Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/nph1508/sql_for_ecommerce_analyzing_sales_customer_behavior_in_bigquery

Designed and executed complex SQL queries on an ecommerce dataset using Google BigQuery to uncover customer behavior patterns, sales performance, and category-level insights. Focused on extracting business value through data exploration and aggregation techniques.
https://github.com/nph1508/sql_for_ecommerce_analyzing_sales_customer_behavior_in_bigquery

analysis bigquery sql

Last synced: 12 months ago
JSON representation

Designed and executed complex SQL queries on an ecommerce dataset using Google BigQuery to uncover customer behavior patterns, sales performance, and category-level insights. Focused on extracting business value through data exploration and aggregation techniques.

Awesome Lists containing this project

README 

          
# SQL for Ecommerce: Analyzing Sales & Customer Behavior in BigQuery

Author: Nguyễn Phương Huy



Date: 2000-15-08



Tools Used: SQL

## 📑 Table of Contents

1. [📌 Background & Overview](#-background--overview)

2. [📂 Dataset Description & Data Structure](#-dataset-description--data-structure)

3. [🔎 Final Conclusion & Recommendations](#-final-conclusion--recommendations)

## 📌 Background & Overview

🎯 Objective

📖 This project uses SQL (BigQuery) to analyze an ecommerce dataset in order to:



✔️ Uncover trends in sales performance across product categories and time periods



✔️ Analyze customer behavior patterns, including frequency and purchase volume



✔️ Identify top-performing products and underperforming segments



✔️ Provide data-driven insights to support inventory, marketing, and sales decisions



💡 Main Business Questions:



What are the best-selling product categories over time?



Which customers contribute most to total revenue?



Are there seasonal patterns in customer purchasing behavior?



How can we segment customers based on their purchase activity?



👤 Who is this project for?



✔️ Data Analysts seeking to practice SQL in a real-world ecommerce context



✔️ Business Analysts / Ecommerce Teams needing insights to optimize operations and marketing



✔️ Decision-Makers who want to understand customer dynamics and product performance



## 📂 Dataset Description & Data Structure

### 📌 Data Source



- **Source:** [Google Analytics Sample Dataset](https://console.cloud.google.com/marketplace/product/bigquery-public-data/google-analytics-sample)  

- **Size:** ~400,000 rows × 15+ columns  

- **Format:** BigQuery table (`.sql`)



### 📊 Data Structure & Relationships**



1️⃣ Tables Used:

  📋 Click to view 



| Field Name | Data Type | Description |

| --- | --- | --- |

| fullVisitorId | STRING | The unique visitor ID. |

| date | STRING | The date of the session in YYYYMMDD format. |

| totals | RECORD | This section contains aggregate values across the session. |

| totals.bounces | INTEGER | Total bounces (for convenience). For a bounced session, the value is 1, otherwise it is null. |

| totals.hits | INTEGER | Total number of hits within the session. |

| totals.pageviews | INTEGER | Total number of pageviews within the session. |

| totals.visits | INTEGER | The number of sessions (for convenience). This value is 1 for sessions with interaction events. The value is null if there are no interaction events in the session. |

| totals.transactions | INTEGER | Total number of ecommerce transactions within the session. |

| trafficSource.source | STRING | The source of the traffic source. Could be the name of the search engine, the referring hostname, or a value of the utm_source URL parameter. |

| hits | RECORD | This row and nested fields are populated for any and all types of hits. |

| hits.eCommerceAction | RECORD | This section contains all of the ecommerce hits that occurred during the session. This is a repeated field and has an entry for each hit that was collected. |

| hits.eCommerceAction.action_type | STRING | The action type. Click through of product lists = 1, Product detail views = 2, Add product(s) to cart = 3, Remove product(s) from cart = 4, Check out = 5, Completed purchase = 6, Refund of purchase = 7, Checkout options = 8, Unknown = 0.
Usually this action type applies to all the products in a hit, with the following exception: when hits.product.isImpression = TRUE, the corresponding product is a product impression that is seen while the product action is taking place (i.e., a "product in list view").
Example query to calculate number of products in list views:
SELECT
COUNT(hits.product.v2ProductName)
FROM [foo-160803:123456789.ga_sessions_20170101]
WHERE hits.product.isImpression == TRUE
Example query to calculate number of products in detailed view:
SELECT
COUNT(hits.product.v2ProductName),
FROM
[foo-160803:123456789.ga_sessions_20170101]
WHERE
hits.ecommerceaction.action_type = '2'
AND ( BOOLEAN(hits.product.isImpression) IS NULL OR BOOLEAN(hits.product.isImpression) == FALSE ) |

| hits.product | RECORD | This row and nested fields will be populated for each hit that contains Enhanced Ecommerce PRODUCT data. |

| hits.product.productQuantity | INTEGER | The quantity of the product purchased. |

| hits.product.productRevenue | INTEGER | The revenue of the product, expressed as the value passed to Analytics multiplied by 10^6 (e.g., 2.40 would be given as 2400000). |

| hits.product.productSKU | STRING | Product SKU. |

| hits.product.v2ProductName | STRING | Product Name. |



2️⃣Table Schema: https://support.google.com/analytics/answer/3437719?hl=en



## ⚒️Main Process

### Query 01: Calculate total visit, pageview, transaction for Jan, Feb and March 2017 (order by month)

**Purpose:** Aggregates 3 key metrics (visits, pageviews, transactions) by month (Jan–Mar 2017).  

**Goal:** Detect seasonal performance trends and identify which months bring better engagement.

```sql

select 

  format_date('%Y%m',parse_date('%Y%m%d', `date`)) as month,

  count(totals.visits) as visits,

  sum(totals.pageviews) as pageviews,

  sum(totals.transactions) as transactions

from `bigquery-public-data.google_analytics_sample.ga_sessions_2017*`

where _table_suffix between '0101' and '0331'

group by month

order by month;

```

** ✅ Results:** 

| month   | visits | pageviews | transactions |

|---------|--------|-----------|--------------|

| 201701  | 64,694 | 257,708   | 713          |

| 201702  | 62,192 | 233,373   | 733          |

| 201703  | 69,931 | 259,522   | 993          |



**📝 Observation:** The table shows monthly aggregated metrics. March (201703) demonstrates an improvement across all key indicators—visits, pageviews, and transactions—compared to January and February.

### Query 02: Bounce rate per traffic source in July 2017 (Bounce_rate = num_bounce/total_visit) (order by total_visit DESC)

**Purpose:** Calculates bounce rate = bounces / visits per source.  

**Goal:** Evaluate traffic quality by source and identify underperforming channels.

```sql

select

    trafficSource.source as source,

    sum(totals.visits) as total_visits,

    sum(totals.bounces) as total_no_of_bounces,

    round((sum(totals.bounces)/sum(totals.visits))* 100.00,2) as bounce_rate

from `bigquery-public-data.google_analytics_sample.ga_sessions_201707*`

group by source

order by total_visits DESC;

```

** ✅ Results:** 

| source | total_visits | total_no_of_bounces | bounce_rate |

| ------ | ------------ | ------------------- | ----------- |

| google | 38400 | 19798 | 51.56 |

| (direct) | 19891 | 8606 | 43.27 |

| youtube.com | 6351 | 4238 | 66.73 |

| analytics.google.com | 1972 | 1064 | 53.96 |

| Partners | 1788 | 936 | 52.35 |

| m.facebook.com | 669 | 430 | 64.28 |

| google.com | 368 | 183 | 49.73 |

| dfa | 302 | 124 | 41.06 |

| sites.google.com | 230 | 97 | 42.17 |

| facebook.com | 191 | 102 | 53.4 |



**📝 Observation:** Google and direct traffic are the main sources by volume, while platforms like Reddit and mail.google.com show significantly lower bounce rates.

### Query 3: Revenue by traffic source by week, by month in June 2017

**Purpose:** Shows revenue distribution by traffic source, split by month and week.  

**Goal:** Understand financial contribution and fluctuations per source over time.

```sql

select 

    'month' as time_type,

    format_date('%y%m', date(parse_date('%y%m%d', date))) as time,

    trafficsource.source as source,

    sum(product.productrevenue) / 1000000 as revenue

from `bigquery-public-data.google_analytics_sample.ga_sessions_201706*`,

unnest(hits) as hit,

unnest(hit.product) as product

where product.productRevenue is not null

group by time, source



union all 



select

    'week' as time_type,

    format_date('%y%w', date(parse_date('%y%m%d', date))) as time,

    trafficsource.source as source,

    sum(product.productRevenue) / 1000000 as revenue

from `bigquery-public-data.google_analytics_sample.ga_sessions_201706*`,

unnest(hits) as hit,

unnest(hit.product) as product

where product.productRevenue is not null

group by time, source



order by time_type, revenue desc;

```

#### ✅ Results:

| time_type | time | source | revenue |

| --- | --- | --- | --- |

| Month | 201706 | (direct) | 97,333.62 |

| Month | 201706 | google | 18,757.18 |

| Month | 201706 | dfa | 8,862.23 |

| Month | 201706 | mail.google.com | 2,563.13 |

| Month | 201706 | search.myway.com | 105.939998 |

| Week | 201724 | (direct) | 30,908.91 |

| Week | 201725 | (direct) | 27,295.32 |

| Week | 201723 | (direct) | 17,325.68 |

| Week | 201726 | (direct) | 14,914.81 |

| Week | 201724 | google | 9,217.17 |



**📝 Observation:** Direct traffic drives the most revenue both monthly and weekly. Google and DFA are also top-performing sources, but with lower contribution.

### Query 04: Average number of pageviews by purchaser type (purchasers vs non-purchasers) in June, July 2017.

**Purpose:** Compares user engagement across groups based on purchase behavior (June–July 2017).  

**Goal:** Identify browsing behavior and its correlation with conversion.

```sql

with 

purchaser_data as(

  select

      format_date("%Y%m",parse_date("%Y%m%d",date)) as month,

      round((sum(totals.pageviews)/count(distinct fullVisitorId)),2) as avg_pageviews_purchase,

  from `bigquery-public-data.google_analytics_sample.ga_sessions_2017*`

    ,unnest(hits) hits

    ,unnest(product) product

  where _table_suffix between '0601' and '0731'

  and totals.transactions>=1

  and product.productRevenue is not null

  group by month

),



non_purchaser_data as(

  select

      format_date("%Y%m",parse_date("%Y%m%d",date)) as month,

      sum(totals.pageviews)/count(distinct fullVisitorId) as avg_pageviews_non_purchase,

  from `bigquery-public-data.google_analytics_sample.ga_sessions_2017*`

      ,unnest(hits) hits

    ,unnest(product) product

  where _table_suffix between '0601' and '0731'

  and totals.transactions is null

  and product.productRevenue is null

  group by month

)



select

    pd.*,

    round(avg_pageviews_non_purchase,2) as avg_pageviews_non_purchase

from purchaser_data pd

full join non_purchaser_data using(month)

order by pd.month; 

```

** ✅ Results:** 

| month  | avg_pageviews_purchase | avg_pageviews_non_purchase |

| ------ | ---------------------- | -------------------------- |

| 201706 | 94.02                  | 316.87                     |

| 201707 | 124.24                 | 334.06                     |



**📝 Observation:** Surprisingly, non-purchasers have much higher average pageviews per user than purchasers, suggesting browsing-heavy behavior without conversion.

### Query 05: Average number of transactions per user that made a purchase in July 2017

**Purpose:** Measures how many transactions each buyer performs on average.  

**Goal:** Understand user lifetime value and buying depth.

```sql

select

    format_date("%Y%m",parse_date("%Y%m%d",date)) as month,

    round(

        sum(totals.transactions)/count(distinct fullVisitorId)

        ,3) as Avg_total_transactions_per_user

from `bigquery-public-data.google_analytics_sample.ga_sessions_201707*`

    ,unnest (hits) hits,

    unnest(product) product

where  totals.transactions>=1

and product.productRevenue is not null

group by month;

```

** ✅ Results:** 

| month  | Avg_total_transactions_per_user |

| ------ | ------------------------------- |

| 201707 | 4.164                           |



**📝 Observation:** On average, each purchasing user completed over 4 transactions, indicating strong repeat buying behavior in July.

### Query 06: Average amount of money spent per session. Only include purchaser data in July 2017

**Purpose:** Calculates monetary value per session with transaction.  

**Goal:** Estimate effectiveness of purchase sessions in terms of revenue.

```sql

select

    format_date("%Y%m",parse_date("%Y%m%d",date)) as month,

    round(

      ((sum(product.productRevenue)/sum(totals.visits))/power(10,6))

      ,2) as avg_revenue_by_user_per_visit

from `bigquery-public-data.google_analytics_sample.ga_sessions_201707*`

  ,unnest(hits) hits

  ,unnest(product) product

where product.productRevenue is not null

  and totals.transactions>=1

group by month;

```

** ✅ Results:** 

| month  | avg_revenue_by_user_per_visit |

| ------ | ----------------------------- |

| 201707 | 43.86                         |



**📝 Observation:** Each purchase session generated an average of $43.86 in revenue, which reflects solid value per visit from buyers.

### Query 07: Other products purchased by customers who purchased product "YouTube Men's Vintage Henley" in July 2017. Output should show product name and the quantity was ordered.

**Purpose:** Identifies other items frequently purchased with the target product.  

**Goal:** Support cross-sell strategies and product bundling.

```sql

with customers as (

  select distinct fullVisitorId

  from `bigquery-public-data.google_analytics_sample.ga_sessions_201707*`,

  unnest(hits) as hits,

  unnest(hits.product) as product

  where product.v2ProductName = "youtube men's vintage henley"

    and product.productRevenue is not null

    and totals.transactions >= 1

)



select 

  product.v2ProductName as other_purchased_products, 

  sum(product.productQuantity) as quantity

from `bigquery-public-data.google_analytics_sample.ga_sessions_201707*`,

unnest(hits) as hits,

unnest(hits.product) as product

join customers using (fullVisitorId)

where product.v2ProductName != "youtube men's vintage henley"

  and product.productRevenue is not null

  and totals.transactions >= 1

group by product.v2ProductName

order by quantity desc;

```

** ✅ Results:** 

| other_purchased_products | quantity |

| --- | --- |

| Google Sunglasses | 20 |

| Google Women's Vintage Hero Tee Black | 7 |

| SPF-15 Slim & Slender Lip Balm | 6 |

| Google Women's Short Sleeve Hero Tee Red Heather | 4 |

| YouTube Men's Fleece Hoodie Black | 3 |

| Google Men's Short Sleeve Badge Tee Charcoal | 3 |

| Crunch Noise Dog Toy | 2 |

| Android Wool Heather Cap Heather/Black | 2 |

| YouTube Twill Cap | 2 |

| Recycled Mouse Pad | 2 |



**📝 Observation:** Customers who bought the YouTube Henley also frequently purchased other branded apparel and accessories, especially Google Sunglasses.

### "Query 08: Calculate cohort map from product view to addtocart to purchase in Jan, Feb and March 2017. For example, 100% product view then 40% add_to_cart and 10% purchase. 

#### Add_to_cart_rate = number product  add to cart/number product view. Purchase_rate = number product purchase/number product view. The output should be calculated in product level."

**Purpose:** Calculates add-to-cart and purchase conversion rates at product level across 3 months.  

**Goal:** Evaluate funnel effectiveness and optimize product-level conversion.

```sql

with product_data as(

select

    format_date('%Y%m', parse_date('%Y%m%d',date)) as month,

    count(case when eCommerceAction.action_type = '2' then product.v2ProductName end) as num_product_view,

    count(case when eCommerceAction.action_type = '3' then product.v2ProductName end) as num_add_to_cart,

    count(case when eCommerceAction.action_type = '6' and product.productRevenue is not null then product.v2ProductName end) as num_purchase

FROM `bigquery-public-data.google_analytics_sample.ga_sessions_*`

,UNNEST(hits) as hits

,UNNEST (hits.product) as product

where _table_suffix between '20170101' and '20170331'

and eCommerceAction.action_type in ('2','3','6')

group by month

order by month

)



select

    *,

    round(num_add_to_cart/num_product_view * 100, 2) as add_to_cart_rate,

    round(num_purchase/num_product_view * 100, 2) as purchase_rate

from product_data;

```

** ✅ Results:** 

| month  | num_product_view | num_add_to_cart | num_purchase | add_to_cart_rate | purchase_rate |

| ------ | ---------------- | --------------- | ------------ | ---------------- | ------------- |

| 201701 | 25787            | 7342            | 2143         | 28.47            | 8.31          |

| 201702 | 21489            | 7360            | 2060         | 34.25            | 9.59          |

| 201703 | 23549            | 8782            | 2977         | 37.29            | 12.64         |



**📝 Observation:** Conversion rates improve over time, with March showing the highest add-to-cart (37.29%) and purchase (12.64%) rates among the three months.

## 🔎 Final Conclusion & Recommendations



👉🏻 Based on the insights and findings above, we would recommend the **Ecommerce & Marketing Team** to consider the following:



### 📌 Key Takeaways:



✔️ **Replicate successful March strategies** across future campaigns, as this month showed consistent growth in visits, engagement, and conversions.  

✔️ **Improve high-bounce traffic sources** such as YouTube and Facebook by optimizing ad content or landing pages to improve engagement.  

✔️ **Introduce product bundles** that include “YouTube Men's Vintage Henley” and frequently co-purchased items (e.g., Google Sunglasses) to increase average order value.