Data Wrangling Final Project - Regork Grocery Chain

Introduction

Objective

As a national grocery chain, Regork must continuously identify new areas of growth to maintain competitive advantage and drive profitability. This analysis was commissioned to identify a potential area of growth where the company could invest future resources to increase revenue and profits.

We have decided to take the following approach to find growth opportunities for Regork Grocery Chain:

1. Which days of the week have the lowest sales performance?
2. What are the daily and hourly traffic patterns?.
3. Which customer segments (family sizes) are underrepresented on weak days?
4. Are there specific day/hour combinations that underperform?

Our goal is to identify specific days, times, and customer segments where targeted promotional investments can generate the highest return on revenue.

Solution Summary

Primary Finding: We identified two consistently underperforming days of the week that represent a significant revenue opportunity.

Three-Tiered Approach:

• Tier 1 Day-Level Targeting : Target Wednesday and Tuesday with day-specific promotions and exclusive loyalty rewards.
• Tier 2 Family Size Targeting: Identify underrepresented family sizes and create family-pack promotions and bulk discounts.
• Tier 3 Time-Based Targeting : Target specific underperforminghour/day combinations during prime hours

Expected Outcome: By bringing weak days up to average performance, we can capture substantial annual revenue growth with focused, low-risk promotional investments.

Business Impact

Quantified Opportunity:

Based on our analysis of one full year of transaction data from 2,700 households, we have identified a clear path to revenue growth:

• Annual Revenue Potential: Significant six-figure opportunity by optimizing weak day performance
• Marketing Efficiency: High - targeted campaigns to specific family segments rather than broad advertising
• Implementation Risk: Low - pilot testing approach before full rollout
• Time to Results: 4-8 weeks for measurable impact

Strategic Value:

This approach provides a blueprint for systematic revenue optimization that can be replicated across all store locations and expanded to other identified opportunities.

---

Packages Required

The following packages are required in order to run the code efficiently without any errors.

library(completejourney)  # dataset source
library(dplyr)            # for dynamic data manipulation
library(tidyr)            # for data tidying operations
library(lubridate)        # for easy date and time manipulation
library(ggplot2)          # for creating visualizations
library(scales)           # for formatting plot scales and labels
library(knitr)            # for dynamic report generation
library(kableExtra)       # for enhanced table styling

# Set global theme for consistent visualizations
theme_set(theme_minimal(base_size = 13))

Package Descriptions:

• completejourney: Provides access to transaction and demographic data
• dplyr: Data manipulation (filtering, grouping, summarizing)
• tidyr: Data reshaping and pivoting
• lubridate: Date/time operations and extraction
• ggplot2: Create visualizations and charts
• scales: Format axes and labels in charts
• knitr/kableExtra: Create professional tables

---

---

Data Preparation

Data Tables and Fields

This analysis uses two primary datasets from the CompleteJourney package:

Transactions Dataset (transactions_sample)

Variable	Data Type	Description
household_id	character	Uniquely identifies each household
basket_id	character	Uniquely identifies each purchase occasion
product_id	character	Uniquely identifies each product
quantity	numeric	Number of products purchased
sales_value	numeric	Amount in dollars from the sale
transaction_timestamp	datetime	Date and time of transaction

Demographics Dataset

Variable	Data Type	Description
household_id	character	Links to transactions dataset
age	character	Age group of household head
income	character	Income bracket
household_size	numeric	Number of people in household (1-5+)

Loading and Preparing Transaction Data

# Load datasets
transactions <- transactions_sample
demographics <- demographics

# Define business parameters
STORE_OPEN_HOUR <- 6   # 6 AM
STORE_CLOSE_HOUR <- 23  # 11 PM

cat("**Dataset Overview:**\n\n")

Dataset Overview:

cat("- Transaction Records:", format(nrow(transactions), big.mark = ","), "\n")

• Transaction Records: 75,000

cat("- Unique Households:", format(n_distinct(transactions$household_id), big.mark = ","), "\n")

• Unique Households: 2,377

cat("- Date Range:", as.character(min(transactions$transaction_timestamp)), "to", 
    as.character(max(transactions$transaction_timestamp)), "\n")

• Date Range: 2017-01-01 07:30:27 to 2017-12-31 22:47:38

cat("- Business Hours: 6:00 AM - 11:00 PM Daily\n")

• Business Hours: 6:00 AM - 11:00 PM Daily

Data Transformation

We transformed the raw transaction data by extracting temporal features and creating meaningful time segments for analysis.

# Prepare transaction data with time features
transactions_clean <- transactions %>%
  mutate(
    # Extract date/time components
    transaction_date = as.Date(transaction_timestamp),
    day_of_week = wday(transaction_timestamp, label = TRUE, abbr = FALSE),
    day_of_week_num = wday(transaction_timestamp),
    hour_of_day = hour(transaction_timestamp),
    month = month(transaction_timestamp, label = TRUE),
    
    # Create time periods
    time_period = case_when(
      hour_of_day >= 6 & hour_of_day < 9 ~ "Early Morning (6-9 AM)",
      hour_of_day >= 9 & hour_of_day < 12 ~ "Mid Morning (9-12 PM)",
      hour_of_day >= 12 & hour_of_day < 15 ~ "Early Afternoon (12-3 PM)",
      hour_of_day >= 15 & hour_of_day < 18 ~ "Late Afternoon (3-6 PM)",
      hour_of_day >= 18 & hour_of_day < 21 ~ "Evening (6-9 PM)",
      hour_of_day >= 21 & hour_of_day < 24 ~ "Late Evening (9-11 PM)",
      TRUE ~ "Closed Hours"
    ),
    
    # Weekend vs Weekday
    day_type = ifelse(day_of_week %in% c("Saturday", "Sunday"), "Weekend", "Weekday")
  ) %>%
  # Filter to business hours only
  filter(hour_of_day >= STORE_OPEN_HOUR & hour_of_day <= STORE_CLOSE_HOUR)

cat("**Data prepared:**", format(nrow(transactions_clean), big.mark = ","), 
    "transactions during business hours\n\n")

Data prepared: 73,813 transactions during business hours

cat("**Coverage:**", round(nrow(transactions_clean)/nrow(transactions)*100, 1), 
    "% of total transactions\n")

Coverage: 98.4 % of total transactions

---

Exploratory Data Analysis

Analysis 1: Weakest Days

Question: Which specific days need promotional investment?

# Calculate daily metrics
day_analysis <- transactions_clean %>%
  group_by(day_of_week, day_of_week_num) %>%
  summarise(
    total_transactions = n(),
    unique_households = n_distinct(household_id),
    total_sales = sum(sales_value, na.rm = TRUE),
    avg_basket_size = mean(sales_value, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  arrange(day_of_week_num) %>%
  mutate(
    pct_transactions = total_transactions / sum(total_transactions) * 100,
    pct_sales = total_sales / sum(total_sales) * 100
  )

# Identify weak days
weakest_day <- day_analysis %>% arrange(total_sales) %>% slice(1)
second_weakest <- day_analysis %>% arrange(total_sales) %>% slice(2)

avg_daily_sales <- mean(day_analysis$total_sales)
weak_gap <- avg_daily_sales - weakest_day$total_sales
second_gap <- avg_daily_sales - second_weakest$total_sales
total_gap <- weak_gap + second_gap

weak_day_data <- day_analysis %>%
  mutate(
    is_weak = day_of_week %in% c(as.character(weakest_day$day_of_week), 
                                   as.character(second_weakest$day_of_week)),
    status = ifelse(is_weak, "Target for Growth", "Current Performance")
  )

ggplot(weak_day_data, aes(x = day_of_week, y = total_sales/1000, fill = status)) +
  geom_col(width = 0.7) +
  geom_hline(yintercept = avg_daily_sales/1000, linetype = "dashed", 
             color = "#D32F2F", size = 1.2) +
  geom_text(aes(label = paste0("$", round(total_sales/1000, 1), "K")),
            vjust = -0.5, size = 4.5, fontface = "bold") +
  annotate("segment", x = 2.5, xend = 4.5, 
           y = avg_daily_sales/1000, yend = avg_daily_sales/1000,
           linetype = "dashed", color = "#D32F2F", size = 1.2) +
  annotate("text", x = 3.3, y = avg_daily_sales/1000 + 2.5, 
           label = "Daily Average", color = "#D32F2F", hjust = 1, 
           fontface = "bold", size = 4.5) +
  scale_y_continuous(labels = function(x) paste0("$", x, "K"),
                     expand = expansion(mult = c(0, 0.18)),
                     limits = c(0, max(weak_day_data$total_sales/1000) * 1.15)) +
  scale_fill_manual(values = c("Target for Growth" = "#E74C3C", 
                               "Current Performance" = "#95A5A6")) +
  labs(
    title = "Daily Revenue Performance Across the Week",
    x = "Day of Week",
    y = "Total Sales (in thousands)",
    fill = ""
  ) +
  theme_minimal(base_size = 14) +
  theme(
    plot.title = element_text(face = "bold", size = 18, color = "#2C3E50"),
    axis.text.x = element_text(angle = 45, hjust = 1, size = 12),
    legend.position = "top",
    legend.text = element_text(size = 12),
    plot.margin = margin(10, 10, 10, 10)
  )

Data Interpretation:

Based on the above graph, We can conclude that Wednesday and Tuesday consistently fall below the daily average revenue benchmark.On Wednesday current Sales is $28.5K/week lower than daily avg sales by $4.9k while for Tuesday current sales is $28.5k/week lower than daily avg sales by $4.4k/week.

These two days represent $9,357 in weekly unrealized revenue potential. When extrapolated across 52 weeks, this gap translates to an annual opportunity of $486,551, representing approximately 211% potential increase in total annual revenue.

A point of observation to be noted here is that the underperformance is consistent rather than sporadic, indicating a systematic pattern rather than random variation. This suggests that targeted interventions can reliably capture this opportunity.

Analysis 2: Traffic Patterns

Question: How does revenue vary across the week?

# Prepare data for dual visualization
daily_plot_data <- day_analysis %>%
  mutate(
    revenue_scaled = total_sales / 1000,
    traffic_scaled = total_transactions
  )

# Calculate scaling factor for secondary axis
scale_factor <- max(daily_plot_data$revenue_scaled) / max(daily_plot_data$traffic_scaled)

# Color palette
revenue_color <- "#4A90E2"
traffic_color <- "#E74C3C"

# Create dual-axis chart
ggplot(daily_plot_data, aes(x = day_of_week)) +
  geom_col(aes(y = revenue_scaled), fill = revenue_color, alpha = 0.8, width = 0.65) +
  geom_text(aes(y = revenue_scaled / 2, label = paste0("$", round(revenue_scaled, 1), "K")),
            size = 4, fontface = "bold", color = "white") +
  geom_line(aes(y = traffic_scaled * scale_factor, group = 1), 
            color = traffic_color, size = 2) +
  geom_point(aes(y = traffic_scaled * scale_factor), 
             color = traffic_color, size = 5, shape = 19) +
  geom_text(aes(y = traffic_scaled * scale_factor, 
                label = format(traffic_scaled, big.mark = ",")),
            vjust = -2.2, size = 3.8, fontface = "bold", color = traffic_color) +
  scale_y_continuous(
    name = "Revenue (Thousands)",
    labels = function(x) paste0("$", x, "K"),
    expand = expansion(mult = c(0, 0.3)),
    sec.axis = sec_axis(
      ~./scale_factor, 
      name = "Number of Shopping Trips",
      labels = comma
    )
  ) +
  labs(
    title = "Daily Performance: Revenue vs Customer Traffic",
    subtitle = "BLUE BARS = Revenue Earned ($) | RED LINE = Customer Shopping Trips",
    x = "Day of Week",
    caption = "Note: Both metrics displayed on separate scales. Bars show total revenue; line shows transaction count."
  ) +
  theme_minimal(base_size = 14) +
  theme(
    plot.title = element_text(face = "bold", size = 18, hjust = 0.5),
    plot.subtitle = element_text(size = 12, color = "gray30", hjust = 0.5, margin = margin(b = 20)),
    plot.caption = element_text(size = 11, color = "gray30", hjust = 0, margin = margin(t = 10)),
    axis.title.y.left = element_text(color = revenue_color, face = "bold", size = 12),
    axis.title.y.right = element_text(color = traffic_color, face = "bold", size = 12),
    axis.text.y.left = element_text(color = revenue_color, face = "bold"),
    axis.text.y.right = element_text(color = traffic_color, face = "bold"),
    axis.text.x = element_text(angle = 45, hjust = 1, size = 11, face = "bold"),
    panel.grid.major.x = element_blank()
  )

Data Interpretation:

This visualization reveals a strong correlation between customer traffic (red line) and revenue generation (blue bars). The parallel movement of both metrics indicates that our revenue challenges on weak days are primarily driven by lower foot traffic rather than reduced spending per customer.

For these weak days, we further deep-dive and find opportunities for growth by analyzing their spend patterns based on factors like family size, time of day, and shopping frequency. The consistent gap between peak days (Saturday, Sunday) and weak days (Wednesday, Tuesday) represents a 49% difference in customer visits - a substantial opportunity for traffic-driving promotions.

Analysis 3: Hourly Traffic Patterns & Time-Based Opportunities

Question: Which exact day/hour combinations should we target with promotions? Part A: Shopping Traffic Heatmap

day_hour_analysis <- transactions_clean %>%
  group_by(day_of_week, day_of_week_num, hour_of_day) %>%
  summarise(
    total_transactions = n(),
    total_sales = sum(sales_value, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  arrange(day_of_week_num)

avg_hourly <- mean(day_hour_analysis$total_transactions)

ggplot(day_hour_analysis, aes(x = hour_of_day, y = day_of_week, fill = total_transactions)) +
  geom_tile(color = "white", size = 1) +
  geom_text(aes(label = comma(total_transactions, accuracy = 1)), 
            size = 3.5, color = "black", fontface = "bold") +
  scale_fill_gradient2(
    low = "#E74C3C", 
    mid = "#F39C12",
    high = "#27AE60", 
    midpoint = avg_hourly,
    labels = comma,
    name = "Transactions"
  ) +
  scale_x_continuous(breaks = seq(6, 23, 1)) +
  labs(
    title = "Shopping Traffic Intensity: Day × Hour Analysis",
    x = "Hour of Day",
    y = "Day of Week"
  ) +
  theme_minimal(base_size = 13) +
  theme(
    plot.title = element_text(face = "bold", size = 16),
    legend.position = "right",
    panel.grid = element_blank(),
    axis.text = element_text(size = 11)
  )

Part B: Specific Time-Based Growth Opportunities

# Calculate hour benchmarks
hour_benchmarks <- day_hour_analysis %>%
  group_by(hour_of_day) %>%
  summarise(
    hour_avg_transactions = mean(total_transactions),
    hour_avg_sales = mean(total_sales),
    .groups = "drop"
  )

# Find opportunities
growth_opps <- day_hour_analysis %>%
  left_join(hour_benchmarks, by = "hour_of_day") %>%
  mutate(
    pct_of_hour_avg = (total_transactions / hour_avg_transactions) * 100,
    revenue_gap = hour_avg_sales - total_sales,
    is_opportunity = hour_of_day >= 9 & hour_of_day <= 21 & pct_of_hour_avg < 85
  ) %>%
  filter(is_opportunity) %>%
  arrange(pct_of_hour_avg)

if(nrow(growth_opps) > 0) {
  
  print(
    kable(
      growth_opps %>%
        head(15) %>%
        mutate(
          time_label = paste0(sprintf("%02d", hour_of_day), ":00"),
          day_hour = paste(day_of_week, time_label)
        ) %>%
        select(
          `Day & Hour` = day_hour,
          `Current Traffic` = total_transactions,
          `Hour Average` = hour_avg_transactions,
          `% of Average` = pct_of_hour_avg,
          `Revenue Gap` = revenue_gap
        ) %>%
        mutate(
          `Hour Average` = round(`Hour Average`, 0),
          `% of Average` = paste0(round(`% of Average`, 1), "%"),
          `Revenue Gap` = dollar(`Revenue Gap`)
        ),
      format.args = list(big.mark = ","),
      caption = "Top 15 Underperforming Day/Hour Combinations (Prime Hours 9 AM - 9 PM)"
    ) %>%
      kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
                    full_width = FALSE) %>%
      row_spec(1:5, bold = TRUE, background = "#F8D7DA")
  )
  
} else {
  cat("\n**No significant time-based opportunities identified.**\n\n")
  cat("Traffic is well-distributed across prime shopping hours (9 AM - 9 PM).\n\n")
  cat("**Strategic Recommendation:** Focus resources on day-level and family-size targeting instead, as these offer greater impact.\n")
}

Top 15 Underperforming Day/Hour Combinations (Prime Hours 9 AM - 9 PM)

Day & Hour	Current Traffic	Hour Average	% of Average	Revenue Gap
Thursday 10:00	330	471	70%	$501.09
Wednesday 13:00	595	823	72.3%	$676.53
Wednesday 10:00	341	471	72.3%	$227.33
Thursday 11:00	505	677	74.6%	$395.38
Thursday 12:00	578	773	74.7%	$607.60
Thursday 14:00	634	839	75.5%	$649.20
Wednesday 09:00	206	272	75.8%	$230.36
Tuesday 14:00	637	839	75.9%	$888.67
Tuesday 12:00	588	773	76%	$625.65
Tuesday 13:00	635	823	77.2%	$692.30
Wednesday 12:00	599	773	77.5%	$674.47
Wednesday 11:00	525	677	77.6%	$507.12
Wednesday 14:00	657	839	78.3%	$434.41
Tuesday 15:00	696	874	79.7%	$731.32
Friday 11:00	540	677	79.8%	$397.27

Data Interpretation:

Based on the above heatmap, we identify shopping intensity patterns across 126 unique day-hour combinations (7 days × 18 hours). The color gradient from red (low traffic) to green (high traffic) reveals that weak days show consistently lower activity across most operating hours.

The table chart helps us to further drill down and identify Top 15 Underperfroming time slots,Prioritize targeted promotions for late-morning to mid-afternoon promotions rather than broad all-day campaigns.

Analysis 4: Family Size Gap

Question: Which family sizes are MISSING on our weakest days?

# Prepare family data
weak_day_trans <- transactions_clean %>%
  filter(day_of_week %in% c(as.character(weakest_day$day_of_week),
                             as.character(second_weakest$day_of_week))) %>%
  left_join(demographics, by = "household_id") %>%
  filter(!is.na(household_size))

all_trans_demo <- transactions_clean %>%
  left_join(demographics, by = "household_id") %>%
  filter(!is.na(household_size))

# Calculate metrics
weak_family <- weak_day_trans %>%
  group_by(household_size) %>%
  summarise(
    transactions = n(),
    total_sales = sum(sales_value, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  mutate(weak_day_pct = transactions / sum(transactions) * 100)

overall_family <- all_trans_demo %>%
  group_by(household_size) %>%
  summarise(transactions = n(), .groups = "drop") %>%
  mutate(overall_pct = transactions / sum(transactions) * 100)

family_comparison <- weak_family %>%
  left_join(overall_family %>% select(household_size, overall_pct),
            by = "household_size") %>%
  mutate(
    gap = weak_day_pct - overall_pct,
    gap_label = paste0(ifelse(gap > 0, "+", ""), round(gap, 1), " pp"),
    status = ifelse(gap < 0, "Missing on Weak Days", "Already Shopping Well")
  ) %>%
  arrange(gap)

most_underrep <- family_comparison %>% filter(gap < 0) %>% slice(1)

ggplot(family_comparison, aes(x = reorder(factor(household_size), gap), y = gap)) +
  geom_col(aes(fill = gap < 0), width = 0.6, show.legend = FALSE) +
  geom_hline(yintercept = 0, size = 1.5, color = "black") +
  geom_text(aes(label = gap_label, hjust = ifelse(gap < 0, 1.0 , -0.3)),
            size = 3.5, fontface = "bold") +
  coord_flip() +
  scale_fill_manual(values = c("TRUE" = "#E74C3C", "FALSE" = "#27AE60")) +
  scale_y_continuous(
    labels = function(x) paste0(x, " pp"),
    breaks = seq(-10, 10, 2)
  ) +
  labs(
    title = "Customer Segment Analysis: Family Size Representation on Weak Days",
    x = "Household Size (Number of People)",
    y = "Gap in Shopping Frequency (percentage points)"
  ) +
  theme_minimal(base_size = 14) +
  theme(
    plot.title = element_text(face = "bold", size = 17, color = "#2C3E50"),
    panel.grid.major.y = element_blank(),
    axis.text = element_text(size = 12, face = "bold"),
    axis.title = element_text(size = 13, face = "bold")
  )

Data Interpretation:

Based on the above graph, we identified customer profiles based on their shopping patterns across weak days versus the overall weekly average. Negative values (red bars) indicate family sizes that shop significantly less on Wednesday and Tuesday compared to their normal behavior.

All the other customers with positive values (green bars) already shop at their typical frequency on these days. For these family size profiles, we further deep-dive and find opportunities for growth by creating targeted promotional bundles and messaging that address their specific household needs.

We observed that family size 3 shows the most significant underrepresentation (0.8 percentage points below normal). These are existing Regork customers who actively avoid weak days - making them prime candidates for targeted “family pack” promotions and personalized incentives to shift their shopping behavior.

---

Summary

Objective and Data Source

The primary objective of the project is to find growth opportunities for Regork Grocery Chain by analyzing customer shopping patterns, identifying weak days, and discovering underrepresented customer segments for the year 2017.

The completejourney dataset is available in R under the package of the same name, and provides access to datasets characterizing household level transactions over one year from a group of 2,467 households who are frequent shoppers at Regork.

Identified Growth Opportunities

1. Two Weakest Days

cat("- **Target Days:** ", as.character(weakest_day$day_of_week), " and ", 
    as.character(second_weakest$day_of_week), sep = "")

• Target Days: Wednesday and Tuesday

cat("- **Combined revenue gap:** $", format(round(total_gap, 0), big.mark = ","), 
    " per week", sep = "")

• Combined revenue gap: $9,357 per week

cat("- **Annual opportunity:** $", format(round(total_gap * 52, 0), big.mark = ","),  sep = "")

• Annual opportunity: $486,551

cat("- **Potential revenue increase:** ", 
    round(total_gap * 52 / sum(transactions_clean$sales_value) * 100, 1), sep = "")

• Potential revenue increase: 211

2. Family Size Targeting

if(exists("most_underrep") && nrow(most_underrep) > 0) {
  underrep_families <- family_comparison %>% filter(gap < -1.5)
  cat("- **Most underrepresented:** Family size ", most_underrep$household_size, "\n", sep = "")
  if(nrow(underrep_families) > 1) {
    cat("- **Total underrepresented segments:** ", nrow(underrep_families), 
        " family sizes identified\n", sep = "")
  }
  cat("- **Action:** Create family-pack promotions and bulk discounts\n")
}

• Most underrepresented: Family size 3
• Action: Create family-pack promotions and bulk discounts

3. Time-Based Opportunities

if(exists("growth_opps") && nrow(growth_opps) > 0) {
  top_15_opps <- growth_opps %>% head(15)
  cat("- **Underperforming time slots:** Top 15 day/hour combinations identified for targeted promotions\n")
  cat("- **Action:** Launch flash sales or limited-time offers during these specific windows\n")
  cat("- **Combined potential:** $", 
      format(round(sum(top_15_opps$revenue_gap), 0), big.mark = ","), 
      " weekly if these top 15 slots reach their hour average\n", sep = "")
} else {
  cat("- **Time distribution:** Well-balanced across prime hours\n")
  cat("- **Focus:** Day-level and family-size strategies will have greater impact\n")
}

• Underperforming time slots: Top 15 day/hour combinations identified for targeted promotions
• Action: Launch flash sales or limited-time offers during these specific windows
• Combined potential: $8,239 weekly if these top 15 slots reach their hour average

---

Strategic Recommendations

Recommendation 1: Launch Day-Specific Loyalty Program

Create “Wednesday & Tuesday Rewards” program:

• Double loyalty points on these two days
• Exclusive deals available only on weak days
• Email/app notifications highlighting special offers
• Expected impact: 40-50% closure of revenue gap

Recommendation 2: Family-Targeted Promotions

if(exists("most_underrep") && nrow(most_underrep) > 0) {
  underrep_sizes <- family_comparison %>% filter(gap < -1.5) %>% pull(household_size)
  cat("Focus on underrepresented family sizes: **", paste(underrep_sizes, collapse = ", "), "**\n\n", sep = "")
  cat("**Specific Tactics:**\n\n")
  cat("- **Targeted Campaigns** - e.g. Family of ", most_underrep$household_size, 
      "? Get 20% off our weekly dinner box!\n", sep = "")
  cat("- **Exclusive deal on 'Family Bundles'** - Create ", most_underrep$household_size, 
      "-pack bundles (meals, snacks, beverages)\n", sep = "")
  cat("- **Bulk buy promotions** - 'Buy 3 Get 1 Free' on staples and household essentials\n")
  cat("- **Highlight value packs** - Emphasize nutritional value and family-friendly items\n")
  cat("- **Partner with family-oriented apps** for targeted digital advertising\n")
} else {
  cat("All family sizes well-represented. Focus on frequency and basket size.\n")
}

Focus on underrepresented family sizes: ****

Specific Tactics:

• Targeted Campaigns - e.g. Family of 3? Get 20% off our weekly dinner box!
• Exclusive deal on ‘Family Bundles’ - Create 3-pack bundles (meals, snacks, beverages)
• Bulk buy promotions - ‘Buy 3 Get 1 Free’ on staples and household essentials
• Highlight value packs - Emphasize nutritional value and family-friendly items
• Partner with family-oriented apps for targeted digital advertising

Recommendation 3: Promotional Calendar

Create a structured approach:

• Week 1: Launch pilot “Weak Day Rewards” with double points
• Week 2-4: A/B test different family promotions (discount % vs. bundle deals)
• Month 2: Scale successful tactics; add time-specific flash sales if needed
• Month 3: Measure results and refine strategy

Recommendation 4: Marketing Channel Strategy

• Email campaigns: Target families by household size with personalized offers
• Mobile app push notifications: Day-specific deals on Tuesday/Wednesday mornings
• In-store signage: Highlight “Family Pack Deals” on weak days
• Social media: “Weak Day Wednesday” and “Tuesday Treats” campaigns

---

Conclusion

Primary Growth Lever: Target the two weakest days (Wednesday and Tuesday) with day-specific promotions ($486,551 annual potential)

Secondary Growth Lever: Create family-size targeted promotions to attract underrepresented household segments on weak days

We have identified top 15 high-priority day/hour combinations, successful targeted campaigns could potentially increase the annual revenue by $428,411

Why This Will Work:

1. Focused: Two specific days vs. broad-based promotions = efficient marketing spend
2. Targeted: Family-size segmentation = relevant offers that resonate
3. Measurable: Clear baseline and KPIs for tracking success
4. Low-risk: Pilot approach allows testing before full rollout
5. Scalable: Success model can extend to other opportunities

Success Metrics:

• Week-over-week traffic increase on target days
• Family size distribution shift on weak days
• Revenue per day trending toward average
• Customer satisfaction scores during promotional periods

---

Analysis Date: 10 October, 2025
Data Source: CompleteJourney Dataset - 2,700 Households, Full Year Analysis
Business Hours Analyzed: 6:00 AM - 11:00 PM Daily