Why do you say that 99% of smart collections are gimmicks?

the collection of consumer finance industry is to seek a dynamic balance between cost and income.

ignoring the cost premise to talk about collection, there is no comparability. Intelligent collection is a partial cost reduction and efficiency improvement, and the methodology of the whole collection has not changed.

the optimization of collection must be combined with the change of bad debt rate, and the cost optimization under the premise of keeping bad debts unchanged is the real cost reduction and efficiency improvement, otherwise it is to give up the cost.

in financial companies, the biggest risk is people.

there is a saying in the financial industry that "three points for loans and seven points for management".

whether it is a bank or a mutual fund institution, the importance of collection is reflected in two aspects: first, it can minimize the loss of bad debts; Second, through strong collection ability to seize the beach with higher risk business and gain income.

the default risk of mutual fund institutions is higher than that of credit card customers, and the slight change of collection effect may cause millions of profits or losses.

With the rise and fall of cash loans, the collection industry seems to have ushered in a spring.

On the one hand, the number of customers has exploded rapidly; on the other hand, customers are becoming more and more difficult to urge, and industrial bad debts are soaring.

as a new way, intelligent collection emerged.

AI is hot in 218. Is intelligent collection really the future of the collection industry? Not necessarily.

what is intelligent collection?

According to public reports, intelligent collection mainly uses artificial intelligence technology to optimize the whole collection process.

Let's take the whole collection process as an example. The following is the definition of overdue indicators for general collection. Usually, 9+(M3+) overdue is defined as bad, which is what the industry often refers to, and 18+ overdue is defined as bad debts.

Of course, the definition of bad debts and the mechanism of accrual and write-off are different for each company, so I won't mention them here for the time being.

Definition of overdue collection indicators

In this process, a few days before each repayment date (m1,m2,m3, etc.), users are usually reminded by telephone that the repayment date is coming, so pay attention to timely repayment. This way, Alipay, IOUs and credit card centers of various banks are all adopting it.

from m1, once the repayment is not made, it will become overdue. Generally, a few days after the deadline, a telephone reminder will also be made. Here, it is not to remind the repayment, but to urge the repayment. Therefore, the intensity of the whole collection in the above picture will gradually increase until it is overdue and will be disposed of by very strong means such as justice and outsourcing. Of course, the cost will be high.

in the whole collection process, we should constantly analyze overdue customers through data reports, formulate targeted collection strategies, and constantly adjust according to the monthly collection indicators.

among them, it includes the work of calling, outsourcing, quality inspection and so on.

Simple diagram of collection process

What links have been changed by intelligent collection?

Now, the main innovations mentioned in the intelligent collection technology are the following links:

1 Intelligent call (incoming and outgoing calls)

2 Intelligent case division (mainly involving collection strategies)

3 Intelligent report

4 Intelligent quality inspection (quality inspection)

Intelligent call: greatly reducing labor costs? Not necessarily

There are two kinds of smart call. One is typical recording and broadcasting, that is, the speech information of voice robots is manually recorded and imported in advance, which is very similar to Taobao's robot customer service.

the other is real-time artificial intelligence voice reply, which can be understood as siri of iPhone or Alexa of Amazon for the time being.

the core of intelligent customer service is to increase the degree of reminding users, increase the frequency and reduce the cost.

traditional collection companies use manual telephone. Many financial companies will outsource this business, and the outsourcing will allocate the collection amount of financial companies according to the number of seats. One seat is one person. The average daily per capita production capacity in the industry is about 2-3 tons.

as for the profit distribution of the collection company, the rebate is basically given according to different proportions of the amount to be urged.

Let me remind you not to be attracted by the samples of many artificial intelligence companies in official website. Generally speaking, official website is a post-polish version. If you want to know how the robot customer service is doing in the collection business, you must do a test.

There is a view in the industry that artificial intelligence significantly reduces labor costs.

From the point of view of the cost of each order, it does have a greater advantage than labor, but if we combine this cost reduction with the late bad debt collection index, the result may not be so optimistic.

At present, different charging standards of the mainstream call platforms in the industry are collected.

The method of manual dialing is basically 2 yuan/Tong according to the dialing rate of each call in the industry. At present, the mainstream charging method of artificial intelligence instead of manual dialing is based on the length of use, which is basically one yuan per minute or about .5 yuan.

Since the average length of each call of a manual telephone is basically about one minute, we can compare it approximately uniformly here.

At present, in the field of artificial intelligence, Iflytek is doing better, and its fees are relatively higher than other suppliers in the industry. Let's take a middle range, 1.5 yuan. Other suppliers also follow the median, .5 yuan.

Let's assume that we need to make 2 phone calls every day as an example, and the different expenses are as follows:

Note: The bad debt rate here refers to the later index changes caused by different collection methods. The source of artificial intelligence data is the business test data made by several mainstream domestic consumer companies.

Pure labor will reduce the bad debt rate by at least one point compared with artificial intelligence. Repeat, do a business test, and don't believe the description in any company's official website. If it is assumed that the amount to be urged is 3, yuan/day, it will be around 9, yuan a month.

If artificial intelligence is selected:

If the bad debt drops by one point, the loss will be 9, * 1% = 9, yuan.

cost reduction, saving: (4-1) yuan *3=9 yuan.

basically tied.

if we lower the unit price of artificial intelligence to .25 yuan. That cost can save 35 yuan * 3 = 1,5 yuan and 9, yuan.

if the gap of bad debts is not one point, it becomes two points. 9, * 2% = 18, yuan and 1,5 yuan.

the above numerical reasoning basically adopts the business average in the industry, and there will be up-and-down deviation, but the amplitude will not be too large.

From the above simple deduction, we can draw at least three conclusions:

1. In terms of unit price, artificial intelligence has obvious advantages over pure labor, which can greatly reduce the average unit cost, no doubt.

2. It remains to be seen whether artificial intelligence is more cost-effective if the late collection effect is integrated and the bad debt rate caused by different ways is also taken into account.

3. From a single point of view, the loss caused by the rising bad debt rate is far more terrible than the cost reduction.

finance is a lagging business with strong risk control, and all changes must take into account the possible changes in bad debts.

if the bad debt rate is not mentioned and the optimization is unilaterally carried out, it is tantamount to putting the cart before the horse.

Intelligent case division (collection strategy): The variable of people cannot be controlled, and the value of artificial intelligence is not great

This is the core work of the whole post-loan collection. What kind of collection methods and frequency do different customers touch at different times? According to different data feedback and industry experience, an optimal collection plan between cost and recovery rate is formulated, which is the value of the strategy.

collection strategy basically involves the following factors: people (collectors), when (collection opportunity) and how (collection intensity/frequency) to handle different cases (collection case characteristics).

whether it's automated wechat, SMS, telephone reminder or manual collection, people need to design the strategy. So far, artificial intelligence can not replace the role of people.

intelligent reports: automated reports have long been popular in the industry

reports run through the whole collection management. At present, the automatic report provided by artificial intelligence is not much different from the report built in the collection system.

for enterprises that are basically MIS systems, the collection system can provide automatic report generation in different dimensions. Many collection systems peddled by third-party enterprises like Huateng in the industry all contain built-in reports, and this technology has matured for many years.

At this point, artificial intelligence has no obvious advantage, and even whether the report level of artificial intelligence can be equal to that of MIS reports needs further observation.

intelligent quality inspection: artificial intelligence is superior to labor

in the field of quality inspection, at present, artificial intelligence has a leading edge.

whether we build our own collection team or outsource it to a third-party collection company, we need to take recordings regularly to control the quality of collection calls, which is called quality inspection. For example, no abuse, threats, clear business statements and so on.

in the traditional quality inspection method, a certain amount of samples are randomly selected from the collection recording library, and a group of quality inspection specialists are recruited to listen to them one by one, record the problematic cases and calculate the proportion.

In the case of artificial intelligence, the sampled voice samples can be converted into text by voice to text, and then the keyword information can be retrieved automatically. At this point, in theory, artificial intelligence can greatly reduce costs.

but there is a problem here, but actually speaking, the quality problem of collection is that there is something wrong with the collector. When the business scale is up and the team is basically stable by manual quality inspection in the early stage, a normal collector will not suddenly deviate from the quality.

this quality inspection is not completely random sampling, and it can be intervened through the management of people, and there is also a law of large numbers. Therefore, as the scale increases and stabilizes, the effect may not be so great.

the truly effective system in the collection field: automatic dialing system

automatic dialing system: after obtaining the expected account list, the accounts that need telephone calls are sorted according to the set rules, and the call sequence is optimized by algorithms to maximize the productivity of each agent.

automatic dialing system is a must for large collection agencies.

Here are some business figures before and after using the system:

Note: The figures in the figure come from the book The True Classic of Consumer Finance

But the application of the system is highly dependent on the professional skills of post-lending strategists. The rules of the automatic dialing system are determined by the managers, so as to ensure that high-quality collectors can answer as many calls as possible, and at the same time maintain a high telephone connection rate.

for example, how to ensure that an excellent collector will send him the latest case as soon as he hangs up the phone, seamlessly connecting. In addition, the automatic dialing system is not a new product brought by AI in the past two years. This is a very mature industry product, which has been used by large credit card centers at home and abroad for many years.

However, this system also has obvious defects. Basically, it takes more than 1, people to make it necessary and possible to use it. Too small a collection company, purchasing such a system, first, the cost is too high, and second, the automatic dialing system needs to rely on different variables in a huge database to make the algorithm make the most decision. If the sample size is small, it doesn't mean much.

what is the core of collection?

collection is a labor-intensive industry, and the core of collection ability is whether it can recover the arrears as low as possible and efficiently.

the core of collection lies in the formulation and practical application of collection strategy. Under certain cost control, how to minimize the potential risk of overdue or the actual damage of overdue behavior. The business ability of a professional modeler and strategist is considered.

as shown in the above figure, in the whole post-loan collection process, there are actually completely different strategies for collection intensity and collection effect.

for consumer finance companies (except cash loans, which have completely different business models), it is not necessary to adopt the strongest collection intensity at the beginning of overdue, but the collection effect should be the best in theory.

As time goes on, the difficulty of collection gradually increases, and the possibility of bad debts is further increased. At this time, the collection should be strengthened.

another important income of collection is fines. Even under some strategies, it is intentional for financial institutions to make a specific part of customers overdue or overdue for a long time.

Financial institutions are not stupid. There are reasons for doing so. Because good collection must be the best balance of bad debt loss, fine income and collection cost.

The above figure simply lists the comparison of three different collection strategies.

the best collection is to make the positive value as large as possible and reduce the negative value. These three factors are interrelated, and it is not easy to affect the whole body.

There are two core links: identifying high-risk and low-risk customers and establishing collection score. At least as far as the current status of mainstream collection agencies is concerned, artificial intelligence can't replace people.

traditional financial institutions will basically send high-risk customers to the most experienced collectors to deal with. Because for high-risk customers, institutions can't wait aimlessly, or increase the intensity of collection after the overdue time is extended, but must try to adopt high-intensity collection strategies at an early stage to get customers to pay back the money.

and how to identify high-risk customers among customers, which requires the use of collection scoring system. The collection scoring system can be based on past customer performance and statistical methods, plus industry collection experience (including changes in the repayment probability of users that may be caused by different collection strategies, etc.).

artificial intelligence has indeed improved efficiency in some single points, but this improvement needs to be taken into account in the overall collection cost. Otherwise, it is meaningless.

In addition, there is still a big gap between the current artificial intelligence outbound call and the real person. When more and more customers recognize that this is a robot talking, will it have the opposite effect?