美國大數據公司聯手京東做信用評估

Financial technology start-ups are creating new models of lending. They mine streams of digital data with clever software to calculate creditworthiness instead of relying on a person’s credit history, the main ingredient in traditional credit scoring.

一些面向金融領域的科技初創公司正在推出新的貸款模式。它們用智能軟件挖掘電子數據流來計算信譽，而不是像傳統的信用評分那樣，以個人的信用記錄爲基礎。

So far, the new breed of big data lenders has focused on niche markets — recent college graduates, immigrants and payday borrowers — where people often have scant or inconsistent repayment records, and the conventional math of risk analysis stumbles.

目前已經有一幫新生的“大數據放款機構”專注於在利基市場上——剛畢業的大學生、移民和發薪日借款人。這類人的還款記錄往往很少，或者不連貫，使用傳統的風險分析數學手段效果不佳。

ZestFinance, a pioneer in the field, is moving into a huge market where credit histories are scarce: China.

ZestFinance是這個領域的先驅之一，目前正步入一個信用記錄稀少的龐大市場：中國。

ZestFinance and , a Chinese online retail giant, are announcing a joint venture to provide a consumer credit scoring service in China. The venture, JD-ZestFinance Gaia, will initially be used to assess credit risk and offer installment loans for purchases on , which has 100 million active customers and generates yearly revenue of $20 billion. The venture intends to eventually offer the credit-analysis service to corporate customers throughout China.

ZestFinance和中國網絡零售巨頭京東宣佈成立一家合資公司，在中國市場上提供消費者信貸評分服務。京東擁有1億活躍用戶，年營收達200億美元。這家合資企業名爲JD-ZestFinance Gaia，最初將爲京東上的分期貸款購物行爲評估信貸風險。公司打算最終爲中國各地的企業客戶提供信用分析服務。

is also making a minority investment in ZestFinance, though the companies would not disclose the size of the investment or the valuation of the start-up.

京東還對ZestFinance進行了少數股權投資，不過雙方沒有透露投資規模或是ZestFinance的估值。

“This is a great validation that what we’ve built works,” said Douglas C. Merrill, founder and chief executive of ZestFinance.

“這是對我們的巨大認可，證明我們的方法是行得通的，”ZestFinance的創始人兼首席執行官道格拉斯·C·梅里爾(Douglas C. Merrill)說。

There is a lot of enthusiasm for the data science approach to credit analysis, and venture funding is flowing into this emerging field. The promise is that high-tech tools can give greater depth and detail to the basic principle of banking: know your customer. Start-ups in the field, beside ZestFinance, include Affirm, Earnest, Elevate and LendUp.

人們對於用數據科學的方法來進行信用分析熱情高漲，風險資本也正在流入這個新興的領域。銀行業的基本原則是瞭解客戶，而高科技工具有望爲此提供更深層次的剖析和更多的細節。除了ZestFinance之外，該領域的初創公司還有Affirm、Earnest、Elevate和LendUp。

The start-ups’ methods vary, as do the data sources they tap. But their algorithms sift through data that can include a person’s social-network connections, web-browsing habits, how they fill out online forms and their online purchases.

這些初創公司的方法各異，利用的數據源也不盡相同。不過，它們用來篩選數據的算法可能會涵蓋個人在社交網絡上的關係、瀏覽網頁的習慣、填寫網上表格的方式，以及網上購物的偏好。

The software looks for patterns and correlations: digital signals that help assess an individual’s willingness and ability to repay. The picture that emerges from the data, enthusiasts say, should result in more accurate risk analysis, thus opening the door to extending consumer credit to millions more people at lower cost.

這種軟件尋找的是模式與相關性，即有助於評估一個人的償還意願和能力的數字信號。追捧者認爲，數據勾勒出來的面貌，應該可以讓風險分析變得更加精準，因此有助於以更低的成本把消費者信貸提供給額外的人，而其中涉及的人數成百上千萬。

Yet public policy experts say the enthusiasm for the new lending models is outrunning the evidence. The accuracy and fairness of big data credit technology is unproven, said Aaron Rieke, a former lawyer for the Federal Trade Commission and director of technology projects for Upturn, a policy consulting firm. Mr. Rieke was a co-author of a report last year, supported by the Ford Foundation, that cited ZestFinance as a prime example of big data underwriting, which deploys “fringe alternative scoring models.”

然而，一些公共政策專家認爲，人們對貸款新模式的熱情跑在了證據的前面。阿隆·裏克(Aaron Riek)稱，大數據信用技術的準確性和公正性尚未經過證實。裏克曾在聯邦貿易委員會(Federal Trade Commission)任律師，目前是政策諮詢公司Upturn的技術項目總監，去年參與撰寫了福特基金會(Ford Foundation)贊助的一份報告。該報告將ZestFinance稱爲大數據貸款審批領域的一個典型，採用“非主流的替代性信用評分模型”。

But sought out ZestFinance, tested its technology and came away impressed. Last fall, Chen Shengqiang, chief executive of the Chinese company’s finance unit, visited the ZestFinance offices in Los Angeles and spoke to Mr. Merrill and members of his team. Soon after, Mr. Merrill traveled to the Chinese company’s headquarters in Beijing to work on setting up a test of ZestFinance’s technology, working with data.

但是京東找到了ZestFinance，測試了它的技術，並對它印象深刻。去年秋天，京東金融集團的首席執行官陳勝強參觀了ZestFinance位於洛杉磯的辦公室，並與梅里爾及其團隊的成員進行交談。不久後，梅里爾前往北京的京東總部，用該公司的數據對ZestFinance的技術進行了一次測試。

ZestFinance, founded in 2009, began making loans itself and underwriting loans made by lending partners in 2010. In the United States, ZestFinance has focused its risk analysis on installment loans that are a lower-cost alternative to payday loans. Those borrowers are in the subprime market, and typically have experienced a credit setback in the past, like a personal bankruptcy.

ZestFinance成立於2009年，從2010年開始自己爲客戶提供貸款，並審批合作伙伴的貸款。在美國，ZestFinance一直專注在分期貸款的風險分析上。對於發薪日貸款，分期貸款是一個成本較低的選擇。其借款人來自次級貸款市場，通常以前都在信用上遭遇過問題，比如個人破產。

In China, had a very different assignment for ZestFinance, using different data sources than in America. Only 20 percent of Chinese adults have a credit score, and they often are given credit through the People’s Bank of China, the nation’s central bank, and through affiliations with large state-owned corporations.

在中國，京東交給ZestFinance的任務則大不相同，而且使用的數據源也有異於美國。在中國成年人中，只有20%擁有信用評分。他們獲得信用的途徑往往是通過央行中國人民銀行，或是與大型國有企業之間的關係。

Across the broader population, lending tends to be more personal and informal — cash loans from networks of friends and relatives.

在更多的中國民衆那裏，貸款往往具有更加個人化的非正式性質——從親戚朋友那裏借錢。

But China’s leaders are seeking to stimulate consumer spending to make its economy less dependent on industrial exports. Expanding consumer credit is part of the formula, and the government is allowing private companies, like , to innovate.

但是中國領導層正在努力刺激消費，以使中國經濟減輕對工業出口的依賴。擴大消費信貸是整個策略的一部分，政府准許如京東這樣的私營企業在這一領域進行創新。

Since early 2014, had been offering its own consumer loans of up to a few thousand dollars for purchases of televisions, smartphones, computers, refrigerators and other merchandise. ’s business model is sometimes compared to a combination of Amazon and UPS.

自2014年初開始，京東一直給它的用戶提供貸款（最高達幾千美元）用以購買電視、智能手機、電腦、冰箱和其他商品。京東的商業模式有時被比作亞馬遜(Amazon)加UPS。

Like Amazon, the company buys goods from manufacturers and has a national network of distribution centers and warehouses. It also has its own fleet of delivery vans. handles more than two million orders a day, and offers next-day delivery in much of China. It is a full-service online retailer, unlike its better-known rival, Alibaba, whose marketplace connects buyers and sellers.

和亞馬遜一樣，京東也是從製造商那裏進貨，並建設了全國性的物流和倉儲網絡。此外，它還有自己的廂式送貨車配送隊伍。京東的日均交易處理量達200多萬單，在中國大部分地區可實現下單次日送達。與它更爲知名的對手阿里巴巴（其業務領域是作爲一個平臺，在買家和賣家之間搭橋）不同，京東是一個提供全方位服務的在線零售商。

In its test run for the Chinese company, ZestFinance built risk models using transaction data: what people buy, when they buy it, what brands they choose, where they live and other nuggets of information in the sales data.

在爲其中國公司進行測試時，ZestFinance利用京東的交易數據——包括人們買什麼、何時買、選什麼品牌、住在哪裏，及交易數據中其他有價值的信息——建立了風險模型。

“There’s signals in there,” Mr. Merrill said. “But what would seem like simple signals can actually be very complex.”

“這些數據裏有一些信號，”梅里爾說道。“但那些看起來簡單的信號，實際上可能非常複雜。”

For example, one might expect that a person purchasing a lot of luxury goods online is a good credit risk. But Mr. Merrill said that often is not the case. It could be a sign of reckless overspending or even fraud, he said, when linked with other data.

比如，人們可能覺得在網上買很多奢侈品的人信用風險小。但梅里爾表示，情況往往並非如此。他說，跟其他數據聯繫起來看，這可能意味着不計後果地過度消費，甚至可能是欺詐。

If a person is making purchases during the day, that could be a signal that the buyer is unemployed. But, Mr. Merrill said, if the purchases are made during the midday lunchtime, from an office computer, it could well be a sign of a hard-working employee squeezing in time to buy necessities.

如果一個人是在白天時間買東西，可能表示這個買家沒有工作。但如果交易是在午餐時間發生，而且是在辦公電腦上進行，梅里爾說，那就很可能代表這是一個勤奮的員工在擠時間買必需品。

In its test, the creditworthiness predictions made by ZestFinance were compared to the results of ’s experience making loans, which was essentially the control group. The ZestFinance algorithms won handily.

在測試中，ZestFinance所作的資信預測，與京東自身放貸的結果作了對比，後者實質上就是對照組。ZestFinance的算法輕鬆勝出。

The Chinese online retailer, said Josh Gartner, senior director for international communications for , hopes to “greatly improve the efficiency of deciding who should be offered credit or not.”

京東國際公關高級總監約什·加德納(Josh Gartner)表示，京東希望能“大大提高其貸款決策的效率”。

Data science methods, Mr. Gartner added, can fill a gap “where traditional metrics tend to be less useful, and China would obviously be one of those places.”

加德納補充道，數據科學的方法可以在“傳統衡量方法表現欠佳的地方”填補一個空白，“中國顯然就是一個這樣的地方”。

In a statement, Mr. Chen pointed to the potential value of the joint venture beyond itself. He called the link-up with ZestFinance “a foundational step toward building a reliable system for assessing credit risk that will help meet the huge market need.”

陳勝強在一份聲明中指出了這一合資公司在京東之外的潛在價值。他將京東和ZestFinance的聯合描述爲“在建立可靠的信用風險評估系統，從而滿足廣闊的市場需求方面，是基礎性的一步。”