A Practical End-to-End R Data Analysis Project

Mini-project overview: from raw data to a shareable report

In this chapter you will run a complete, realistic analysis workflow on a small dataset. The goal is not to learn new syntax, but to integrate skills into a repeatable process: define inputs/outputs, organize files, write reusable functions, validate assumptions, produce plots, and generate a final report grounded in computed results.

Project scenario and questions

Scenario: You manage a small online store and want to understand what drives revenue and whether discounts are helping or hurting profitability.

Dataset: Order-level data with product category, region, dates, quantities, prices, and discounts.

Questions to answer:

• How has revenue changed month-to-month overall and by region?
• Which categories contribute most to revenue and profit?
• Do higher discounts correlate with lower profit margin?
• Are there regions or categories where discounting appears beneficial?
• What are the top 10 products by revenue, and do they have healthy margins?

Scalable project structure (folders, naming, and outputs)

Folder layout

Create a project folder (ideally an RStudio Project) with a predictable structure. This keeps paths stable and makes it easy to rerun or share the work.

Continue in our app.

• Listen to the audio with the screen off.
• Earn a certificate upon completion.
• Over 5000 courses for you to explore!

Or continue reading below...

Download the app

r-mini-project/  data/    raw/    processed/  R/    00_setup.R    01_import_clean.R    02_analysis.R    03_visualize.R  reports/    report.Rmd  outputs/    figures/    tables/  README.md

Naming conventions that scale

• Files: prefix scripts with numbers to enforce run order (00_, 01_, 02_).
• Objects: use snake_case and be explicit: orders_raw, orders_clean, monthly_kpis.
• Columns: keep consistent units and meaning (e.g., discount_pct as 0–1, not 0–100).
• Outputs: save with informative names and dates if needed: monthly_revenue_by_region.png.

Dataset: create a realistic CSV (so everyone can run it)

To make this chapter fully runnable without external downloads, you will generate a dataset and write it to data/raw/orders.csv. In a real project, this step would be replaced by placing your actual raw file in the same folder.

Script: R/00_setup.R

# R/00_setup.R  # Packages used across the project  library(dplyr)  library(ggplot2)  library(readr)  library(lubridate)  library(stringr)  library(tidyr)  # Create folders if missing  dir.create("data/raw", recursive = TRUE, showWarnings = FALSE) dir.create("data/processed", recursive = TRUE, showWarnings = FALSE) dir.create("outputs/figures", recursive = TRUE, showWarnings = FALSE) dir.create("outputs/tables", recursive = TRUE, showWarnings = FALSE)  # Reproducible example dataset generator  set.seed(42)  n <- 1200  products <- tibble::tibble(   product_id = sprintf("P%03d", 1:60),   category = rep(c("Accessories", "Electronics", "Home", "Office"), length.out = 60),   base_price = round(runif(60, 8, 300), 2) )  orders <- tibble::tibble(   order_id = sprintf("O%05d", 1:n),   order_date = sample(seq.Date(as.Date("2024-01-01"), as.Date("2024-12-31"), by = "day"), n, replace = TRUE),   region = sample(c("North", "South", "East", "West"), n, replace = TRUE, prob = c(0.27, 0.23, 0.28, 0.22)),   product_id = sample(products$product_id, n, replace = TRUE),   quantity = sample(1:6, n, replace = TRUE, prob = c(0.35, 0.25, 0.18, 0.12, 0.07, 0.03)),   discount_pct = pmin(pmax(rnorm(n, mean = 0.12, sd = 0.10), 0), 0.45) ) %>%   left_join(products, by = "product_id") %>%   mutate(     unit_price = base_price * runif(n(), 0.95, 1.10),     unit_cost = base_price * runif(n(), 0.55, 0.80),     revenue = quantity * unit_price * (1 - discount_pct),     cost = quantity * unit_cost,     profit = revenue - cost   ) %>%   select(order_id, order_date, region, product_id, category, quantity, unit_price, unit_cost, discount_pct, revenue, cost, profit)  # Introduce a few realistic data issues to practice cleaning  orders_dirty <- orders %>%   mutate(     region = ifelse(runif(n()) < 0.01, "", region),                 # blank region     discount_pct = ifelse(runif(n()) < 0.01, -0.05, discount_pct),       # invalid negative discount     unit_price = ifelse(runif(n()) < 0.01, NA, unit_price),              # missing price     order_date = ifelse(runif(n()) < 0.01, "2024-13-40", as.character(order_date)) # invalid date string   )  write_csv(orders_dirty, "data/raw/orders.csv")

Inputs/outputs defined: input is data/raw/orders.csv; processed data will be saved to data/processed/orders_clean.csv; plots and tables will be exported to outputs/.

Import, validate, clean, and transform (with reusable functions)

Cleaning should be explicit and testable. A good pattern is to write small functions for repeated steps (e.g., checking required columns, standardizing text, validating ranges).

Script: R/01_import_clean.R

# R/01_import_clean.R  source("R/00_setup.R")  # Reusable checks -------------------------------------------------------  assert_has_cols <- function(df, cols) {   missing <- setdiff(cols, names(df))   if (length(missing) > 0) stop("Missing columns: ", paste(missing, collapse = ", "))   df }  clamp <- function(x, lower, upper) pmin(pmax(x, lower), upper)  # Import ----------------------------------------------------------------  orders_raw <- read_csv("data/raw/orders.csv", show_col_types = FALSE)  orders_raw <- assert_has_cols(orders_raw, c(   "order_id", "order_date", "region", "product_id", "category", "quantity",   "unit_price", "unit_cost", "discount_pct", "revenue", "cost", "profit" ))  # Clean -----------------------------------------------------------------  orders_clean <- orders_raw %>%   mutate(     # Parse date safely; invalid strings become NA     order_date = suppressWarnings(lubridate::ymd(order_date)),      # Standardize text fields     region = str_squish(region),     region = na_if(region, ""),     category = str_to_title(str_squish(category)),      # Fix numeric issues     discount_pct = clamp(discount_pct, 0, 0.60),     quantity = as.integer(quantity),     unit_price = as.numeric(unit_price),     unit_cost = as.numeric(unit_cost)   ) %>%   # Remove rows that cannot be analyzed reliably   filter(!is.na(order_date), !is.na(region), !is.na(unit_price), quantity > 0) %>%   # Recompute derived fields from cleaned inputs (avoid trusting dirty totals)   mutate(     revenue = quantity * unit_price * (1 - discount_pct),     cost = quantity * unit_cost,     profit = revenue - cost,     profit_margin = ifelse(revenue > 0, profit / revenue, NA_real_),     month = floor_date(order_date, unit = "month")   )  # Quick data quality summary -------------------------------------------  dq <- orders_raw %>%   summarise(     n_rows = n(),     n_bad_dates = sum(is.na(suppressWarnings(ymd(order_date)))),     n_blank_region = sum(is.na(na_if(str_squish(region), ""))),     n_missing_price = sum(is.na(unit_price)),     n_negative_discount = sum(discount_pct < 0, na.rm = TRUE)   )  write_csv(orders_clean, "data/processed/orders_clean.csv") write_csv(dq, "outputs/tables/data_quality_summary.csv")

Why recompute revenue/profit? In real datasets, totals may be inconsistent with line-item fields. Recomputing ensures your analysis is based on a single, consistent definition.

Analysis plan: define metrics and aggregation levels

Before plotting, decide the metrics and the grain (order-level, product-level, monthly, etc.). Here you will compute KPIs at multiple levels so each question has a clear data source.

Script: R/02_analysis.R

# R/02_analysis.R  source("R/00_setup.R")  orders_clean <- read_csv("data/processed/orders_clean.csv", show_col_types = FALSE)  # Core KPIs -------------------------------------------------------------  overall_kpis <- orders_clean %>%   summarise(     n_orders = n(),     revenue = sum(revenue),     profit = sum(profit),     avg_order_value = mean(revenue),     avg_margin = weighted.mean(profit_margin, w = revenue, na.rm = TRUE)   )  # Monthly revenue and profit by region ---------------------------------  monthly_kpis <- orders_clean %>%   group_by(month, region) %>%   summarise(     orders = n(),     revenue = sum(revenue),     profit = sum(profit),     margin = ifelse(sum(revenue) > 0, sum(profit) / sum(revenue), NA_real_),     .groups = "drop"   )  # Category contribution -------------------------------------------------  category_kpis <- orders_clean %>%   group_by(category) %>%   summarise(     revenue = sum(revenue),     profit = sum(profit),     margin = ifelse(sum(revenue) > 0, sum(profit) / sum(revenue), NA_real_),     .groups = "drop"   ) %>%   arrange(desc(revenue))  # Discount vs margin (order-level) -------------------------------------  discount_margin <- orders_clean %>%   transmute(     discount_pct = discount_pct,     profit_margin = profit_margin,     revenue = revenue,     region = region,     category = category   ) %>%   filter(!is.na(profit_margin), revenue > 0)  # Top products ----------------------------------------------------------  top_products <- orders_clean %>%   group_by(product_id, category) %>%   summarise(     revenue = sum(revenue),     profit = sum(profit),     margin = ifelse(sum(revenue) > 0, sum(profit) / sum(revenue), NA_real_),     orders = n(),     .groups = "drop"   ) %>%   arrange(desc(revenue)) %>%   slice_head(n = 10)  # Export tables ---------------------------------------------------------  write_csv(overall_kpis, "outputs/tables/overall_kpis.csv") write_csv(monthly_kpis, "outputs/tables/monthly_kpis.csv") write_csv(category_kpis, "outputs/tables/category_kpis.csv") write_csv(top_products, "outputs/tables/top_products.csv")

Assumption checks and sanity checks (practical, not theoretical)

Assumption checks are about avoiding misleading results. For this mini-project, focus on checks that commonly break business analyses: extreme outliers, tiny denominators, and unstable margins when revenue is near zero.

Checks to run (and how to act on them)

• Revenue near zero: profit margin becomes unstable. Filter or weight by revenue when summarizing.
• Outliers: a few orders can dominate totals. Inspect high-revenue orders and confirm they are plausible.
• Discount range: ensure discounts are within expected bounds (e.g., 0–60%).
• Group sizes: avoid interpreting patterns for groups with very few orders.

Script snippet: add to R/02_analysis.R (or a separate R/02b_checks.R)

# Sanity checks ----------------------------------------------------------  checks <- list(   revenue_negative = sum(orders_clean$revenue < 0, na.rm = TRUE),   profit_margin_outside = sum(orders_clean$profit_margin < -1 | orders_clean$profit_margin > 1, na.rm = TRUE),   discount_outside = sum(orders_clean$discount_pct < 0 | orders_clean$discount_pct > 0.60, na.rm = TRUE) )  checks_df <- tibble::enframe(checks, name = "check", value = "count") write_csv(checks_df, "outputs/tables/sanity_checks.csv")  # Inspect top 20 orders by revenue (for plausibility review)  top_orders <- orders_clean %>%   arrange(desc(revenue)) %>%   select(order_id, order_date, region, product_id, category, quantity, unit_price, discount_pct, revenue, profit) %>%   slice_head(n = 20)  write_csv(top_orders, "outputs/tables/top_orders_by_revenue.csv")

Visualize key patterns and export plots

Exporting plots is part of the deliverable. Keep figure sizes consistent and use a single theme so the report looks cohesive.

Script: R/03_visualize.R

# R/03_visualize.R  source("R/00_setup.R")  monthly_kpis <- read_csv("outputs/tables/monthly_kpis.csv", show_col_types = FALSE) category_kpis <- read_csv("outputs/tables/category_kpis.csv", show_col_types = FALSE) top_products <- read_csv("outputs/tables/top_products.csv", show_col_types = FALSE) orders_clean <- read_csv("data/processed/orders_clean.csv", show_col_types = FALSE)  theme_set(theme_minimal(base_size = 12))  # 1) Monthly revenue trend by region ------------------------------------  p_monthly <- ggplot(monthly_kpis, aes(x = month, y = revenue, color = region)) +   geom_line(linewidth = 1) +   scale_y_continuous(labels = scales::dollar) +   labs(title = "Monthly revenue by region", x = NULL, y = "Revenue")  ggsave("outputs/figures/monthly_revenue_by_region.png", p_monthly, width = 9, height = 5, dpi = 150)  # 2) Category revenue and profit ----------------------------------------  category_long <- category_kpis %>%   select(category, revenue, profit) %>%   pivot_longer(cols = c(revenue, profit), names_to = "metric", values_to = "value")  p_category <- ggplot(category_long, aes(x = reorder(category, value), y = value, fill = metric)) +   geom_col(position = "dodge") +   coord_flip() +   scale_y_continuous(labels = scales::dollar) +   labs(title = "Revenue and profit by category", x = NULL, y = NULL)  ggsave("outputs/figures/category_revenue_profit.png", p_category, width = 9, height = 5, dpi = 150)  # 3) Discount vs profit margin (weighted view) ---------------------------  # Use point size by revenue to emphasize business impact  p_discount <- ggplot(orders_clean, aes(x = discount_pct, y = profit_margin)) +   geom_hline(yintercept = 0, linewidth = 0.6, color = "grey50") +   geom_point(aes(size = revenue), alpha = 0.25) +   scale_x_continuous(labels = scales::percent) +   scale_y_continuous(labels = scales::percent) +   guides(size = "none") +   labs(title = "Discount vs profit margin (point size = revenue)", x = "Discount", y = "Profit margin")  ggsave("outputs/figures/discount_vs_margin.png", p_discount, width = 9, height = 5, dpi = 150)  # 4) Top products by revenue with margin --------------------------------  p_top <- ggplot(top_products, aes(x = reorder(product_id, revenue), y = revenue)) +   geom_col(fill = "steelblue") +   coord_flip() +   scale_y_continuous(labels = scales::dollar) +   labs(title = "Top 10 products by revenue", x = NULL, y = "Revenue")  ggsave("outputs/figures/top_products_by_revenue.png", p_top, width = 9, height = 5, dpi = 150)

Build the final report: computed results and interpretation prompts

Your report should read like a short decision memo: what you analyzed, what you found, and what the numbers show. Keep interpretation tied to exported tables and plots so it is reproducible.

Report skeleton: reports/report.Rmd (body sections)

```{r setup, include=FALSE} library(readr) library(dplyr) library(ggplot2) library(scales) ```  ```{r load-data} overall_kpis <- read_csv("../outputs/tables/overall_kpis.csv", show_col_types = FALSE) monthly_kpis <- read_csv("../outputs/tables/monthly_kpis.csv", show_col_types = FALSE) category_kpis <- read_csv("../outputs/tables/category_kpis.csv", show_col_types = FALSE) top_products <- read_csv("../outputs/tables/top_products.csv", show_col_types = FALSE) ```  ## Objective and questions - How has revenue changed month-to-month overall and by region? - Which categories contribute most to revenue and profit? - Do higher discounts correlate with lower profit margin? - What are the top products by revenue and their margins?  ## Data quality notes ```{r dq} read_csv("../outputs/tables/data_quality_summary.csv", show_col_types = FALSE) ```  ## Overall KPIs ```{r overall} overall_kpis ```  ## Trends by month and region (see exported figure) ```{r monthly-table} monthly_kpis %>% arrange(month, region) %>% head(12) ```  ## Category contribution ```{r category} category_kpis ```  ## Top products ```{r top-products} top_products ```  ## Interpretation prompts (fill using computed results) - Identify the highest-revenue region-month combinations from `monthly_kpis`. - Compare category margins from `category_kpis` and call out categories with high revenue but low margin. - Use the discount vs margin plot to describe whether margin tends to decrease as discount increases, focusing on high-revenue points. - For the top 10 products, note any high-revenue items with weak margins and quantify the margin from `top_products`.

Tip: Keep the report focused on answering the original questions. If you add extra analysis, label it clearly as exploratory and ensure it still references computed outputs.

Run order: make the project reproducible

A simple run order ensures anyone can reproduce the same outputs from scratch.

• Run R/00_setup.R (creates folders and the example raw CSV).
• Run R/01_import_clean.R (creates processed data and data quality table).
• Run R/02_analysis.R (creates KPI tables).
• Run R/03_visualize.R (exports figures).
• Knit reports/report.Rmd to HTML/PDF (depending on your setup).

Deliverable checklist (what to submit)

• Runnable script set: R/00_setup.R, R/01_import_clean.R, R/02_analysis.R, R/03_visualize.R run in order without manual edits.
• Processed dataset: data/processed/orders_clean.csv.
• Exported tables: outputs/tables/overall_kpis.csv, monthly_kpis.csv, category_kpis.csv, top_products.csv, plus data_quality_summary.csv and sanity_checks.csv.
• Exported plots: PNG files in outputs/figures/ (monthly trend, category comparison, discount vs margin, top products).
• Knitted report: rendered output from reports/report.Rmd that references the exported tables/figures.
• Short interpretation section: 5–10 bullet points in the report that cite specific computed values (e.g., top category margin, lowest-margin region, month with peak revenue) rather than opinions.