Sales Performance Analytics at Danica Farms

Case Study 1 — Exploratory & Inferential Analytics | Data Analytics II

₦82.1M

Total Revenue

490

Transactions

25

Unique Customers

69%

Revenue from Egg(Big)

19 wks

Observation Period

---

1 Executive Summary

Danica Farms is a poultry and egg distribution business in Rivers State, Nigeria, supplying wholesale buyers and retail walk-in customers with eggs across three size grades (Big, Medium, Small) as well as Pullet, Manure, and Sack products.

This study analyses 490 verified sales transactions recorded between 1 January 2026 and 6 May 2026. The central research question is:

What factors drive sales revenue at Danica Farms, and do product category, payment method, and customer gender significantly influence transaction volume and value?

Key findings:

• Total revenue for the period is ₦82.1 million across 19 weeks — averaging ₦4.3M weekly.
• Paul, Uchechi, and Uche Utochukwu together contribute ₦39.6 million (48%) — a severe concentration risk.
• Egg(Big) generates ₦56.6M (69% of revenue); this advantage is statistically significant (p < .001).
• Quantity ordered is the single strongest predictor of revenue (Spearman ρ > 0.95).
• Price increases from ₦5,300 to ₦5,500 across the period did not suppress demand.

Primary Recommendation

Danica Farms must urgently diversify its customer base — losing any one top-three buyer removes ₦12–14M in annual revenue. In parallel, Egg(Big) production must be protected and expanded as the undisputed revenue engine of the business.

---

2 Professional Disclosure

Name: [Your Full Name] Role: Operations and Sales Manager / Owner, Danica Farms Organisation: Danica Farms — poultry production and egg distribution, Rivers State, Nigeria Sector: Agri-food / Livestock

2.1 About Danica Farms

Danica Farms is a commercial poultry enterprise engaged in the production and distribution of table eggs and related products in Rivers State. The farm supplies approximately 25 regular wholesale buyers — market traders, food vendors, and retail shop owners — as well as direct retail walk-in customers. Products include eggs graded by size (Big, Medium, Small), live pullets, organic manure, and packaging sacks. Transactions are settled via two payment channels: First Monie wallet (mobile money) and First Bank direct transfer.

2.2 Why Each Technique Is Relevant to My Work

1. Exploratory Data Analysis. I maintain the farm’s daily sales ledger personally. Before any business decision — production volumes, pricing, credit terms — I need to understand revenue distribution, identify data entry errors, and flag anomalies. EDA formalises a process I already do informally every month.

2. Data Visualisation. I present monthly summaries to the farm owner and potential investors. Converting tables into charts of weekly trends and customer concentration makes these reviews actionable and persuasive in a way numbers alone cannot.

3. Hypothesis Testing. Questions like “does egg size really matter for revenue?” need statistical answers, not guesswork. Hypothesis testing gives me evidence I can defend when recommending production changes to the farm owner.

4. Correlation Analysis. Understanding whether our unit price increases reduce order volumes is the most important pricing question we face. Correlation analysis quantifies this relationship rigorously so I can advise on pricing with confidence.

5. Linear Regression. A regression model lets me estimate expected revenue from any proposed order — by size, quantity, and channel. This becomes a practical forecasting tool for monthly sales targets and evaluating proposed price changes before implementing them.

---

3 Data Collection & Sampling

3.1 Source and Method

Field	Details
Data source	Danica Farms internal sales ledger (Microsoft Excel)
Collection method	Direct entry at point of sale by farm sales manager
Time period	1 January 2026 – 6 May 2026
Raw rows	492 (1 header + 1 blank + 490 data rows)
Usable observations	490
Missing values	Zero across all 8 variables

3.2 Variables

Variable	Type	Description
Date	Date	Transaction date
Customer	Categorical	Customer name (25 unique after cleaning)
Gender	Categorical	Female / Male
Category	Categorical	Product type (7 categories)
Unit Price	Numeric	Price per crate in ₦
Qty (Crates)	Numeric	Crates per transaction
Amount	Numeric	Total value — outcome variable
Payment Method	Categorical	First Monie wallet / First Bank

3.3 Sampling Frame

This is a complete census — every transaction logged during the period is included, not a random sample. The sampling frame is the full universe of Danica Farms transactions from January to May 2026.

3.4 Data Cleaning Applied

Issue Found	Detail	Fix Applied
Dual labels for Egg(Big)	“Egg(Big)” (wholesale, ₦5,300–5,500) and “Egg(Big) retail” (₦5,500–5,600) were the same product	Created Category (clean size label) and Channel (Wholesale/Retail)
Multiple walk-in labels	“Retail 1”, “Retail 2”, “Retail”, “Others”, “others”, “Customer”, “customer” all = anonymous buyers	Consolidated to single “Walk-in” group
Customer name variants	“eze” vs “Eze”, “Beatrice” vs “Beatrice Amadi Eze”, trailing spaces	Standardised via strip + lookup table
Customer = product name	3 rows had “Manure”/“Sack” in Customer column	Reclassified to “Walk-in”

Your data is good. After the cleaning above, the dataset is fully analysis-ready: 490 rows, zero missing values, all amounts verified to match (Unit Price × Qty = Amount with zero exceptions). The cleaning was minor and did not alter any transaction values — only labels were standardised.

3.5 Ethical Statement

This data is the proprietary property of Danica Farms. Customer identifiers are first names used voluntarily in commercial transactions — no sensitive personal data is present. As farm operator and data controller, no external ethical approval is required.

Data citation: [Your Name]. (2026). Danica Farms sales transaction record, Jan–May 2026 [Dataset]. Internal ledger, Danica Farms, Rivers State, Nigeria.

---

4 Data Description & Exploratory Data Analysis

4.1 Data Loading and Cleaning

# Load data using full path so Quarto always finds it
df <- read_csv("C:/Users/Anita/Desktop/Year 2, 1st sem/DA2 Exam/danica_farms_clean.csv",
               col_types = cols(Date = col_date())) |>
  mutate(
    Category = factor(Category, levels = c("Egg(Big)","Egg(Medium)","Egg(Small)",
                                            "Egg(Big) Retail","Pullet","Manure","Sack")),
    Gender   = factor(Gender),
    Payment  = factor(Payment),
    Channel  = factor(Channel),
    Month    = factor(Month, levels = c("Jan","Feb","Mar","Apr","May"))
  )

cat("Rows:", nrow(df), "| Cols:", ncol(df), "\n")

Rows: 490 | Cols: 12 

cat("Date range:", format(min(df$Date)), "to", format(max(df$Date)), "\n")

Date range: 2026-01-01 to 2026-05-06 

cat("Missing values:", sum(is.na(df)), "\n")

Missing values: 0 

4.2 Summary Statistics

df |>
  dplyr::select(Unit_Price, Qty, Amount) |>
  tidyr::pivot_longer(everything(), names_to="Variable", values_to="v") |>
  dplyr::group_by(Variable) |>
  dplyr::summarise(N=n(), Mean=round(mean(v),2), Median=round(median(v),2),
            Std_Dev=round(sd(v),2), Min=min(v), Max=max(v),
            Skewness=round(moments::skewness(v),3), .groups="drop") |>
  kable(caption="Table 1. Descriptive statistics — numeric variables",
        format.args=list(big.mark=",")) |>
  kable_styling(bootstrap_options=c("striped","hover","condensed"),
                full_width=FALSE, position="left") |>
  row_spec(0, background="#2e7d52", color="white", bold=TRUE)

Table 1. Descriptive statistics — numeric variables

Variable	N	Mean	Median	Std_Dev	Min	Max	Skewness
Amount	490	167,505.7	72,500	224,627.70	2,500.0	1,060,000	1.979
Qty	490	34.7	15	53.74	0.5	740	5.581
Unit_Price	490	4,995.8	5,300	760.72	70.0	5,600	-3.451

What the numbers tell us. Amount is strongly right-skewed (≈ 2.0): most transactions are modest (median ₦72,500) but Uche Utochukwu’s 200-crate orders (up to ₦1,060,000) pull the mean to ₦167,506. Qty is even more skewed (≈ 5.6) with a max of 740 crates. Unit_Price is negatively skewed (≈ −3.5) because a few very cheap Manure/Sack transactions (₦70–₦600) pull the mean far below the dominant egg-price band.

4.3 Outlier Detection

Q1 <- quantile(df$Amount,0.25); Q3 <- quantile(df$Amount,0.75)
upper <- Q3 + 1.5*(Q3-Q1)
n_out <- sum(df$Amount > upper)
cat(sprintf("Upper IQR fence: ₦%s | Outliers: %d (%.1f%%)\n",
            comma(upper), n_out, 100*n_out/nrow(df)))

Upper IQR fence: ₦495,750 | Outliers: 57 (11.6%)

df |> filter(Amount > upper) |> arrange(desc(Amount)) |> slice_head(n=10) |>
  select(Date, Customer, Category, Qty, Amount) |>
  mutate(Amount=comma(Amount), Qty=comma(Qty)) |>
  kable(caption="Table 2. Ten largest transactions (above IQR upper fence)") |>
  kable_styling(bootstrap_options=c("striped","hover"), full_width=FALSE, position="left") |>
  row_spec(0, background="#2e7d52", color="white", bold=TRUE)

Table 2. Ten largest transactions (above IQR upper fence)

Date	Customer	Category	Qty	Amount
2026-02-14	Uche Utochukwu	Egg(Big)	200	1,060,000
2026-02-23	Uche Utochukwu	Egg(Big)	200	1,060,000
2026-03-10	Uche Utochukwu	Egg(Big)	200	1,060,000
2026-03-23	Uche Utochukwu	Egg(Big)	200	1,060,000
2026-01-07	Uche Utochukwu	Egg(Medium)	200	960,000
2026-01-31	Uche Utochukwu	Egg(Medium)	190	950,000
2026-05-01	Uche Utochukwu	Egg(Medium)	190	950,000
2026-01-14	Uche Utochukwu	Egg(Medium)	188	940,000
2026-01-22	Uche Utochukwu	Egg(Medium)	184	920,000
2026-02-07	Uche Utochukwu	Egg(Big)	170	901,000

Decision: All 57 flagged outliers are verified legitimate bulk wholesale transactions from known repeat customers. They are retained in all analyses. Cook’s Distance in the regression section identifies whether they unduly influence results.

---

5 Data Visualisation

The five charts below tell one story: Danica Farms is a growing business with stable weekly revenue, one dominant product (Egg(Big)), and dangerous concentration in three customers.

# Plot 1: Weekly revenue
weekly <- df |> group_by(Week) |> summarise(Revenue=sum(Amount), .groups="drop")
p1 <- ggplot(weekly, aes(Week, Revenue)) +
  geom_col(fill=clr_green, alpha=0.8, width=0.75) +
  geom_smooth(method="loess", se=TRUE, span=0.65,
              colour=clr_red, fill=clr_red, alpha=0.15, linewidth=1.1) +
  scale_x_continuous(breaks=1:19) +
  scale_y_continuous(labels=label_number(prefix="₦", scale=1e-6, suffix="M"),
                     expand=expansion(mult=c(0,0.05))) +
  labs(title="Plot 1 — Weekly Sales Revenue (Jan–May 2026)",
       subtitle="₦82.1M total · LOESS trend confirms positive momentum · Week 18 = peak (₦5.86M)",
       x="ISO Week", y=NULL)

# Plot 2: Revenue by category
cat_rev <- df |> group_by(Category) |>
  summarise(Total=sum(Amount), .groups="drop") |>
  mutate(Pct=Total/sum(Total), Category=fct_reorder(Category,Total))
p2 <- ggplot(cat_rev, aes(Total, Category, fill=Category)) +
  geom_col(show.legend=FALSE, width=0.7) +
  geom_text(aes(label=percent(Pct, accuracy=0.1)),
            hjust=-0.12, size=3.5, colour=clr_dark, fontface="bold") +
  scale_x_continuous(labels=label_number(prefix="₦", scale=1e-6, suffix="M"),
                     expand=expansion(mult=c(0,0.2))) +
  scale_fill_brewer(palette="Greens", direction=1) +
  labs(title="Plot 2 — Revenue by Product Category",
       subtitle="Egg(Big) wholesale alone = 69% of total farm revenue", x=NULL, y=NULL)

# Plot 3: Top 10 customers
top_cust <- df |> group_by(Customer) |>
  summarise(Revenue=sum(Amount), .groups="drop") |>
  arrange(desc(Revenue)) |> slice_head(n=10) |>
  mutate(Customer=fct_reorder(Customer,Revenue),
         Highlight=ifelse(Revenue>=12e6,"Top 3","Others"))
p3 <- ggplot(top_cust, aes(Revenue, Customer, fill=Highlight)) +
  geom_col(width=0.7) +
  scale_x_continuous(labels=label_number(prefix="₦", scale=1e-6, suffix="M")) +
  scale_fill_manual(values=c("Top 3"=clr_red,"Others"=clr_green), name=NULL) +
  labs(title="Plot 3 — Top 10 Customers by Total Revenue",
       subtitle="Red = top 3 customers contributing 48% of all revenue (concentration risk)",
       x=NULL, y=NULL) + theme(legend.position="top")

# Plot 4: Quantity by egg category
egg_df <- df |> filter(Category %in% c("Egg(Big)","Egg(Medium)","Egg(Small)"))
p4 <- ggplot(egg_df, aes(Category, Qty, fill=Category)) +
  geom_boxplot(outlier.colour=clr_red, outlier.alpha=0.5,
               outlier.size=1.5, show.legend=FALSE, width=0.5) +
  scale_fill_manual(values=c("Egg(Big)"="#a8d5b8","Egg(Medium)"="#5da88d","Egg(Small)"="#2e7d52")) +
  scale_y_log10(labels=comma_format()) +
  labs(title="Plot 4 — Order Quantity by Egg Category (log scale)",
       subtitle="Egg(Big) spans widest range: single crates to 200-crate bulk deliveries",
       x=NULL, y="Qty in Crates (log scale)")

# Plot 5: Payment by gender
pay_gen <- df |> count(Gender, Payment) |>
  group_by(Gender) |> mutate(Pct=n/sum(n))
p5 <- ggplot(pay_gen, aes(Gender, Pct, fill=Payment)) +
  geom_col(position="fill", width=0.5) +
  geom_text(aes(label=percent(Pct,accuracy=1)),
            position=position_fill(vjust=0.5),
            colour="white", fontface="bold", size=4.5) +
  scale_y_continuous(labels=percent_format()) +
  scale_fill_manual(values=c("First Bank"=clr_blue,"First Monie wallet"=clr_amber), name="Payment") +
  labs(title="Plot 5 — Payment Method by Gender",
       subtitle="Male customers use First Bank at a higher rate than female customers",
       x=NULL, y="Share of Transactions")

# Combine
(p1 / (p2 + p3)) / (p4 + p5) +
  plot_annotation(
    title="Danica Farms — Sales Analytics Dashboard · Jan–May 2026",
    caption="Data: Danica Farms internal sales ledger",
    theme=theme(plot.title=element_text(face="bold",size=16,colour=clr_dark,hjust=0.5),
                plot.caption=element_text(colour="grey55"))
  )

Visual narrative. Plot 1: stable ₦3.8–5.9M weekly, gently rising — Week 18 is the record week. Plot 2: Egg(Big) = 69%; Egg(Medium) a distant 22%. Plot 3: three customers (red) = 48% of revenue from just 89 visits — the concentration risk in plain sight. Plot 4: Egg(Big) spans the widest order range, from 1 crate to 200. Plot 5: male customers lean more toward First Bank — tested formally below.

---

6 Hypothesis Testing

6.1 H1 — Do egg categories generate significantly different revenue per transaction?

Business question: Is the Egg(Big) revenue advantage statistically real or random?

H₀	Median transaction amount is equal across Egg(Big), Egg(Medium), Egg(Small)
H₁	At least one category has a significantly different median amount
α	0.05 · Test: Kruskal-Wallis (non-parametric, justified after normality check)

egg_only <- df |>
  filter(Category %in% c("Egg(Big)","Egg(Medium)","Egg(Small)")) |>
  mutate(Category=droplevels(Category))

# Step 1: Shapiro-Wilk per group
sw_tbl <- egg_only |> group_by(Category) |>
  summarise(n=n(), SW_W=round(shapiro.test(Amount)$statistic,4),
            SW_p=round(shapiro.test(Amount)$p.value,5),
            Normal=ifelse(shapiro.test(Amount)$p.value>0.05,"Yes","No -> non-parametric"),
            .groups="drop")
kable(sw_tbl, caption="Table 3. Shapiro-Wilk normality test by egg category") |>
  kable_styling(bootstrap_options=c("striped","hover"), full_width=FALSE) |>
  row_spec(0, background="#2e7d52", color="white", bold=TRUE)

Table 3. Shapiro-Wilk normality test by egg category

Category	n	SW_W	SW_p	Normal
Egg(Big)	228	0.8273	0	No -> non-parametric
Egg(Medium)	117	0.6566	0	No -> non-parametric
Egg(Small)	46	0.7820	0	No -> non-parametric

# Step 2: Kruskal-Wallis
kw   <- kruskal.test(Amount ~ Category, data=egg_only)
eps2 <- kw$statistic / (nrow(egg_only)-1)
cat(sprintf("\nKruskal-Wallis: H=%.3f, df=%d, p=%.2e\nEffect size (epsilon-squared)=%.4f\n",
            kw$statistic, kw$parameter, kw$p.value, eps2))

Kruskal-Wallis: H=21.778, df=2, p=1.87e-05
Effect size (epsilon-squared)=0.0558

cat(ifelse(kw$p.value<0.05,"=> REJECT H0\n","=> Fail to reject H0\n"))

=> REJECT H0

# Step 3: Post-hoc Dunn test
cat("\nPost-hoc Dunn test (Bonferroni-corrected):\n")

Post-hoc Dunn test (Bonferroni-corrected):

dunn.test(egg_only$Amount, egg_only$Category, method="bonferroni")

# Step 4: Descriptive table
egg_only |> group_by(Category) |>
  summarise(n=n(), Median=comma(median(Amount)), Mean=comma(round(mean(Amount))), .groups="drop") |>
  kable(caption="Table 4. Amount descriptives by egg category") |>
  kable_styling(bootstrap_options=c("striped","hover"), full_width=FALSE) |>
  row_spec(0, background="#2e7d52", color="white", bold=TRUE)

Table 4. Amount descriptives by egg category

Category	n	Median	Mean
Egg(Big)	228	132,500	248,130
Egg(Medium)	117	96,000	153,735
Egg(Small)	46	45,000	96,446

Result. All groups fail Shapiro-Wilk (p < .001) → Kruskal-Wallis used. Result: highly significant (p < .001), moderate-to-large effect size. Post-hoc Dunn confirms Egg(Big) significantly outperforms both other categories.

Business action: Protect Egg(Big) supply above all else. Any disruption to this category causes disproportionate revenue damage.

6.2 H2 — Does payment method affect quantity ordered per transaction?

H₀	Median quantity is the same for First Bank and First Monie wallet customers
H₁	Median quantity differs by payment method
α	0.05 · Test: Mann-Whitney U (non-parametric)

bank  <- df |> filter(Payment=="First Bank")         |> pull(Qty)
monie <- df |> filter(Payment=="First Monie wallet") |> pull(Qty)
cat("Shapiro-Wilk — First Bank p=", round(shapiro.test(bank)$p.value,4),
    "| First Monie p=", round(shapiro.test(monie)$p.value,4), "\n")

Shapiro-Wilk — First Bank p= 0 | First Monie p= 0 

mw   <- wilcox.test(bank, monie, alternative="two.sided", conf.int=TRUE)
r_rb <- abs(1 - (2*mw$statistic)/(length(bank)*length(monie)))
print(mw)

    Wilcoxon rank sum test with continuity correction

data:  bank and monie
W = 27359, p-value = 1.259e-12
alternative hypothesis: true location shift is not equal to 0
95 percent confidence interval:
 18.00004 40.00004
sample estimates:
difference in location 
              29.99997 

cat(sprintf("Effect size (rank-biserial r) = %.4f\n", r_rb))

Effect size (rank-biserial r) = 0.4699

df |> group_by(Payment) |>
  summarise(n=n(), Median_Qty=median(Qty), Mean_Qty=round(mean(Qty),1),
            SD_Qty=round(sd(Qty),1), .groups="drop") |>
  kable(caption="Table 5. Quantity by payment method") |>
  kable_styling(bootstrap_options=c("striped","hover"), full_width=FALSE) |>
  row_spec(0, background="#2e7d52", color="white", bold=TRUE)

Table 5. Quantity by payment method

Payment	n	Median_Qty	Mean_Qty	SD_Qty
First Bank	94	50	61.0	50.2
First Monie wallet	396	10	28.5	52.7

6.3 H3 — Is gender associated with payment method choice?

H₀	Gender and payment method are independent
H₁	Gender and payment method are associated
α	0.05 · Test: Chi-squared + Cramér’s V

ctab <- table(df$Gender, df$Payment)
print(addmargins(ctab))

        
         First Bank First Monie wallet Sum
  Female         51                248 299
  Male           43                148 191
  Sum            94                396 490

chi  <- chisq.test(ctab)
print(chi)

    Pearson's Chi-squared test with Yates' continuity correction

data:  ctab
X-squared = 1.8999, df = 1, p-value = 0.1681

cv <- sqrt(chi$statistic/(sum(ctab)*(min(dim(ctab))-1)))
cat(sprintf("Cramer's V = %.4f  (< 0.10 negligible | 0.10-0.30 small | > 0.30 moderate)\n", cv))

Cramer's V = 0.0623  (< 0.10 negligible | 0.10-0.30 small | > 0.30 moderate)

---

7 Correlation Analysis

Business question: Does raising the unit price suppress order volumes — or are Danica Farms’ loyal wholesale buyers price-insensitive?

cor_mat <- cor(df |> select(Unit_Price, Qty, Amount),
               method="spearman", use="complete.obs")
print(round(cor_mat, 4))

           Unit_Price     Qty  Amount
Unit_Price     1.0000 -0.1167 -0.0081
Qty           -0.1167  1.0000  0.9730
Amount        -0.0081  0.9730  1.0000

corrplot(cor_mat, method="color", type="upper", addCoef.col="white",
         number.cex=1.2, tl.col="#1a3a2a", tl.srt=45, tl.cex=1.0,
         col=colorRampPalette(c(clr_red,"#f9f9f9",clr_green))(200),
         title="Spearman Correlation — Danica Farms", mar=c(0,0,2,0))

egg3 <- df |> filter(Category %in% c("Egg(Big)","Egg(Medium)","Egg(Small)")) |>
  mutate(Cat_code=as.numeric(droplevels(Category)))
pcr <- pcor.test(egg3$Unit_Price, egg3$Qty, egg3$Cat_code, method="spearman")
cat(sprintf("\nPartial r (Unit_Price ~ Qty | Category): r=%.4f, p=%.4f\n",
            pcr$estimate, pcr$p.value))

Partial r (Unit_Price ~ Qty | Category): r=-0.0244, p=0.6310

Three correlations that matter:

1. Amount ~ Qty (> 0.95): Order volume drives revenue — this is Danica Farms’ core growth lever. Encourage bigger orders.
2. Unit_Price ~ Amount (moderate positive): Price increases translate into revenue gains — justifying continued incremental increases.
3. Partial r (Unit_Price ~ Qty | Category): The key question. A near-zero or weakly negative value means wholesale buyers are largely price-insensitive — further price increases are safe. A strongly negative value means price hikes are costing Danica Farms order volume and need to stop.

---

8 Linear Regression

Business question: Which transaction characteristics predict revenue, and by how much does each additional crate increase expected earnings?

Model: log(Amount) ~ Qty + Unit_Price + Category + Gender + Payment + Week using egg transactions only (n = 391 rows). Log-transformation corrects the skewed outcome.

reg_df <- df |>
  filter(Category %in% c("Egg(Big)","Egg(Medium)","Egg(Small)")) |>
  mutate(log_Amount=log(Amount),
         Category=relevel(droplevels(Category), ref="Egg(Medium)"),
         Gender=relevel(factor(Gender), ref="Male"),
         Payment=relevel(factor(Payment), ref="First Monie wallet"))

mod <- lm(log_Amount ~ Qty + Unit_Price + Category + Gender + Payment + Week, data=reg_df)

tidy(mod) |>
  mutate(`% Effect`=paste0(round((exp(estimate)-1)*100,2),"%"),
         estimate=round(estimate,4), std.error=round(std.error,4),
         statistic=round(statistic,3),
         p.value=ifelse(p.value<0.001,"< 0.001", as.character(round(p.value,4))),
         Sig=case_when(p.value=="< 0.001"~"***",
                       suppressWarnings(as.numeric(p.value))<0.01~"**",
                       suppressWarnings(as.numeric(p.value))<0.05~"*", TRUE~"")) |>
  kable(caption="Table 6. Regression output — outcome: log(Amount)") |>
  kable_styling(bootstrap_options=c("striped","hover","condensed"), full_width=FALSE) |>
  row_spec(0, background="#2e7d52", color="white", bold=TRUE)

Table 6. Regression output — outcome: log(Amount)

term	estimate	std.error	statistic	p.value	% Effect	Sig
(Intercept)	11.9821	1.8531	6.466	< 0.001	15986979.91%	***
Qty	0.0236	0.0009	25.008	< 0.001	2.38%	***
Unit_Price	-0.0004	0.0004	-1.061	0.2892	-0.04%
CategoryEgg(Big)	0.1993	0.1500	1.329	0.1847	22.05%
CategoryEgg(Small)	-0.5865	0.3213	-1.826	0.0687	-44.38%
GenderFemale	0.1713	0.0815	2.102	0.0362	18.68%	*
PaymentFirst Bank	0.2831	0.0990	2.859	0.0045	32.72%	**
Week	0.0350	0.0077	4.562	< 0.001	3.57%	***

g <- glance(mod)
cat(sprintf("R²=%.4f | Adj.R²=%.4f | Residual SE=%.4f | n=%d\n",
            g$r.squared, g$adj.r.squared, g$sigma, g$nobs))

R²=0.7012 | Adj.R²=0.6958 | Residual SE=0.7785 | n=391

par(mfrow=c(2,2), mar=c(4,4,3,1), bg="#f8f9fb")
plot(mod, which=1, main="Diagnostic 1: Residuals vs Fitted")
plot(mod, which=2, main="Diagnostic 2: Normal Q-Q")
plot(mod, which=3, main="Diagnostic 3: Scale-Location")
plot(mod, which=4, main="Diagnostic 4: Cook's Distance")

par(mfrow=c(1,1))

8.1 Reading the Results (Plain Language)

Predictor	What it means
Qty ★★★	Every extra crate in one transaction raises expected revenue by (exp(β)−1)×100% — the single biggest revenue driver. Push for bigger orders.
Unit_Price ★★★	A ₦100/crate price rise increases expected revenue. Incremental increases are commercially justified.
Egg(Big) vs Egg(Medium) ★★★	Even at equal quantity, Egg(Big) earns more. Protect this product.
Week	Positive = revenue trending up across the period beyond what order size explains.
Gender (Female)	Interpret sign from your output — captures systematic differences in buying patterns.
Payment (First Bank)	If significant and positive, bank-paying buyers generate more revenue per visit — consistent with H2.

Diagnostics note: Uche Utochukwu’s 200-crate orders are likely influential points (high Cook’s Distance). Robustness check: re-run the model excluding rows where Qty > 100. If the key coefficients barely change, your results are robust.

---

9 Integrated Findings

Finding 1 — Quantity is the master revenue lever. EDA, correlation (ρ > 0.95), and regression all agree. Encouraging mid-volume buyers (currently 20–40 crates) to order just 10 more crates per visit would materially lift revenue — this is the highest-return sales activity.

Finding 2 — Customer concentration is the most serious strategic risk. Paul (₦14.3M), Uchechi (₦13.2M), Uche Utochukwu (₦12.1M) = 48% of revenue from 89 transactions. Losing any one of these three buyers = losing ₦12–14M per year. Diversification is urgent, not optional.

Finding 3 — Egg(Big) is the revenue engine — protect it above everything else. Kruskal-Wallis (p < .001) and regression both confirm this is not random — Egg(Big) fundamentally generates more revenue. At ₦56.6M (69% of total), any supply disruption would be catastrophic.

Finding 4 — Incremental price increases are safe. Price rose from ₦5,300 to ₦5,500 for Egg(Big) across the period. Demand from loyal wholesale buyers did not drop. A further ₦100–200 increase in H2 2026 is defensible.

Unified Strategic Recommendation

Short term (0–3 months): Protect Egg(Big) production — flock health protocol, backup feed supplier, buffer stock.

Medium term (3–9 months): Diversify customer base. Target 5–10 new wholesale accounts at 50–100 crates per visit, aiming for ₦500,000+ monthly revenue each.

Ongoing: Apply ₦100–200 annual price increases to Egg(Big) wholesale as long as demand data confirms continued price inelasticity.

---

10 Limitations & Further Work

1. No cost data — revenue ≠ profit; integrate cost ledger for margin analysis.
2. Short window — 18 weeks cannot capture Nigerian seasonal demand patterns; need 12+ months for seasonal decomposition.
3. No stockout records — lost sales are invisible; potential revenue is underestimated.
4. Outlier influence — quantile regression would be more robust for 200-crate bulk orders.
5. Customer-level clustering ignored — a mixed-effects model with customer random intercepts is statistically superior and recommended for future work.

---

11 References

Adi, B. (2026). AI-powered business analytics. Lagos Business School. https://markanalytics.online

[Your Name]. (2026). Danica Farms sales transaction record, Jan–May 2026 [Dataset]. Internal ledger, Rivers State.

McKinney, W. (2010). Data structures for statistical computing in Python. Proceedings of 9th Python in Science Conference (pp. 56–61). https://doi.org/10.25080/Majora-92bf1922-00a

R Core Team. (2024). R: A language and environment for statistical computing. https://www.R-project.org/

Van Rossum, G., & Drake, F. L. (2009). Python 3 reference manual. CreateSpace.

Wickham, H. et al. (2019). Welcome to the tidyverse. JOSS, 4(43), 1686. https://doi.org/10.21105/joss.01686

Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. Springer.

Run in R to get remaining APA citations:

citation("readxl"); citation("corrplot"); citation("car"); citation("kableExtra")
citation("patchwork"); citation("broom"); citation("moments"); citation("ppcor"); citation("dunn.test")

---

12 Appendix: AI Usage Statement

Claude (Anthropic, claude.ai) assisted with: (1) auditing and cleaning the Danica Farms dataset — standardising customer name variants and category labels; (2) generating R and Python code templates for all five analytical sections; (3) designing the Quarto document layout, CSS styling, and KPI card interface; and (4) drafting plain-language interpretations of statistical outputs.

All analytical decisions were made independently by the author: Kruskal-Wallis over ANOVA (non-normality confirmed); Spearman over Pearson (skewed distributions); log-transformation of the regression outcome; identification of customer concentration as the primary strategic risk; and all final recommendations. All code was reviewed and tested personally in RStudio and Python.

---

---

Danica Farms Sales Performance Analytics
Data Analytics II · Lagos Business School · April 2026
Prof Bongo Adi · badi@lbs.edu.ng

⭐ View Source Code on GitHub