CommerceLens
Data → Customer Behavior → Association

Association

Measuring how pairs of variables move together — conditional purchase rates across groups, and correlation between numeric behaviors. Every relationship here is observational: association, not causation.

Group-Level Purchase Rates

Device Type vs. Purchase Rate
P(Purchase | Group)
Business Question
Does the device customers shop on relate to whether they purchase?
Metric
Conditional purchase rate, P(Purchase | Group)
Observation
Desktop converts at 30.4% (n = 14,692) versus Mobile at 26.4% (n = 40,903) — a gap of 4.0%.
Interpretation
Purchase likelihood clearly differs across device type groups in this sample, which is useful for targeting and channel/segment prioritization.
Caveat
This is an observational, cross-sectional association in the dataset — it does not establish that this factor causes purchase behavior. Confounding variables (e.g. customer intent, price, timing) could drive both.
Income Level vs. Purchase Rate
P(Purchase | Group)
Business Question
Does income level relate to purchase conversion?
Metric
Conditional purchase rate, P(Purchase | Group)
Observation
High converts at 34.5% (n = 11,431) versus Low at 20.9% (n = 9,607) — a gap of 13.7%.
Interpretation
Purchase likelihood clearly differs across income level groups in this sample, which is useful for targeting and channel/segment prioritization.
Caveat
This is an observational, cross-sectional association in the dataset — it does not establish that this factor causes purchase behavior. Confounding variables (e.g. customer intent, price, timing) could drive both.
Discount Exposure vs. Purchase Rate
P(Purchase | Group)
Business Question
Are customers exposed to a discount more likely to purchase?
Metric
Conditional purchase rate, P(Purchase | Group)
Observation
True converts at 33.5% (n = 28,062) versus False at 22.3% (n = 31,938) — a gap of 11.3%.
Interpretation
Purchase likelihood clearly differs across discount exposure groups in this sample, which is useful for targeting and channel/segment prioritization.
Caveat
This is an observational, cross-sectional association in the dataset — it does not establish that this factor causes purchase behavior. Confounding variables (e.g. customer intent, price, timing) could drive both.

More Associations with Purchase

Categorical × Numerical: Segment vs. Order Value

Order Value Distribution by Customer Segment
Box plot: numerical variable (order value) split by categorical group (segment)
₹0₹34,969₹69,938₹1,04,907₹1,39,876₹1,74,844LoyalNewOccasionalRegular
Business Question
Does typical order value differ across customer segments?
Metric
Mean order value (AOV) by segment, ₹
Observation
Regular customers have the highest average order value at ₹4,730, versus Loyal at ₹3,944 — a difference of ₹787. The box plot shows the full spread (Q1 –Q3, median, and outliers) behind each of these segment averages.
Interpretation
Segment membership is associated with meaningfully different typical basket sizes, not just different conversion rates — useful for tailoring merchandising or promotions by segment.
Caveat
This is an observational, cross-sectional association in the dataset — it does not establish that this factor causes purchase behavior. Confounding variables (e.g. customer intent, price, timing) could drive both.

Numeric Correlations

Business Question
Do longer sessions tend to view more products?
Metric
Covariance, Pearson correlation coefficient (r), and OLS fitted line
Observation
cov(X, Y) = 2.055, r = 0.355, fitted line y = 0.233·x + 1.27, R² = 0.126.
Interpretation
A positive relationship of moderate strength: as session duration (min) increases, products viewed tends to increase on average, though R² = 0.126 means the linear fit explains only 12.6% of the variance — most of the spread is unexplained by this single variable.
Caveat
This is an observational, cross-sectional association in the dataset — it does not establish that this factor causes purchase behavior. Confounding variables (e.g. customer intent, price, timing) could drive both.
Session Duration vs. Products Viewed
Sampled points with ordinary least-squares fitted line overlay (n = 800 plotted)
Business Question
Do customers who view more products end up placing larger orders?
Metric
Covariance, Pearson correlation coefficient (r), and OLS fitted line
Observation
cov(X, Y) = 217.051, r = 0.009, fitted line y = 47.001·x + 4538.03, R² = 0.000.
Interpretation
A essentially no relationship of negligible strength: as products viewed increases, order value (₹) tends to increase on average, though R² = 0.000 means the linear fit explains only 0.0% of the variance — most of the spread is unexplained by this single variable.
Caveat
This is an observational, cross-sectional association in the dataset — it does not establish that this factor causes purchase behavior. Confounding variables (e.g. customer intent, price, timing) could drive both.
Products Viewed vs. Order Value
Sampled points with ordinary least-squares fitted line overlay (n = 800 plotted)
Business Question
Do customers with more past orders have higher lifetime spend?
Metric
Covariance, Pearson correlation coefficient (r), and OLS fitted line
Observation
cov(X, Y) = 2512.899, r = 0.133, fitted line y = 944.790·x + 1813.85, R² = 0.018.
Interpretation
A positive relationship of weak strength: as previous orders increases, total spend (₹) tends to increase on average, though R² = 0.018 means the linear fit explains only 1.8% of the variance — most of the spread is unexplained by this single variable.
Caveat
This is an observational, cross-sectional association in the dataset — it does not establish that this factor causes purchase behavior. Confounding variables (e.g. customer intent, price, timing) could drive both.
Previous Orders vs. Total Customer Spend
Sampled points with ordinary least-squares fitted line overlay (n = 800 plotted)

Category × Device (Relative Frequency)

Device Share within Each Purchase Category
Row percentages: P(Device | Category)
CategoryDesktopMobileTablet
Beauty & Personal Care25.0%67.7%7.3%
Books & Stationery23.9%68.9%7.2%
Electronics24.5%67.8%7.7%
Fashion23.7%69.0%7.3%
Footwear24.6%68.5%6.9%
Grocery24.9%67.5%7.6%
Home & Kitchen24.7%67.7%7.5%
Mobiles24.4%68.2%7.3%
Sports & Fitness24.5%68.1%7.4%
Toys & Baby24.5%68.3%7.2%

Device share is remarkably stable across categories — Mobile accounts for roughly 68–69% of sessions in every category, with Desktop around 24% and Tablet around 7%. This near-uniformity is itself informative: it suggests device choice is driven by a customer’s general shopping habit rather than the specific product category being browsed, i.e. category and device appear close to independent in this sample.

Correlation
Correlation measures how strongly two numeric variables move together, on a scale from -1 to +1. Close to +1 means they rise together; close to -1 means one rises as the other falls; close to 0 means little to no linear relationship. It never tells you which variable, if either, is causing the other to change.
Every association on this page — categorical group-rate comparisons and numeric correlations alike — is an observational relationship computed on a synthetic sample. None of it should be read as evidence that one variable causes another; confounders, sampling, and the synthetic data-generating process can all produce apparent associations that would not hold, or would hold differently, in a controlled experiment or a different population.