WILMA

What do you want to become when you grow up? The typical answer would be a teacher, a doctor or an engineer. But as early as 10 years old, I knew I wanted to be a Customs Broker. My childhood dream was pursued by taking a Bachelor’s Degree in Customs Administration. When I was in my college days I fell in love learning Tariff and Customs Law and Computation. My ONE love! Have I told you that I got 91% in Computation in the customs broker board exam? Naks! Tooting my horn!

CCBI

I am delighted to announce that after many years living abroad, the Chamber of Customs Broker (Accredited Professional Organization of Philippine’s Professional Regulation Commission) created its Singapore Chapter. And my name was inked in history as one of its founding officers.

I am blessed to be appointed as PRO and as one of Singapore Chapter officers, we will do our best to help and be a contributor in developing and enhancing Customs Broker as a reputable profession in and out of the Philippines.

CCBI SINGAPORE CHAPTER —APPOINTED OFFICERS

FREIGHT/LOGISTICS ANALYST

As professionals we need to upgrade our skills and being an analyst is an outstanding way to develop our brokerage expertise.

Below is the career map shown by Mr. Roger Lee in CCBI Goes International:

Most of the companies here in Singapore are looking not only for a Customer Service or Coordinator but they are eyeing people that possess knowledge on how to handle tools like pivot table for reporting or data handling/tracking.

CUSTOMS BROKERS OF THE FUTURE

As I have said in this post titled Evolution of Logistics Analyst, analyst is the new and best thing not only for career but also for businesses involved in logistics/warehousing/freight forwarding businesses.

And I do hope that in the near future, logistics, supply chain, brokerage companies will be aware of the power of data they own right now and how to harness it to solve their problems and how to take advantage to increase a firm’s profit.

For Customs Brokers in the Philippines that want to kickstart their upskilling or retooling, you can check-out DataSense Analytics or Project SPARTA for free courses with certificates and for those who have the budget to learn, you can look for Eskwelabs!

That’s all for now! See you in the FUTURE FELLOW CUSTOMS BROKERS!

INTRODUCTION

In 2016, Cyclistic launched a successful bike-share offering. Since then, the program has grown to a fleet of 5,824 bicycles that are geotracked and locked into a network of 692 stations across Chicago. The bikes can be unlocked from one station and returned to any other station in the system anytime.

Cyclistic categorized the riders as Casual who purchased single-ride or full-day passes and Members who purchased annual memberships.

Cyclistic’s marketing strategy relied heavily on building general awareness and appealing to broad consumer segments. Marketing team created flexibility on pricing plans like single-ride passes, full-day passes, and annual memberships. But there is no effort for an upselling plan to convert the casual riders into members.

Cyclistic’s finance analysts have concluded that annual members are much more profitable than casual riders.

Ms. Moreno, Director of Marketing Team, has a clear goal: Design marketing strategies aimed at converting casual riders into annual members. As a Junior Analyst in the Marketing Team, I was assigned to answer: How do annual members and casual riders use Cyclistic bikes differently?

BUSINESS TASK

Identify, analyze, and assess the bike usage trends and patterns of casual riders/members for the Cyclistic future marketing program.

DATA SOURCES

Internal data of Cyclistics was used in this case study under Motivate International Inc, under license. It is the original data from the source. A 12-month (November 2021 to October 2022) historical dataset was downloaded and stored in Google Drive and organized in a YYYYMM_CompanyName manner. To avoid sampling bias we will use the dataset that will represent the population and anonymization was used to implement data privacy.

Reliable — it is first-party data. It is owned by Cyclistic.

Original — Dataset is straight from the source. Although already anonymized the other data is still intact.

Comprehensive — There is some nulls but the importance of it can be checked in “Process” stage.

Current — Data is up to date and has the current information of the cyclers’ usage.

Cited — The City of Chicago owns all right, title, and interest in the Data. Thus, this is a reliable data source.

Since the finance analysts already concluded the annual memberships are much more profitable, anonymized data provided will help to focus on the main task which is to define the different usage of the bike renters.

CLEANING OR MANIPULATION OF DATA

Since the data exceeds the allowable complimentary BigQuery sandbox, I loaded the dataset in pgAdmin4 to use SQL. Tableau Public was used to create an interactive visualization or a dashboard and for an easier way to check the relations between columns, I used Anaconda.

STEP 1: Creating a table named bike.

Since all of our downloaded files have similar features, it will be logical to create only one table. Using this kind command, you can easily manipulate the features into correct data types.

CREATE TABLE IF NOT EXISTS public.bike
(
    ride_id character varying COLLATE pg_catalog."default" NOT NULL,
    rideable_type character varying COLLATE pg_catalog."default",
    started_at timestamp without time zone,
    ended_at timestamp without time zone,
    start_station_name character varying COLLATE pg_catalog."default",
    start_station_id character varying COLLATE pg_catalog."default",
    end_station_name character varying COLLATE pg_catalog."default",
    end_station_id character varying COLLATE pg_catalog."default",
    start_lat numeric COLLATE pg_catalog."default",
    start_lng numeric COLLATE pg_catalog."default",
    end_lat numeric COLLATE pg_catalog."default",
    end_lng numeric COLLATE pg_catalog."default",
    member_casual character varying COLLATE pg_catalog."default",
    CONSTRAINT ride_pkey PRIMARY KEY (ride_id)
)

Ensure to use the correct data type for future data aggregation/manipulation and Tableau visualization.

STEP 2: Copying and merging data.

We will not use any JOIN or UNION command because we created only one table. Stacking the 12-month in the ride table created.

COPY bike
--FROM '/Users/wilmalapuz/Downloads/202111_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202112_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202201_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202202_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202203_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202204_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202205_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202206_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202207_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202208_Cyclistic.csv' 
--FROM '/Users/wilmalapuz/Downloads/202209_Cyclistic.csv' 
FROM '/Users/wilmalapuz/Downloads/202210_Cyclistic.csv'
DELIMITER ','
CSV Header;

STEP 3: Data Familiarization.

Schema of the table created.

SELECT *
FROM information_schema.columns
WHERE table_name = 'bike';

Checking the total rows of the table

SELECT COUNT(*)
FROM bike

Getting the head of the table

SELECT *
FROM bike 
LIMIT 10

We need to know how many distinct observation per column

SELECT 
    COUNT(DISTINCT ride_id) AS rideid
    ,COUNT(DISTINCT rideable_type) AS ridetype 
    ,COUNT(DISTINCT started_at) AS startat
    ,COUNT(DISTINCT ended_at) AS endedat
    ,COUNT(DISTINCT start_station_name) AS startstation
    ,COUNT(DISTINCT start_station_id) AS startid
    ,COUNT(DISTINCT end_station_name) AS endtstation
    ,COUNT(DISTINCT end_station_id) AS endid
    ,COUNT(DISTINCT start_lat) AS startlat
    ,COUNT(DISTINCT start_lng) AS startlng 
    ,COUNT(DISTINCT end_lat) AS endlat
    ,COUNT(DISTINCT end_lng) AS endlng       
 ,COUNT(DISTINCT member_casual) AS membercasual   
FROM bike

Since the count and count distinct of ride_id is the same, we will not use any delete duplicate function for this dataset.

How many nulls per column

select 
  sum(case when ride_id is null then 1 else 0 end) rideid
  ,sum(case when rideable_type is null then 1 else 0 end) ridetype
  ,sum(case when started_at is null then 1 else 0 end) startat
  ,sum(case when ended_at is null then 1 else 0 end) endat
  ,sum(case when start_station_name is null then 1 else 0 end) startstation
  ,sum(case when start_station_id is null then 1 else 0 end) startid
  ,sum(case when end_station_name is null then 1 else 0 end) endtstation
  ,sum(case when end_station_id is null then 1 else 0 end) endid
  ,sum(case when start_lat is null then 1 else 0 end) startlat
  ,sum(case when start_lng is null then 1 else 0 end) startlng
  ,sum(case when end_lat is null then 1 else 0 end) endlat
  ,sum(case when end_lng is null then 1 else 0 end) endlng
  ,sum(case when member_casual is null then 1 else 0 end) membercasual
from bike

We will go on to the next dataset feature which is the rideable_type.

SELECT rideable_type, COUNT(rideable_type)
FROM bike
GROUP BY rideable_type
ORDER BY COUNT(rideable_type) ASC

There are 3 classes of types in this data; docked, classic and electric.

SELECT started_at, ended_at
FROM bike
ORDER BY started_at ASC

STEP 4: Adding/Dropping columns for data analysis.

Going back to our business question, which is to analyze the bike usage of members and casual riders, we will now list all the underlying queries that we can formulate to answer the business task.

-What is the popular start time of the cyclers?

-What is the average cycling time?

-What is the busiest and slowest day/month?

-What is the average distance per cycle?

-Where is the most active station/dormant station?

By those questions above, we will need to drop the end_station_names and station_id column which consists of almost 1 million NULLS per column. We will add columns like starttime (extracted from started_at), cycletime (elapsed time between ended_at and started_at), month/day/year (convert the timestamp to their designated month/day/year), and distance traveled (using point command).

SELECT ride_id, rideable_type, started_at, ended_at, start_station_name AS station_name, start_lat, start_lng, end_lat, end_lng, member_casual
    ,started_at::time AS starttime
    ,extract
     (
             epoch from ((ended_at - started_at)/60)::interval
        )
         as cycletime
 ,TO_CHAR(started_at, 'fmDay') AS "day"
 ,TO_CHAR(started_at, 'fmMonth') AS "month"
    ,TO_CHAR(started_at, 'YYYY') AS "year"
 ,((point(start_lng, start_lat) <@> point(end_lng, end_lat)) * 1.609)as distance
FROM bike 
WHERE start_lat != end_lat AND start_lng != end_lng

We will not stop there, we want to clean it further by using the ORDER BY command on our newly added column. We found out that there are negative cycle times and less than 5 meters cycling distance in some recorded rides. We will exclude those using the command below:

SELECT *
from (SELECT ride_id, rideable_type, started_at, ended_at, start_station_name AS station_name, start_lat, start_lng, end_lat, end_lng, member_casual
    ,started_at::time AS starttime
    ,extract
     (
             epoch from ((ended_at - started_at)/60)::interval
        )
         as cycletime
 ,TO_CHAR(started_at, 'fmDay') AS "day"
 ,TO_CHAR(started_at, 'fmMonth') AS "month"
    ,TO_CHAR(started_at, 'YYYY') AS "year"
 ,((point(start_lng, start_lat) <@> point(end_lng, end_lat)) * 1.609)as distance
FROM bike 
WHERE start_lat != end_lat AND start_lng != end_lng) as Q 
WHERE Q.cycletime >2 AND Q.distance > 0.005