114th Daily Writing Challenge

By Bernardus Ari Kuncoro

Marketing Department Hierarchy: Marketing and Sales

Commonly, marketing and sales are combined in one department. Why? Because the goal and focus are the same. To drive revenue.

Marketing has more roles in educating the market, creating campaigns, and acquiring potential customers. While Sales responsibilities are converting the potential customers into the real ones.

Since the roles are more look-alike, the coordination and collaboration between marketing and sales are inevitable. Thus, the use cases mentioned on the graph might be interchangeable and applied in both.

There are also some companies that divide marketing into the B2C (Business to Customer) approach. While the sales division is included in B2B (Business to Business) approach.

Here are numerous references for data science use cases in marketing:

Kalideres, 22 July 2021


113th Daily Writing Challenge

By Bernardus Ari Kuncoro

Data Science Use Cases in Legal Department: Legal Document Analysis and Fraud Detection

There are plenty of legal cases which are related to data analytics. Remember the problem of Cambridge Analytica? Indonesia Ecommerce data leak? BPJS data breach? You name it.

For your reference, the Legal Industry has also been started discovering the standards of data analytics. All aspects in legal such as strategy developments, client interaction and project discovery are transformed digitally.

The data points collected in the digital transformation of legal can be analyzed by Data Scientists such as :

  1. Legal Document Analysis. One of the organization like Data Science for Lawyers provides several material to be learnt by lawyers above. One of the cases is Document comparisons, which basically find similarities and difference between legal texts.
  2. Fraud Detection. This use case is usually combine with outside legal department such as accounting and HR. Why? Because the fraud motive usually about money. And the ones who involve are human. However, in the digital era, hackers are pretty
Data Science for Lawyers Learning Materials (Source: Here)

Kalideres, 21 July 2021


112th Daily Writing Challenge

By Bernardus Ari Kuncoro

Accounting Department Use cases

What is accounting department?

Accounting department refers to the division in a firm that looks after the preparation of financial statements, maintenance of general ledger, payment of bills, preparation of customer bills, payroll, and more. In other words, they are responsible for managing the overall economic front of the business. (Source: here)

What is the function of accounting department?

Well, to make sure all the financial sustainability of business organization

What are the data analytics use cases in accounting department?

  1. Financial Outlook Analysis: Data scientists can utilize financial data to spot trends and extrapolate into the future, helping their employers and clients make the best investing decisions.
  2. Financial Statement Analysis: Data scientists who can read financial statement can help investors or venture capitals decide to invest. This is more like the fundamental analysis in stock market business.
  3. Cost and Budget analysis: Some of data scientists might not be familiar with term budget. Consider it is like a plan for food budget to be spent during the week end. You get the benefit in relationship or social needs. Budgeting is a powerful tool that helps the management in performing its functions such as planning, coordinating, and controlling the operations efficiently. While cost, it is actual spending. If cost lesser than budget, then you still have remaining fund or surplus. Conversely, deficit. If you can predict using the behavior of spending factors plus external trending variables, you can help accounting department to warn the C-levels about their financial status.
  4. Asset allocation analysis: Who have asset? Organization or every persons in the world, right? To be exact, the citizen tho resides in the country that have the law of ownership. The proportion of industry asset allocation can be set to reach its financial goal. It is true that asset allocation is fundamental for successful investing. Data scientists together with Accountants help organization choose your investment strategy wisely and build a profitable portfolio.

Kalideres, 20 July 2021


111th Daily Writing Challenge

By Bernardus Ari Kuncoro

Human Resources Division’s Data Analytics Use cases

All about people is in the human resources. Hence, all the use cases are more less making people keep motivated and satisfied in their work. Including environment, working culture, and last but not least, salary and benefit.

As a data scientist, what are the data analytics use cases that you can cultivate?

  1. Workforce planning. Most of the time HR people are overwhelmed with the various and abundance of tasks. You might give them the external data analysis towards the human resource needs of company. Job market analysis of the specific roles. And so on. Thus HR team can have better understanding on how do they can plan with the workforce.
  2. Salary and Compensation Planning. Again, HR frequently needs some recommendation analysis, especially about how much do they can put a salary towards a new or existing employee. Job salary market analysis compare to the industry or competitors are relevant. If an employee is underpaid, he tend to resign quickly. Turn over rate of the employees is high. As a good data scientist, you must keep your honesty and integrity above all things. You might see the confidential data in this use case.
  3. Employee Evaluation Scorecard. To evaluate the employee, performance review should be done in a regular basis. Some companies do this every six months. Others are 3 months. Or even annually. Well, if a data scientist can predict the scorecard evaluation based on historical data and some external factor, you can help HR to prepare the budget for bonus, for example.
  4. Employee Churn Prediction. This case is very common. Basically you predict when the employee will leave the company based on several historical data that is particularly mapped into the employees. Randy Lao give you great example in doing this project.

Kalideres, 19 July 2021


109th Daily Writing Challenge

By Bernardus Ari Kuncoro

In IT Ops division

It is inevitable for all companies must meet the certain level of IT maturity. In banking, telco, and e-commerce, they are operating the IT system in a way to fulfill all demand of customers and their internal employees’ needs. The use case sample in IT ops are related to the following

a. Enterprise Resource Planning (ERP) Analytics

Oracle mentioned that ERP refers to a type of software that organizations use to manage day-to-day business activities such as accounting, procurement, project management, risk management and compliance, and supply chain operations. A complete ERP suite also includes enterprise performance management, software that helps plan, budget, predict, and report on an organization’s financial results.

To perform ERP analytics means getting deep into the following activities

  • describe and predict the health of enterprise performance management
  • plan the software needs
  • budget allocation prediction
  • financial health issues

It is very unlikely that those ERP dataset are allowed to be published completely towards all division. IT ops, in this case, can provide the dataset to each of division such as HR for performance management, finance for budget allocation prediction / financial health issues, and procurement to purchase the softwares.

b. IT Hardware Maintenance Support Analytics

It is clear that IT hardware needs to be maintained. Laptops of each employees need to be taken care, for example. It will support the operation process. How can data science help? By predicting what and when the hardware need to be maintained. It helps IT Ops a lot.

c. Software Maintenance Analysis

Most of the time, software updates are tedious. Even though we are updating software into the latest version, there might be issues and sometimes in the end, not useful enough. Data Science team can perform analysis on how the employees or customers need the features in the near future. In the end the updates are well received and implemented.

To be continued …

Kalideres, 17 July 2021


108th Daily Writing Challenge

By Bernardus Ari Kuncoro

In Supply Chain Subdivision

In commerce, a supply chain is a system of organizations, people, activities, information, and resources involved in supplying a product or service to a consumer. (Wikipedia)

To supply a good or service to a consumer, procurement and logistics must be provided. To ease the accounting, most of the time procurement activities are included in Finance, instead.

a. Supply Chain Planning

The goal of planning in supply chain is to make sure that the goods are well delivered in a good condition and timely manner. But due to big data with 4V characteristics, the problems are complex and can be derived into the following challenges that data science can solve.

  • Making the supply chain greener to minimize the environmental impact of global sourcing (e.g., shorter distances or consolidated shipments)
  • Increasing visibility into the supply chain and response time (e.g., through blockchain)
  • Adapting to demographic changes and customer expectations (e.g., free same day deliveries)
  • Allowing manufacturers to decrease their product life-cycle times (e.g., through better market insights and smart sourcing) to react to trends and demand more quickly
  • Increasing the product portfolio to serve not only the mass market but the entire demand curve (e.g., through mass-customization)

(source: Here)

b. Capacity and Inventory Planning

Most of the time, manufacturing company that utilize warehouse, they need to perform capacity and inventory planning. The idea is simple. Making sure that the flow of goods are in control based on demand and production. They will distribute based on the forecasting of goods, hence consumers are happy. They can buy and get the product in desired time.

The challenges of capacity and inventory planning are optimizing the interval of 80 – 90% capacity.

Interested in digging up this case? Find the sample MIT thesis of Raytheon’s Circuit Card Assembly (CCA) factory here.

To be continued

Kalideres, 16 July 2021


107th Daily Writing Challenges

By: Bernardus Ari Kuncoro

In production subdivision, there are following data science use cases.

a. Demand Forecasting

In a way to support company’s vision and mission, production team have to make sure how many adequate results of the products. This case is a bit complicated, because not only historical data will be significant, but also external data including the buying power, season, and trend of the market. For data professionals, you need to be aware of those data sources, even ‘unknown data’.

The Demand Forecast vs Actual Sales (Photo was taken from here)

b. Strategy and Operation Planning

To become an impactful and efficient company, strategy and planning in operation are fundamental. Every step of production must meet the criteria of the optimization. In a service company such as banking, for example. There is a specific need on each branch to prepare enough cash money.

This use case is real in one of the biggest bank in Indonesia (Bank Rakyat Indonesia) whose branches located in all of the Sub regency in Indonesia Archipelago. They perform machine learning modeling by using the historical data on each branches and unit. The cash money prediction result will be reference for them to stock the cash money in the office.

Cash optimisation for BRI Branch Offices and Business Units

c. Operational KPIs Dashboard

Each divisions has goals to reach the company’s mission. Usually they use Key Performance Indicator to track the routine tasks. The dashboard will show how many products have been produced, challenges and problem lists, and so on.

d. Product Allocation

Mapping the number of produced items based on the condition in the market is very crucial. Imagine the Unilever produce the bar soaps. Usually they will use historical data, either it is based on seasonality or trend.

e. Predictive Maintenance of Manufacture Machines

Imagine you have the machines from a toaster that you use for breakfast to a computer that you utilize for working. Realize or not, if they fail, what will be the impact? We might not be productive, right? You will not have a toast bread ready. Hungry. Cannot work. Therefore, the quality of a machine is not only based on how useful and efficient, but also how reliable. To be reliable, maintain the machines is obligatory. I recommend you to read this article to jump start the business problem and solution options.


106th Daily Writing Challenge

By: Bernardus Ari Kuncoro

2. Use Cases in Operation

Operation division’s goal is to make sure all the products are ready. The results should met the quality and quantity expectation of the market.

This division is where inputs, or factors of production, is converted to outputs, which are goods and services.

Operations is the heart of a business providing goods and services in a quantity and of a quality that meets the needs of the customers. Operations control the supply chain, including procurement and logistics.

Use Cases in Operation

To be continued

Kalideres, 13 July 2021


104th Daily Writing Challenge

By: Bernardus Ari Kuncoro

It is essential for data professionals know and be familiar with the bank of business problems. The more and various problems they face, the more they become experienced. In this article, I would like to frame the overwhelm thinking about business for data professionals.


  1. Introduction
  2. Use cases in Operation
  3. Use Cases in Finance
  4. Use Cases in Marketing
  5. Use Cases in R&D
  6. Conclusion

1. Introduction

Three common divisions in profit-oriented organization that include Operation, Finance, and Marketing

Most of the time, we found that profit-oriented organizations have at least three division. The one that is responsible for production that includes supply chain, and IT Ops. We frequently call it as Operation. The Finance team that includes Human Resources, accounting, and procurement. And also marketing, the division which is responsible for sales, marketing, and customer service. Last but not least there are some fully funded organizations who have budget for innovation, thus they add more division like Research and Development (R&D).

Additional R&D division that might be more relevant to organization that prioritize the innovation. Note *HR can sometimes standalone or be under operation.

2. Use Cases in Operation

To be continued.

Kalideres, 13 July 2021


Tantangan Menulis Hari ke-104

Oleh: Bernardus Ari Kuncoro

Dalam acara TalksON ke-41 minggu lalu, saya dipercaya sebagai moderator. Topiknya tentang Optimasi persediaan uang cash pada kantor cabang atau unit di BRI. Mas Risal Andika yang menjadi nara sumber.

Dalam webinar tersebut, machine learning digunakan untuk memodelkan berapa uang tunai yang mesti dipersiapkan di tiap kantor cabang BRI. Ada 5000-an model machine learning yang dibangun. Menarik, bukan?

Terdapat 200-an peserta yang mengikuti dari mana-mana. Lewat Zoom. Berikut ini adalah pertanyaan-pertanyaan yang muncul.

Beberapa ada yang sudah dijawab. Namun mengingat keterbatasan waktu, banyak pertanyaan yang belum sempat dibahas.

Wirasta C. Pambudi to Me (Direct Message) (7:32 PM)
Kak mau nanya, dalam melakukan prediksi machine learning, tools apa yang di gunakan?
Wirasta C. Pambudi to Me (Direct Message) (7:35 PM)
Seberapa besar peran data science di BRI?
Wirasta C. Pambudi to Me (Direct Message) (7:38 PM)
karena di perusahaan saya hal prediktif seperti ini sudah pernah dilakukan namun hasil perhitungan berdasarkan statistic seperti ini mentah di mata direksi, sehingga peran data science di perusahaan saya tidak powerfull, mohon penjelasannya

Yulinda to Everyone (7:42 PM)
seberapa presisi prediksi data dr machine learning dg eksisting

Arif RH to Everyone (7:42 PM)
Itu metode evaluasi yg bagus itu yg mana ?

Edo Pratama to Everyone (7:43 PM)
mas, model ini tiap berapa kali dilakukan update nya?

Bima Priambodo to Everyone (7:44 PM)
mas bagaimana life cycle maintenance modelnya?

Rio to Everyone (7:44 PM)
[tanya] evaluasi yang MASE tadi untuk MAE2nya dari mana ya?

Fakhri Rizal Santosa to Me (Direct Message) (7:46 PM)
Siapakah inisiator dalam merubah metode moving average menjadi ML, apakah karena temuan dari team data scientist atau memang menjadi issue di level management / unit bisnis?

Kurnia Andre Febrian to Everyone (7:46 PM)
mas, terkadang saya mengalami stuck, mengaplikasikan persamaan matematis ke codingnya. ada kah tips untuk mengurangi itu? ehehhe

Rio to Everyone (7:50 PM)
[tanya] kalau di BRI, dalam satu project data science kan tadi dibilang ada data engineer, data analyst, data scientist, sama dari product. biasanya berapa orang yang terlibat ya dalam satu project data science tersebut? terus waktunya berapa lama? apakah sepanjang tahun atau beberapa bulan?

Bima Priambodo to Everyone (7:52 PM)
model sampai 5000 unique,, bagaimana cara maintenance modelnya mas?? life scyclenya bagaimana ms?

Nofriandi to Everyone (7:59 PM)
materi persentasi nanti dishare nggak mas?

David Kurniawan

  1. Bahasa pemrograman dan platform apa yang sering digunakan BRI?
  2. Machine learning mengikuti trend and seasonality, lalu bagaimana solusi untuk pandemi seperti sekarang ini dimana datasetnya akan sangat berbeda dan akan susah untuk diprediksi?
  3. Bagaimana mengetahui atau tracking model kita perlu retraining setelah deploymenet?

Iren Ramadhan:
Halo mas, tadi sempat disebutkan bahwa cukup sulit untuk memprediksi di saat Corona seperti ini (dependen kepada peraturan pemerintah). Untuk di BRI sendiri ketika keadaan seperti sekarang, apakah metode predictionnya masih sama seperti yang dipresentasikan saat ini atau ada metode lain ya? Terima kasih

selamat malam, izin bertanya, apakah pada case ini, BRI memasukkan semua variable yang ada pada slide, atau ada eliminasi sehingga tidak semua fitur dimasukkan ke model? lalu jika dilakukan menggunakan metode apa?

Abduh Riski:

  1. bagi yang tertarik untuk melakukan penelitian juga tentang prediksi kas bri seperti ini, apakah bisa mendapatkan data yang diperlukan? bagaimana caranya?
  2. apa Bahasa pemrograman yang digunakan oleh teman2 bri dalam menndevelop model machine learningynya.

Heri Wahyu:
Pada Feature Engineering, jika kita membuat feature baru, apakah berarti pada saat model digunakan nanti datasetnya juga harus dilakukan feature engineering juga untuk membuat feature baru tersebut ya ?

Utk lebih mempelajari materi hari ini, apakah dimungkinkan kami mendapatkan data2 primer yg dpt diolah seperti yg disampaikan narasumber??

Selamat malam host IYKRA, berikut beberapa pertanyaan yang saya ajukan. Semoga bisa diajukan, terimakasih.

  1. Berapa hari prediksi kedepan yang ideal dan hari pengiriman uang untuk Cash In dan Cash Out pada unit kerja maupun kantor cabang BRI?
  2. Algoritma apa yang paling tepat dan sering diterapkan dalam prediksi Cash in dan Cash out?
  3. Berapa hari jeda remodeling yang sering dilakukan di BRI?
  4. Ada banyak unit kerja dan kantor cabang, berarti perlu analisis prediksi 1 per 1 unit kerja dan kantor cabang di BRI?
  5. Error rate (MAPE, MAE, MASE) apa yang biasa digunakan? Berapa batas error rate yang ideal yang digunakan dalam prediksi Cash in dan Cash out di BRI?

Muhammad Ashabul Kahfi:
Mas, mau tanya apakah ada best practice untuk mengetahui bahwa data yang dimiliki bisa dilakukan prediksi menggunakan machine learning

Kevin Prasetio:
mengapa menggunakan mape ? tidak menggunakan LSE atau MSE?


  1. Apakah ada sample dataset yang bisa digunakan untuk melakukan prediction seperti kasus yang dilakukan kak risal?misalkan di kaggle atau lainnya. 2. tools apa aja yang digunakan kak risal supaya menghasilkan prediction dg machine learning tersebut? 3. metode evaluasi yang mana yg paling bagus dalam penggunaan machine learning?

Tris DIanasari
Selamat malam bapak Risal, berikut beberapa pertanyaan yang saya ajukan:

  1. Setelah diperoleh model akan dilakukan iterasi, kapan iterasi akan berhenti?
  2. Apakah ada treatment data sebelum masuk model? seperti data preparation (missing value, outlier dll)
  3. Berapa proporsi data training dan data testing?
  4. Bagaimana cara melakukan seleksi variabel sebelum masuk model?
  5. Setelah pembentukan model, apakah dilakukan proses backtesting untuk menguji kestabilan model?
  6. Hasil proyeksi dapat memprediksi sampai jangka waktu kapan?

Terima Kasih


  1. Prediksi terkait kebutuhan kas unit kerja dapat tersaji H- berapa kah dr kebutuhan? untuk mengantisipasi kekurangan kas agar unit kerja punya waktu untuk melakukan Tambahan Kas 2. jika terjadi Rush/penarikan banyak dan tiba2 karena adanya kebutuhan mendadak/ ada perubahan data bisa langsung mengupdate kah? tks

Hi, pengen nanya Kak, terkait unexpected factor seperti keputusan libur pemerintah. Bagaimana treatmentnya ya?

Aditya EKa:
apakah melakukan perhitungan prediksi machine learning setiap hari? atau tiap minggu?

Hi, pengen nanya Kak, terkait unexpected factor seperti keputusan libur pemerintah. Bagaimana treatmentnya ya?

I Wayan:
Dalam pengembangan terlihat menggunakan beberapa metode. seperti ARIMA, Linear Regression, dll. Apa yg menjadi kelebihan dan kekurangan masing2 metode sehingga perlu dilakukan kombinasi?

Kurnia Andre:
mas, terkadang saya mengalami stuck, mengaplikasikan persamaan matematis ke codingnya. ada kah tips untuk mengurangi itu? ehehhe

Adinda Oktavia:
bagaimana tim data scientist melakukan penyesuaian rumus/metodologi yang diapplikasikan pada saat kondisi ekonomi tidak menentu (contoh saat Cov 19)? dan seberapa besar effect dari kebijakan cashless yang saat ini banyak berlaku untuk transaksi yang ada.

Ocha Alieffi:
aya mau tanya kak proses pola kolaborasi dengan tim lain (DE,DA,dll) seperti apa ya gambarannya

Izin bertanya mas, kan tadi disebutkan menggunakan OLS. Berarti untuk asumsi-asumsi klasiknya harus terpenuhi. Bagaimana ketika terdapat beberapa asumsi-asumsi tersebut tidak terpenuhi? Sehingga, apa yg harus dilakukan untuk dapat dilakukan analisis lebih lanjut?

`Nanya lagi kak, tadi dikatakan bahwa proporsi predictior itu 20%nya merupakan input dari subject-matter expert. bagaimana menentukan apakah “wejangan” dari subject matter-expert di-accept atau enggak ??

Kalideres, 12 Juli 2021