What is the "concentration" in customer data and why is it important

If you've been looking at "life poison" or some other series of videos all weekend, you can enjoy it, because you're not alone. Now everyone is in a "centralized" way of consuming electronic products and services, this way in the long term, consumers will be in a short period of time to concentrate on the purchase of products.

Eric Bletterau, Eric Bradlow, a professor of marketing at Wharton, says that once marketers are aware of the phenomenon and they have the data to follow up on the phenomenon, they can discover the development of a rich new type of electronic behavior. He points out that one of the basic tools for measuring customer value over the decades has been the RFM model, the most recent consumption (recency), consumption frequency (Frequency) and consumption amount (monetary value). "However, my research suggests that the system is not comprehensive and should be coupled with a ' C ' that represents concentration (clumpiness)," he added. Bletterau The concept in an interview with Wharton Online, a concept derived from his research paper, "a new standard for measuring the crowding of Influence Data", co-authored by the Swiss credit group partner Zhang Yao and the Wharton School of Data Professor Duren Smow (Dylan Sgt). Clumpiness for incidence Data.

Following is an edited interview excerpt:

What is the "concentration" in customer data and why is it important:

The most common practice in the field of marketing and customer value is through the RFM model, that is, the most recent consumption, consumption frequency and the amount of consumption to summarize the consumer behavior, that is, using all the information about customers, to calculate the customer in the following three aspects of the data, that is, the last time consumption is when, How often and how much money you spend ... This is the basic metric that most companies use to measure valuable customers and other customers.

My research found that this is not all the characteristics of customers, RFM model should be added with the letter "C", representative concentration, indicating that some customers purchase behavior has a certain regularity. For a long time, customers buy orange juice, diapers and other items will follow certain rules. But concentration refers to people scrambling to buy goods. This period of concentrated buying can reflect the distinctive characteristics of customers, who may be extremely valuable customers.

Important Conclusion:

The important conclusions drawn from my study are easy to understand. Imagine that you want to predict who is a valuable customer in the future. You have four indicators that you can use to predict: the most recent consumption, frequency of consumption, amount of consumption, and marketing to customers. This is the scoring model that many companies design. But I think you need to add one more indicator, which is concentration-measuring how high a customer's purchase is. This metric is as easy to compute as the R, F, and M metrics, and can be done in Excel and calculated quickly to compute 100 million customer data in one second.

My research also found that in the experiment the concentration of customers with higher levels of value in the future, even in the use of RFM model and marketing cost analysis, will come to the same conclusion. This means we've found another variable for the customer, and the company should follow up on the variable and use it to predict the value of the customer in the future.

The most surprising conclusion:

I was amazed by the two points in the study's conclusions. One is that I found out that this model based on RFM has been around for many years now and is being used by many companies, and in this sense has been recognized as the easiest way to help explain customer value. You can do a variety of web search, you can set up a variety of other variables, but the concentration of this variable is the most simple.

So first of all, let me be surprised that the concentration has been no attention, in other words, hot and cold period can reflect the characteristics of customers. The other thing that surprises me is that from the data I've analyzed, this concentration applies to electronic and online consumer goods, not to conventional consumer goods. In other words, I can see why the traditional model is suitable for conventional consumer goods, because consumers buy toilet paper, orange juice consumption behavior is regular, but on the site Hulu consumption is not regular. Consumer Auctions on ebay or buying books on Amazon have no regularity.

Concentration does not apply to traditional goods, but it is a compelling data set for emerging products and emerging economies that I have analyzed.

Practical value:

I think this is probably the most practical research I've done in my more than 20 years of career. I am more inclined to call my work a complex statistical model. This work is not just about statistical modeling, but also about data-concentration, and today's companies can do real operations without the need to collect additional data and use the same data that calculates R, F, M, and customer lifetime values. Use this model to determine how much the customer's predictive value is, how the customer's value ranking will change, and how you will change your understanding of which customers have a redeeming value.

I have come to the conclusion that while the highly concentrated customers are not stable, they are still the most rewarding. If you win these customers again, they will become highly concentrated and will spend a lot of money in the future. So I think this model has great practical value. The glamour of this model can be seen on my website with an Excel table that lists some success stories. You can download this form and start using the focus today to analyze it.

What new rules, procedural changes or tricks can you draw from your research?

Nowadays many people are talking about big data. I like big data, but I want to say there is something more to my liking than big data, that is data enrichment. Data enrichment refers to the fact that you can now collect thousands of variables from people and identify the source of these variables, what items they buy, what pages they browse. But this cannot be called science, it is merely collecting data. The question is what information is really useful for solving immediate business problems? This is what I call data enrichment.

So I think of concentration as a supplement to traditional variables like RFM, marketing campaigns and similar activities, a form of data enrichment. What I want to emphasize is that you need to master more data and not just condense things to three variables, but condense them into four.

I've done a lot of research on concentration, and I know it exists in various industries and has predictive value. But I don't know what spawned the concentration degree ... I link marketing activities to concentration. The company can make you spend your brain by writing emails, sending pictures, making you a target customer.

Although I know the presence of concentration, I have not yet done research on the best way for companies to locate targeted customers. For example, I never paid attention to whether consumers would have a higher concentration when buying a series of products. Imagine what it would be like to watch "The Geek", "Mad Men" or a similar series of videos. If, for example, a company tries to sell a series of products such as facial care, moisturizing series and other similar products, will it pack all the products into a suit so that it seems to have the moral to work towards a goal?

I know how to do math on concentration, and I know it's trivial for companies, but it's predictable. But I don't know how to explain it psychologically, and this is one of the areas that I've been studying with other research colleagues in consumer psychology. We're going to do a lot of behavioral experiments in the lab, trying to get the psychology behind the people's concentration behavior.

"Surrounded" vs. "Crazy"

I like the word "focus", others like "crazy". The reason I like this word is that it's relative to "not concentrating," like a regular way of getting into a place or shopping regularly. We don't hear stories about concentration, but when you see stories about people frantically doing something or consuming something like, "a student stays up all night watching videos for 18 hours," You know what "crazy" means. The concept of "madness" is also persuasive: every time I talk to managers, students, or college students about the concept, everyone believes it exists.

Dispel misunderstanding:

My research is to dispel the misconception that, to some extent, the customer's classification is based solely on a series of numbers. We need to think about more than just the simple theory of human behavior that I'm talking about. If we study the last three indicators of consumption, consumption frequency and consumption amount, we will find that the basis of this traditional consumption behavior ignores the time interval I propose. It shows that fundamentally, I can collect all the data, such as the consumer at this window of the consumption interval is two days, then four days, three days or six days, but I can ignore all of this, just remember the consumer last time to spend and the number of times they come to spend.

In addition, there is a misunderstanding: there is insufficient information about the law of consumers to shop. In fact, the consumer to store information is very detailed. People flooded into stores, left, poured in, and left ... These people are very different. I think that people can be divided into two types: centralized and non-centralized.

Our results also show that concentration varies depending on the type of product. For example, we found that women are more concentrated than men, and young people are older. Therefore, I think we should clarify that people are not only born equal, but there are many simple ways to divide people into several types.

How this study is different:

Although a number of mathematical models have existed for 50 of years, they are still popular, and in the past 10 they have been called "Hidden Markov Models" (hidden Markov model). Suppose there are two countries in the world, a hot country, a cold country. You take turns to stay in hot and cold countries. The mathematical model is concentration, you are in hot country, you will do a lot of things, in cold country situation is diametrically opposite, so in hot and cold alternating ... All I have to do is provide a way for professionals to compute simple numbers, that is, statistics. This is not a statistical paper, but a treatise on numbers and statistics that we are talking about.

You just count the numbers and use the numbers to do your own research, use it to predict customer value, determine whether men are more concentrated than women, and use it to divide people into different types. This is typical and unusual for this study, a metric based approach that insiders can use rather than a variety of models, but both are designed to solve the same problem.

Examples that might be useful in proving "concentration":

We have conducted research on centralized consumers to prove that email, brochure or other different types of marketing channels are more effective. We found that e-mail has not unexpectedly become a short-term effect of a better means of publicity, and the long-term effect is better publicity pictures.

What remains to be seen is whether some of the words in emails, brochures, or video marketing are related to people's concentration or have a higher degree of concentration. are some topics of concentration? are some product types necessarily related to concentration? What we are doing is proving the existence of this phenomenon, And I know that many industries have this phenomenon, and some types of people are more likely to become concentrated people.

The task that I have yet to complete is to discuss the marketing implications of concentration, and as a marketing professor It is really surprising that this has not been done. I also need a larger and newer set of data to be used to link information about marketing activities to people's centralized consumption behavior. I know how to do it, but I need richer, better data to do it.

Follow-up work?

I am considering following up on three different aspects of the study. For the first time, I am thrilled to be able to analyze more datasets and prove how persuasive the metrics of concentration are. I analyzed data from Amazon, CDNow, EBay, Hulu, YouTube and other traditional consumer goods companies. Now I just need to apply this data to the new dataset.

Secondly, I want to understand the psychological process of people. Why do people's consumption appear to be centralized? The third and final aspect is that I want to link marketing activities to concentration. It's not just about how people behave, what they do, what websites they browse, what things they buy, but also about the marketing campaigns themselves, perhaps the promotional leaflets for marketing campaigns, and the channels through which they are distributed. All of this helps me find the best solution-to provide companies with optimized marketing campaigns to activate people's consumption concentration.