seg.df[seg.df["segment_name"] == "Heavy"].index.to_series().to_csv(filename, index=False) # Display the CSV with open(filename) as f: print("".join(f.readlines()[:5]))

customer_id
1566
925
2095
1623

Coming soon¶

• Quickly segmenting new vs returning customers
• More charting options
• Activating segments to Facebook and other social platforms
• ... and more