CommerceLens
Data → Customer Behavior → Statistical Description

Descriptive Statistics

Summarizing the shape, center, and spread of session-level behavior and order value across 60,000 synthetic e-commerce sessions before any modeling is attempted.

Business Question
What does a typical order look like?
Metric
Mean vs. median order value; modal category and device
Observation
Mean order value is ₹4,685 while the median is only ₹1,350 — the mean is about 3.5× the median. The most common purchase category is Footwear, and the most common device is Mobile.
Interpretation
Mean substantially exceeding median indicates a right-skewed (long-tailed) order value distribution: most orders are modest, but a minority of large orders pull the average upward. The median is the more representative 'typical order' figure here; the mode describes the single most frequent category/device rather than a central value.
Caveat
Order value here is computed only over purchased sessions (n = orders), not all sessions.

Numeric Variable Explorer

Mean
5.24
Median
4.56
Std. Deviation
2.97
IQR
3.33
Session Duration (min) — Full Summary
n = 60,000 sessions
nMeanMedianStdMinMaxRangeQ1Q2Q3IQRP90P95P99
60,0005.244.562.970.4747.1346.663.184.566.513.338.9910.8315.53

Categorical Frequencies

Customer Segment Distribution
What percentage of customers fall into each behavioral segment?
Income Level Distribution
What percentage of customers fall into each income bracket?
Device Distribution
What device do customers use to shop?

Order Value Spread & Outliers by Segment

Order Value Distribution by Customer Segment
Box shows Q1–Q3 (IQR); whiskers extend to the most extreme non-outlier values; dots are Tukey-fence outliers
₹0₹34,969₹69,938₹1,04,907₹1,39,876₹1,74,844LoyalNewOccasionalRegular
Business Question
What customers lie beyond the IQR-based outlier thresholds?
Metric
Tukey fences: lower = Q1 − 1.5·IQR, upper = Q3 + 1.5·IQR
Observation
For order value, Q1 = ₹644, Q3 = ₹3,759, IQR = ₹3,116, giving an upper fence of ₹8,433 (the lower fence, ₹-4,030, is negative and therefore non-binding since order values cannot be negative). 2,253 orders (13.6%) exceed this upper fence. A sample of the largest outlier values includes ₹1,00,323, ₹1,02,864, ₹1,02,892, ₹1,04,412, ₹1,04,625, ₹1,05,236.
Interpretation
These are statistically extreme orders relative to the bulk of the distribution, not necessarily errors. They likely correspond to bulk purchases, high-value electronics/fashion baskets, or premium customers, and merit separate treatment in AOV or revenue-per-session calculations since they can dominate the mean.
Caveat
Tukey fences are a heuristic convention (1.5×IQR), not a statistical test of anomaly; some of these are legitimate large orders.

Order Value Thresholds

Business Question
What proportion of orders fall below ₹1,000?
Metric
P(order_value < 1000)
Observation
40.0% of orders are below ₹1,000.
Interpretation
A substantial share of purchases are low-ticket, consistent with the median (₹1,350) sitting well under the mean and reinforcing the right-skew story above.
Caveat
This is a simple threshold count on the observed sample, not a forecast of future order mix.
Business Question
What proportion of orders fall above ₹5,000?
Metric
P(order_value > 5000)
Observation
20.6% of orders exceed ₹5,000.
Interpretation
Roughly one in five orders is a high-ticket purchase; these disproportionately drive total revenue even though they are a minority of transactions, which is typical of skewed value distributions in e-commerce.
Caveat
Category or seasonal mix shifts could change this proportion; it is a snapshot of the current sample.
Percentiles, Quartiles & IQR
A percentile tells you what fraction of values fall below a point. The median (50th percentile) splits the data in half. Quartiles split the data into four equal chunks (Q1 = 25th percentile, Q2 = median, Q3 = 75th percentile). The interquartile range (IQR = Q3 − Q1) captures where the 'middle half' of the data sits, ignoring extreme values.
Outlier Detection (Tukey Fences)
Tukey's method flags a value as an outlier if it's 'too far' from the middle 50% of the data — specifically, more than 1.5 times the IQR beyond Q1 or Q3. It's a simple, distribution-free rule of thumb used to spot unusually small or large values worth a closer look.
All figures on this page are computed from a 60,000-row synthetic e-commerce sessions dataset generated for this project. Distribution shapes (e.g. right-skew in order value) are realistic of e-commerce data generally, but exact percentile values, outlier counts, and category/device splits are specific to this sample and should not be read as real market statistics.