April 10, 2012 at 11:46Client analytics: Big Brother knows what you will buy tomorrow and when you change provider
If you know that one of your customers will wake up tomorrow with the thought of buying a new tablet, you can send him an email with a discount code today. If you understand that by all indications the client is going to switch to another provider, you can increase his speed, lower the price or offer something else. This is customer analytics .
If you don’t know which of the three proposed tariff plans should be launched throughout the country, you need to use client analytics, which will take each individual person from the database, evaluate the emotional and practical motives for the transition, and make it possible to understand how many people will use this tariff.
This is what Data Mining looks like in client analytics.And just like that, it already works in practice in hundreds of large companies around the world and in our country.
In the last topic, I already wrote about how much client analytics can transform a call center when you know everything about the caller. This is a piece of the big puzzle about customer analytics.
Not all subscribers are equal in terms of marketing: they differ in profitability, the list of consumed services, loyalty or aptitude for leaving. You need to be able to share audiences and work with each segment separately.
Why? Because the more accurately we focus on the client, the greater will be the return. For example, you can offer a particular service or product not to everyone, but to those who are most likely to accept the offer by segmentation.
Does this work in practice? Yes. In recent years, I have been implementing tools for analyzing client requests and I can say that this is definitely a very powerful tool for large businesses.
The basic parameter - the value of subscribers - can be determined by a set of these data:
Next, you can highlight the groups that you should work with in the first place. This is determined on the basis of the current strategy of the company (capturing market share, maintaining a customer base, increasing profitability, and so on) and the tactical situation (degree of customer satisfaction, quality of services, and more).
An individual message can be generated for each individual client, taking into account his profile.
Below is a typical client’s departure schedule and a new schedule of the situation in which a trend was detected in time and the client returned.
In my practice, there was a case where only the program for finding the right time to contact the client (and searching for the best offer) allowed me to increase the profit from the loyalty program by 4 times. Previously, the loyalty program worked, but it did not work very accurately. Here we are talking about more accurate focusing of the stock (early detection of a trend, understanding how much a client can make profit, plus drawing up an optimal offer that increases the likelihood of retention or return).
In a broader sense, what we are doing is looking for customers who can offer something a day before they go looking for it themselves. The second option is to find market areas where the company has a clear advantage and quickly reach all potential customers.
Integration of this piece with CRM allows you to deliver analytics exactly on the spot for operational decisions: for example, it really works for communication providers when choosing a tariff, in a bank when choosing options for an account, in retail - for product offers, and so on.
Here is an example of a calculation piece when a client needs to offer one of two products (for example, tariff plans):
There is a combination with a marketing engine. For example, you can stimulate a customer to make new purchases based on customer profiles that are similar in behavior (what would you buy tomorrow if you behaved like the whole group?). There is also a response engine, where you can submit data such as conditions for marketing campaigns, and at the output get a sample of customers for whom each stock will be optimal. The result is lower holding costs, high returns.
Another interesting thing is testing hypotheses about strategic decisions. For example, you need to calculate the cost of implementing a very expensive loyalty program: you can get a profitability forecast. More broadly, you can evaluate the entire existing customer base and understand the company's capitalization. Another interesting example: when selling a business, you can take into account all the potential income from the client base.
On the one hand, the more data sources, the more accurately you can segment the database and select offers for each segment. The first scenario - the amount of data is limited, and processing takes place offline (since the process of collecting data from all sources is gradual), this is the last century and poor integration. The second scenario - working with all data sources in the "natural environment" and in real time - it is with these predictive solutions that I now work.
As a rule, such systems are implemented in the infrastructure of companies with 1 or more million customers, but often they work successfully starting from 100 thousand customers.
Is there such a practice in our country? Yes, definitely. I have personally been involved in integration for telecom operators , banking companies, and retailers and I know about many similar projects of my colleagues, so you can be sure that not only Google uses your data to offer you something.
Is it possible to build such a system yourself for a small business? Yes, you can: you already know the basic principles, deep integration is not needed, infrastructure, unlike market giants, rises “on the knee”. In general, if you use the most obvious things, there is a chance to quickly increase efficiency.
If you don’t know which of the three proposed tariff plans should be launched throughout the country, you need to use client analytics, which will take each individual person from the database, evaluate the emotional and practical motives for the transition, and make it possible to understand how many people will use this tariff.
This is what Data Mining looks like in client analytics.And just like that, it already works in practice in hundreds of large companies around the world and in our country.
In the last topic, I already wrote about how much client analytics can transform a call center when you know everything about the caller. This is a piece of the big puzzle about customer analytics.
Let's start with the basics: working with segments
Not all subscribers are equal in terms of marketing: they differ in profitability, the list of consumed services, loyalty or aptitude for leaving. You need to be able to share audiences and work with each segment separately.
Why? Because the more accurately we focus on the client, the greater will be the return. For example, you can offer a particular service or product not to everyone, but to those who are most likely to accept the offer by segmentation.
Does this work in practice? Yes. In recent years, I have been implementing tools for analyzing client requests and I can say that this is definitely a very powerful tool for large businesses.
What can be used as source data?
The basic parameter - the value of subscribers - can be determined by a set of these data:
- Tendencies to continue to use the services of the company or go to another based on the average term of service of the client of the group and the elapsed time.
- Social status based on billing, loyalty programs and social graph.
- According to the current portfolio of services (order history).
- Based on the predictive model - according to the list of services or goods that will be most likely to be used by the client in the future.
- Loyalty (based on the history of events from CRM).
- Plus for dozens of parameters that depend on a particular business.
Next, you can highlight the groups that you should work with in the first place. This is determined on the basis of the current strategy of the company (capturing market share, maintaining a customer base, increasing profitability, and so on) and the tactical situation (degree of customer satisfaction, quality of services, and more).
An individual message can be generated for each individual client, taking into account his profile.
Retention and Refunds Example
Below is a typical client’s departure schedule and a new schedule of the situation in which a trend was detected in time and the client returned.
In my practice, there was a case where only the program for finding the right time to contact the client (and searching for the best offer) allowed me to increase the profit from the loyalty program by 4 times. Previously, the loyalty program worked, but it did not work very accurately. Here we are talking about more accurate focusing of the stock (early detection of a trend, understanding how much a client can make profit, plus drawing up an optimal offer that increases the likelihood of retention or return).
In a broader sense, what we are doing is looking for customers who can offer something a day before they go looking for it themselves. The second option is to find market areas where the company has a clear advantage and quickly reach all potential customers.
Integration and practical use
Integration of this piece with CRM allows you to deliver analytics exactly on the spot for operational decisions: for example, it really works for communication providers when choosing a tariff, in a bank when choosing options for an account, in retail - for product offers, and so on.
Here is an example of a calculation piece when a client needs to offer one of two products (for example, tariff plans):
There is a combination with a marketing engine. For example, you can stimulate a customer to make new purchases based on customer profiles that are similar in behavior (what would you buy tomorrow if you behaved like the whole group?). There is also a response engine, where you can submit data such as conditions for marketing campaigns, and at the output get a sample of customers for whom each stock will be optimal. The result is lower holding costs, high returns.
Another interesting thing is testing hypotheses about strategic decisions. For example, you need to calculate the cost of implementing a very expensive loyalty program: you can get a profitability forecast. More broadly, you can evaluate the entire existing customer base and understand the company's capitalization. Another interesting example: when selling a business, you can take into account all the potential income from the client base.
Monitoring Results
On the one hand, the more data sources, the more accurately you can segment the database and select offers for each segment. The first scenario - the amount of data is limited, and processing takes place offline (since the process of collecting data from all sources is gradual), this is the last century and poor integration. The second scenario - working with all data sources in the "natural environment" and in real time - it is with these predictive solutions that I now work.
So what does this give?
- You can look into the future in terms of tactical, operational and strategic decisions.
- The system allows you to rely on accurate data, rather than intuitive.
- You can work out the base very deeply and accurately, in fact, build the optimal model for each client.
- Business analysts rejoice because the output is data that is understandable even to a high school student.
- All this works in real time based on available data.
- Relatively easy integration with any IT environment is made.
- Calculations allow you to cover your ass: it’s better to count and do than not to count and sit still.
As a rule, such systems are implemented in the infrastructure of companies with 1 or more million customers, but often they work successfully starting from 100 thousand customers.
Is there such a practice in our country? Yes, definitely. I have personally been involved in integration for telecom operators , banking companies, and retailers and I know about many similar projects of my colleagues, so you can be sure that not only Google uses your data to offer you something.
Is it possible to build such a system yourself for a small business? Yes, you can: you already know the basic principles, deep integration is not needed, infrastructure, unlike market giants, rises “on the knee”. In general, if you use the most obvious things, there is a chance to quickly increase efficiency.