關於 AI 和 ML 在開放銀行的運用，你不可不知的二三事
Open Banking. Everyone talks about it, everyone has a positive view of it, yet, seldom anyone tells you what problems you may face when you start to get your hands dirty.
The idea of this article is coming from a Forbes piece “AI And ML Can Transform Financial Services, But Industry Must Solve Data Problem First“, in which author Carson Lappetito, President of the Califonia-based Sunwest Bank, describes how AI (Artificial Intelligence) and ML (Machine Learning) can benefit the financial industry and why industry players should watch out about the data problem.
Problem with Clean Data
As someone who worked with data for quite some time in an international bank, I couldn’t agree more on the part that “clean data” is a must-have when you build a model to generate actionable insights.
Whether it’s a small dataset with which you can work on a spreadsheet for data cleaning, or it involves big data for which you need to use programming tools like R or Python, you must organize and clean your data first to avoid any data bias.
Google’s Data Analytics course on Coursera has also emphasized the importance of using clean data and pointed out the implications if we don’t do it properly. You can read about them in my earlier article about the course.
Problem with Data Privacy
That being said, another data issue that was rarely brought up, is how to deal with the data privacy issue.
Visa released a Consumer Payment Attitudes Study in Dec 2020. What I found interesting was the part about customers’ attitudes toward sharing data. During the interviews, even though the majority of people expressed a positive position towards digital banking and would love to enjoy the benefits brought by open banking, there were 1/3 of interviewees reluctant to share their personal and financial data.
And this tendency results in bank customers choose to opt-out different channels that banks use to communicate with them. So when data analysts try to predictcustomers’ banking behavior, or when marketing professionals send campaign messages to the selected customers via different channels, there may be sampling bias. You can imagine how things would get even more complicated when API, financial ecosystem, and affiliate marketing come in and create more connection points.
For example, when a data analyst identified a group of customers who will be more likely to respond to the bank’s multiple product ads, he or she has to decide whether to deliver the message via email or via the push notification of the bank’s mobile banking app. What if the customer agreed to receive notification regarding the latest banking offer via email, but said no to credit card promotion on all channels?
Final Thoughts
Ever since the release of the General Data Protection Regulation in Europe in 2016, customers begin to think twice before giving permission to access their personal and financial profiles. Banks must have clear guidelines that incorporate customers’ data-sharing preferences when AI or ML are used to process those data.
The power of data is a force that cannot be ignored when the industry players start to embrace Open Banking. Every bank is trying to take the benefit by hiring more AI/ ML experts or data scientists to better understand their customers. While the models can identify the most efficient and cost-effective ways to interact with customers, banks have to take these data problems seriously so that they can maintain sustainable relationships with customers for many years to come.
Hope my article can give you some different perspectives on AI and ML’s application in Open Banking. Give me a shout-out below or drop me an email to tell me all about it!
Meanwhile, don’t forget to subscribe to my newsletter so that you won’t miss the latest update. You can also check out my other industry sharing here.
Last but not least, if you get some useful insights from my article, please feel free to click the icon below and support my writing project. This way, I can continue to create high-quality content on this website without ads.
開放銀行 (Open Banking)。每個人都在談它，每個人都保持正面看法，但卻鮮少有人告訴你如果真的採用了開放銀行的運作模式，你可能會遇到哪些問題。
本文的靈感來自富比世雜誌的一篇文章「AI And ML Can Transform Financial Services, But Industry Must Solve Data Problem First (人工智慧和機器學習可以改變金融服務，但是金融產業一定要先解決數據問題)」，該文作者是加州 Sunwest Bank 的銀行總裁 Carson Lappetito。文章內容在探討 AI (Artificial Intelligence) 人工智慧和 ML (Machine Learning) 機器學習將如何幫助金融產業，而金融業者又為什麼要考慮其中可能產生的數據問題。
乾淨數據的問題
在外商銀行跟數據打交道多年，我對Lappetito 的說法真是再同意不過了。如同文中強調的，確保使用乾淨數據 (clean data) 來建模型，得以產生有用的洞見 (insight) 是處理數據過程中絕對不可忽略的一環。
不論你手上是可以用試算表就完成數據清洗 (data cleaning) 的小型數據，亦或需要動用 R 或 Python 等程式語言來處理的大數據，一定要先整理跟清洗你的數據以避免任何數據偏見 (data bias) 的產生。
Google 跟 Coursera 合作的數據分析課程 (Data Analytics Professional Certificate) 中就一再強調使用乾淨數據的重要，同時一併指出如果不好好處理可能會產生哪些後果。這部分可以參考我之前上這門課寫的心得文。
數據隱私的問題
說到這裡，另一個很少有人提出的數據問題，就是數據隱私 (data privacy)。
Visa 在 2021 年初發表過一份消費者支付態度報告2.0，其中有一段我覺得很有趣的資料，是關於消費者對分享數據的態度調查。在訪談過程中，即使大多數人對數位銀行保持正面看法，也想要享受開放銀行帶來的便利，但卻有 1/3 的受訪者表示不願意分享他們的任何資訊。
這樣的趨勢造成客戶可以拒絕銀行用某一個平台跟他們聯絡，而當數據分析師試圖預測消費者的金融行為，或當行銷人員欲透過不同平台寄發活動訊息給特定客戶時，都可能產生樣本偏差。所以你可以想像當 API, 金融生態圈, 聯盟行銷一起納入考量時，隨著客戶接觸點的增加，情況又會變得有多複雜。
舉例來說，在數據分析師篩選出一群對銀行各種產品廣告更容易有回應的客戶後，他接著要選擇用 email 或網路銀行的推播通知來傳遞訊息，此時若客戶同意接受銀行寄發活動 email，但卻不同意接收來自任何平台的信用卡優惠時，該怎麼辦呢？
寫在最後
自從歐洲在 2016 年發佈一般資料保護規範後，消費者在交出自己個人及金融數據的使用權前不免考慮再三，因此銀行在使用 AI 或 ML 處理客戶數據時，應該針對消費者的數據分享意願有明確的規範。
數據的重要在金融業者紛紛擁抱開放銀行後是不可忽略的一股力量。每家銀行都試著招募更多的 AI/ML 專家或數據科學家來享受瞭解客戶帶來的好處。專家建造的模型固然可以找出最有效率和最符合經濟效益的客戶溝通方式；然而，銀行仍須嚴肅看待這些數據問題，才能維持繫長遠的客戶關係。
希望我的文章可以讓你對 AI 和 ML 在開放銀行的運用及考量上有不同的啟發，歡迎在下方留言或透過 email 跟我分享你的心得！
若想收到 Second Star To The Right 的最新文章，別忘了訂閱我的電子報；你也可以在這裡閱讀我的其他產業分享。
最後，如果覺得我的文章對你有幫助，歡迎透過以下綠界或讚賞公民 Liker 的連結支持我後續的寫作計畫，也讓這個網站能以無廣告的方式走得更遠，謝謝。
