ParcFinder Proposal

HOME

PROJECT DETAILS

POSTER

APPLICATION

RESEARCH PAPER

Project Background

Flowchart of Traditional SingPost Delivery Service in E-Commerce

Issues and Problems

With the expansion of E-Commerce in Singapore, there is a growing demand for the provision of effective logistical services to facilitate the delivery and receiving of goods and services to consumers.

As we analysed the entire customer experience journey in the e-commerce industry, we realised that there is a gap in the service delivery process in the event of a missed delivery. A study by the NUS Logistics Institute - Asia Pacific shows that as of 2016, the delivery failure rates in Singapore hover at more than 15%.

Traditionally, customers who missed a delivery from their logistics providers are redirected either to the post office to collect their parcels. They could also be required to make a call to their logistics providers and rescheduling for a redelivery, and at times required to pay an additional fee for the services. As for the logistics providers, carrying out redeliveries incurs additional operational costs in terms of man-hours and resources.

In our opinion, the last-mile delivery process should be improved such that consumers are not required to go through the additional hassle and incurring additional costs of collecting their missed parcels, and logistics providers can ensure that their customers get their parcels without any additional incurring of operational costs in carrying out redeliveries.

Motivation

Project Aim

Our team has explored the Self-Collection Points as a viable solution in addressing the issue of missed deliveries. We recognise the importance for Logistic Companies to be able to determine the location of their self-collection points, in order to maximise coverage as well as improving their last-mile delivery service experience for their customers.

POPStation, a 24/7 self-collection point from SingPost

Proposed Solution

Through the conceptualisation of our application - ParcFinder, we provide users with the tools to visualise the geographical accessibility and generate spatial analysis reports of their existing self-collection points. We hope to provide the necessary insights for our users in their decision-making process in the location of their self-collection points.

Our project will provide an application that will present to users the following insights and analyses:

The number of residential locations that each self-collection point caters to within a specified buffer distance

The accessibility scores of the respective self-collection points

The heatmaps to visualise hotspots of self-collection points and determining areas that are underserved

Approach - GIS and Accessibility Models Used

Overview

ParcFinder is an application that uses various Accessibility Models and give the user different levels of understanding the geographical accessibility of the self-collection points. All the models are executed and have the following assumptions in common:

Each residential building is regarded as one demand point, be it a public or private housing.

The demand and usage of each self-collection point is uniform across all the locations.

These models are as follows, based on the level of analysis and insights each method brings:

Level 1: Catchment Area Buffer Analysis

Circular catchment area of Noerrebro Station in Copenhagen

Objective of Analysis: To display and capture the number of residential locations that each point caters to, based on the specified buffer distance.

A Catchment Area Buffer of a self-collection point is defined as the vicinity of the point and its surroundings. It can be viewed as the base of which the residential locations have accessibility to. There are two types of catchment area buffers - Circular Area and Service Area. In the context of ParcFinder, it uses the circular buffer analysis of the Catchment Area Buffer Analysis.

There are two phases of the circular catchment area buffer analysis. Firstly, we set the radius of the buffer distance to determine the geographical catchment area of the self-collection point. The size of the buffer distance is dependent on the willingness to walk criteria of a consumer to the respective self-collection point in their catchment area. Secondly, we determine the number of residential locations that fall within the specified buffer distance, to determine the demand catered by each self-collection point.

The Catchment Area Buffer analysis is executed on the assumption that the self-collection service area is circular, and the collection point is centered on the residential locations.

Level 2: Hansen Potential Accessibility Model

Objective of Analysis: Each SCP is given an accessibility score to see how well it performs regarding its accessibility to the residential location.

Level 3: Kernel Density Estimation

Objective of Analysis: Looks at the macro level the areas in which the existing SCPs clusters and disperses to determine which areas are adequately served or underserved.

Project Proposal

Project Milestones

Project Storyboard

Project Task Allocation

S/N	Task	Done by	Week	Dates	Status
1	Topic Brainstorming	All	2 & 3	15 Jan - 26 Jan	Completed ✔
2	Drafting and refinement of Project Proposal	All	2 & 3	15 Jan - 26 Jan	Completed ✔
3	Consultation with Prof Kam for Feedback on Proposal	All	3	22 Jan - 26 Jan	Completed ✔
4	Finalisation of Project Topic and Focus	All	3	26 Jan - 28 Jan	Completed ✔
5	Compilation and Cleaning of Datasets (SingPost Post Office, POPStation, EzBuy)	Shu Yan & Zhi Hui	4 - 6	29 Jan -18 Feb	Completed ✔
6	Creation of Wiki Page	Aaron & Shu Yan	5 - 6	5 Feb - 14 Feb	Completed ✔
7	Generation of Storyboard	Aaron	5	5 Feb - 9 Feb	Completed ✔
8	Inddependent learning of R and R Shiny on DataCamp	All	6 - 9	12 Feb - 9 March	Completed ✔
9	Research on Tools for Data Conversion	Shu Yan & Zhi Hui	6 - 7	12 Feb - 19 Feb	Completed ✔
10	Wiki Content Update	Aaron	7	19 Feb - 23 Feb	Completed ✔
11	Preparation for Interim Presentation	Aaron and Shu Yan	7 - 8	19 Feb - 2 Mar	Completed ✔
12	Consolidation and complete conversion of data into SHP File	Shu Yan and Zhi Hui	8	26 Feb - 4 March	Completed ✔
13	Interim Presentation with Prof Kam	All	9	6 Mar	Completed ✔
14	Map and Interface Development Buffer Analysis: Shu Yan and Zhi Hui Hansen Potential Analysis: Zhi Hui Kernal Density Estimation: Shu Yan User Interface (Web Layout): Shu Yan User Interface (Description) : Aaron	All	9-13	5 Mar - 6 Apr	Completed ✔
15	Debugging and Analysis of Results	All	10-14	12 Mar - 10 Apr	Completed ✔
16	Creating and Submission of Townhall Poster	Aaron	13	2 Apr - 6 Apr Submission: 9 Apr	Completed ✔
17	Uploading of App on Shinyapps.io	Zhi Hui	13 - 14	2 Apr - 15 Apr	Completed ✔
18	Updating of Project Wiki Page	Aaron	13 - 14	2 Apr - 15 Apr	Completed ✔
19	Townhall Poster Presentation @ SLA	All	14	11 Apr	Completed ✔
20	Finalizing Wiki & Research Paper	All	14	12 Apr - 15 Apr	Completed ✔

Data Sources

S/N	Title	Format	Website Link / Sources
1	Master Plan 2014 Planning Area	SHP	https://data.gov.sg/dataset/master-plan-2014-planning-area-boundary-web
2	SingPost Post Office	Unformatted	https://www.singpost.com/list-of-post-offices
3	SingPost POPStation	Unformatted	https://www.mypopstation.com/locations
4	EzBuy	Unformatted	https://ezbuy.sg/Help/QuickGuide#Delivery
6	Residential Location	csv	Public Housing: https://www.ema.gov.sg/statistic.aspx?sta_sid=20150617kEhn53Jk6sDQ Private Housing: https://www.ema.gov.sg/statistic.aspx?sta_sid=20150209DnSuIwVsNHBY

Case Study

We have selected the following Self-Collection models in generating our analysis from ParcFinder.

Case Study 1: Singapore Post POPStation

Over 140 POPStation locations islandwide

• Operates 24/7
  
• Allow customers to take back control and self-collect their items at their preferred time and location
  
• Customer can now unlock the POPStation locker with their mobile device to retrieve their parcel using the Bluetooth function on their mobile device
  

Case Study 2: EzBuy

EZBuy has 277 collection points islandwide spread across MRT stations, neighbourhood collection points and warehouse collection centres

EZCollect (New initiative by EZBuy):

• Collaboration with third party retail shop owners for the collection of parcels
• 50 collection points under our ezCollect collaboration with 12 of such collection points operating at 24 hours

ParcFinder Application

Application Architecture

Application Overview

Application Features

Results and Findings

Singapore Post

From the three levels of analyses, we can derive that there is much room for improvement for SingPost to improve the geographical accessibility of its post offices to residential locations. Coupled with the visualisation results from the KDE analysis, the areas in which SingPost should look into expanding the number of SingPost Post office should primarily be either in the West Area (Jurong East, Jurong West, Bukit Batok), and in the North Area (Woodlands, Yishun, Sembawang).

However, we also note that the cost of setting up and operating a new post office is high for SingPost. Therefore with this analysis, it supports the scope of our project, that to extend its last-mile delivery service efficiency, SingPost should look more into the expansion of its SCPs – POPStation, to enhance the overall accessibility of its services to consumers.

Singapore Post POPStation

From the three levels of analyses, we can derive that the extension of SingPost last mile delivery solution through POPStation has supported the improvement in the geographical accessibility to residential locations. Nevertheless, with the visualisation results both from the HPA and KDE Analyses, there are certainly areas of improvement regarding improving the accessibility scores of POPStations overall, especially those in the Western and Northern Regions of Singapore.

The Catchment area buffer analysis and the KDE analysis both suggest that there is currently an extensive coverage of POPStations SCPs in Singapore. However, as seen in the HPA Analysis in Figure 6, there is still the low accessibility scores of POPStations overall. One of the possible insights would be due to the insufficient coverage of transportation networks in these areas, which may have affected the overall performance of these POPStations. Hence, what SingPost can do regarding improving the accessibility scores of POPStations, would be to build more POPStations SCPs in these areas, preferably in locations that are nearer to the residential locations such as neighbourhood shopping centres and transport terminals. POPStations can also be opened in areas within private estates, to improve the overall accessibility coverage especially to private residential locations.

POPStations as an SCP is situated mostly in the central district, pertinently in the CBD areas, and in the center of various towns. One explanation for this could be that SingPost takes into consideration of the travel patterns of its consumers who utilise their service, where consumers would be travelling from their areas of work or school, rather than from their areas of residence.

EzBuy

From the three levels of analyses, we can derive that the EzBuy CDPs have the best performance between both SingPost Post Offices and POPStations in terms of its geographical accessibility to residential locations. With the extensive coverage that EzBuy provides, especially in the heartlands, it demonstrates the effectiveness of adopting CDPs in the Self-Collection model to improve efficiency in the last-mile delivery service.

These findings are supported by all 3 levels of the analyses generated. With the HPA Analysis showing EzBuy Overall Accessibility performing better than average (Figure 11) and the probability density clusters from the KDE Analysis (Figure 12) showing a proportionate distribution of CDPs in Singapore.

An area of improvement would be for EzBuy to extend its CDPs to the private estates like Bukit Timah, as these areas are where the accessibility scores and buffer distance coverage fare the lowest.

Challenges and Limitations

Challenges

Data Cleaning and Transformation

• Data retrieved not available in KML/SHP/XML format
  • Team effort to convert to SHP format
  • Documentation to keep track of changes.
• Data (longitude and latitude) not provided
  • Write script to transform postal codes
  • For postal codes that wasn't able to transform using the script, manually retrieve longitude and latitude from Google Map

No Prior Experience in Geospatial and R Programming

• Team effort to learn from Prof Kam and attend online classes from DataCamp
• Complete the hands on practice to familiar ourselves in the R environment
• Read relevant packages' documentations

Limitations

Due to the scope of the project, our team decided to focus on residential areas and did not include the rest, such as business offices/buildings. Distances were measured in straight line distance for Buffer Analysis and Hansen Potential Estimation.

Feedback from Industry Showcase

We had the privilege to present our prototype and our findings to industry experts and practitioners at GeoWorks @ Singapore Land Authority on the 11th April 2018. These experts and practitioners came from several government agencies, of which includes the Singapore Land Authority, Urban Redevelopment Authority, Housing and Development Board, Ministry of Transport, Land Transport Authority, National Environment Agency, Ministry of Social and Family Development, and the People’s Association.

A. Looking at the accessibility of the various SCPs to Office locations

Our team received comments from a few groups of industry professionals that we should take into account of the patterns of usage of SCPs by consumers in our analysis. Typically, consumers who opt for SCPs as their delivery collection method are more likely to select SCPs that are located near at their workplaces or places of interest they typically visit after work, such as the gas station or a supermarket, rather than an SCP located near their residency. While we raised to these groups the scope of which our system aims to focus on, we also agreed with their comments. Through the exchange, a constructive suggestion was to include the analysis of visualising accessibility of the existing SCPs to Office Locations, to provide an incremental layer of our analysis and see how logistics companies fare concerning their Self Collection delivery service both to individual consumers as well as to corporate consumers.

B. Looking into the demand and utilization rate of each SCP and the accessibility at different timeslots of the day

We also received comments that our analysis could take into account of the demand and utilization rate of each SCP from POPStation and EzBuy. Through the analysis of the demand and utilization rate, we can extrapolate the findings and determine which are the areas where the demand for the SCPs services are higher than others, and which existing SCPs percentage of user rates. Thereby it will help in determining which areas would require expanding more SCPs, be it parcel lockers from POPStation or CDPs from EzBuy.

Also, an additional comment for the analysis of the EzBuy self-collection model would be to include a feature that shows the accessibility of the CDPs at the different times of the day, as the various existing CDPs have different time slots with regards to its operations. By comparing how each CDPs fare in the different parts of the day, it could enhance the overall analysis and help EzBuy or other logistics companies using CDPs to determine which areas should new CDPs be set up based on the different timings of the day.

C. Use of colours in mapping methods from ParcFinder

One of the industry professionals from the Singapore Land Authority gave us feedback with regards the usage of colours in our thematic mapping of both the Hansen Potential Accessibility Analysis and the Kernel Density Analysis. He commented that the use of the colour “Red” in our analyses should be reconsidered, as red is a symbol of “danger” or “warning” that calls for the users’ attention. As shown in the two diagrams below, we previously used the colour red to represent the areas of the SCPs in their accessibility scores in HPA, and the various point of clusters in KDE.

We found this feedback to be relevant for our HPA analysis, as the colours are meant to highlight more on the accessibility score of the SCPs. Hence, we changed the RColorBrewer colour palette of the HPA analysis from “Reds” to “Greens”. However, with regards to the KDE analysis, we explained that the use of the heat colours was more appropriate as the purpose of the analysis was to determine the areas where the SCPs cluster, which in other words meant determining the “hotspots” of the existing SCPs. Hence as the colour red was suitable and appropriate mainly for the use of a heat map context, we decided to stick to our initial colour palette for the KDE Analysis.

Future Work

Following the successful development and showcase of our project, our team has outlined the following areas of future work in the enhancement of ParcFinder:

1. Two-step floating catchment area method (2SFCAM)

The 2SFCAM measure the spatial accessibility to each self-collection point, by calculating the ratio of the self-collection point to the residential population in that specified area

2. Upload dataset (SHP Files, XML format)

Allow users to upload their data into the web application and perform the three layers of analysis provided

Providing analyses of the geographical accessibility of existing datasets to different locations asides residential locations such as Office Locations, school locations etc.

This will enable users to be able to interact with various datasets and obtain different analyses that suit to their needs.

3. Slider for KDE Function

Allow users to adjust the bandwidth for their analysis

Select Color Scheme

Allow users to change the color for HPA and KDE for comfortable viewing purpose

Meet the ParcFinder Team

From left to right: Lee Zhi Hui, Chien Shu Yan, Aaron Ching Kwun Hin

Comments

Please insert your comments here:

No.	Name	Date	Comments
1.	Insert your Name here	Insert Date here	Insert Comment here
2.	Insert your Name here	Insert Date here	Insert Comment here
3.	Insert your Name here	Insert Date here	Insert Comment here