Mining website user information

How much do you really know about your online customers? If you think that the data you collect through server logs is enough, think again. Usual server logs don't provide enough information about your visitors to support you in making important strategic decisions.

In today's fast-paced economy, companies must be able to develop and mine value-added data through their websites, and combine this data with offline information to provide managers with the ability to move from "score statistics" to strategic analysis.

If your business uses a multi-channel distribution or communication approach, you must examine how your website performs compared to various direct (letters, phone calls, emails, etc.) and indirect (distributors, retailers, etc.) channels . You can't judge a website to be profitable just because the revenue is greater than the cost; instead, you need to encourage the use of more efficient channels to generate revenue based on customer preferences and business policies. With the proliferation of customer touchpoints, the focus of customer analytics has shifted from detecting sales to understanding trends.

Be clear about your goals

Obviously, when you collect and analyze data from your website, you must take into account the more complexities that come with finer site design, greater traffic, and channel expansion. You can no longer rely solely on log file data.

Traditional direct marketing principles can help you here. They help you understand what it means to define data. Data definition is the first step in analysis by which individual behaviors can be tracked, responses and profitability models can be modeled, customer categories can be established and their lifetime value estimated.

The main problem with applying direct marketing principles to a website is that they evolved in a strictly controlled environment. Website users themselves choose search engines, links, advertisements or other mediums. They enter or leave the site anytime and anywhere and decide for themselves what to watch. The user, not the website, is in control. This means that, in fact, there can be an infinite number of variations of data elements, and while variations are good for analysis, too many variations become meaningless. Also, storing every possible variable leads to memorizing too much data and consuming too much processing time.

Does this mean that direct marketing principles have no place on the Internet? No, this means that these principles must be actively extended and applied during the conceptualization and design phase of the website.

Achieving this goal requires three basic steps: define goals and key actions; define user criteria; and classify according to customer status.

The core of knowing what your goal is to know what data to collect and how to define it. However, it is more important to determine which actions to measure, as goals overlap more easily than actions. The reason for the overlapping goals is that every company needs to accomplish something similar and understand who a customer is, how to track online browsing and shopping habits, how to manage content, how to change advertising and promotion strategies, and how to build personalized service.

Defining Customer Behavior

But while every company needs this knowledge, every business acquires it in different ways, according to its philosophy and values. Only when you identify which actions are unique and critical in helping to achieve your business goals and understand how to measure their impact can you ultimately create differentiation.

Consider the following categories of customers: If Jeff bought an item last year, we call him a "customer." But what if he bought something 8 years ago?

Rick bought something 3 months ago but returned it and got the money back. Is he still a "customer"?

As for Tina, she bought something on the same day as Rick and asked for a refund instead of returning the item. What is she?

Don't forget Douglas, he paid the first installment for the walker and never heard of him again.

And two new customers, Tom and Al, who just received their first mail order item, didn't pay in time, but already ordered another item from the business website.

While each definition has its own strengths, it is also apparent that future analyses based on them will vary widely. The easiest solution is to confirm a definition. But what if you change your mind after half a year? The core of the problem is that it is not enough to know the characteristics of individual customers, it is really important to understand their behavior, we need to define their actions.

In the above example, each customer performed or omitted an action, the nature of which can be captured. You can store this information in an order and design different customer processing plans based on the order status. In a multi-channel environment, you need to take the extra step of combining order status information from different channels to assign an overall customer status score to drive your customer contact strategy.

Categorize customers

Once the customer actions and corresponding data elements have been identified, the next step is to determine how to summarize them. This is the key to condensing the data, because that's when you're determining the data category. These categories are based on goals or actions. Example:

User Motivation. If the site is primarily goal-oriented, such as searching the site, it is important to aggregate data at the traffic level. However, if the goal of the site is to measure sustained engagement, cumulative metrics are more appropriate.

User classification. Aggregate data at multiple levels based on customer status.

Content. Website owners who rely on advertising and links for income are categorization experts. They collect users with common interests and work to market them to merchants who want to sell them. To be successful, these sites must aggregate visitors based on metrics such as content, length of time exposure to a particular ad, links clicked, and more.

Regardless of the method of aggregation, you must maintain consistency in your metrics. If you track sales at the individual customer level, you should also include costs.

Gather key data

When it comes to online behavior, you're going to have to keep your fingers crossed. In collecting website data, this goal is to collect only critical data that is compatible with other data sets and that is easy to analyze in a form.

For example, don't collect the entire URL address, but save information based on page type or content category. This approach reduces processing time, reduces storage requirements, and makes it easier to identify critical paths.

Obviously, website management is impossible without knowing what users are doing or looking at. The only way to reveal it is to directly intervene in the interaction flow. Fortunately, this is fairly easy: all you have to do is add instructions for what data to collect and where to store it.

The beauty of this process is that you're collecting data at the point of interaction. You can go back to steps 1, 2, and 3 of the conceptual design and use the criteria established at the time to help create your own instructions and extract only the data you want. The downside of this method is that since the original data is not preserved, you cannot corroborate or adjust the data later, and additional data processing can slow down the website. The more data you collect and control, the greater the potential negative impact on website performance. Therefore, decisions about website building and data development must be synchronized.

A very basic database structure for a website that accepts orders and collects user interaction information might look like this:

A database accepts and returns information related to inventory.

Database B identifies customers and returns customer-customized content based on historical information.

The C database mainly acts as a container. It accepts clickstreams and content summaries from unidentifiable customers.

From a technical and analytical standpoint, it makes sense to have separate databases perform different types of work. Each database can be tuned and optimized for its specific purpose. But care must be taken when you design such a division of tasks to ensure that it still makes sense to spread the data across multiple databases. It's important to keep the categories consistent.

Doing in-depth analysis

Bad information leads to bad decisions, and lack of information leads to guesswork. It is easy to confuse having the amount of data with having the amount of information. Sadly, building a framework for data collection and being able to boast about having high-quality data doesn't make you any smarter. But it does do what was previously impossible: build confidence in the integrity, accuracy, and relevance of data. This method removes speculation and minimizes bad decisions due to bad data.

Once the integrity of the data has been established, the next step is in-depth analysis. Often, the first two things businesses want to know are: how to convert prospective customers into customers; and how to build customer loyalty. Both goals require analysis, the conclusions of which allow companies to gradually build qualified customer conditions that lead to specific actions. In a traditional direct marketing environment, these actions relate first and foremost to a contact strategy; in the context of a website, they also include personalized content presentation.

At this point, you're beginning to reap the benefits of working hard to define your site's data. The development of personalised website content or personalised information requires the integration of financial, marketing, operational and statistical information. This means knowing what customers have seen and done, not only during the current visit, but also their online and offline behavior in historical contexts. Personalization is based on categorizing, recognizing patterns, and sensing shifts in behavior. Without a comprehensive view of your customers, you're stuck in guesswork -- which is ultimately unreliable.

It is easy to confuse the amount of data with the quality of information. But just because data sources, flows, and inventories have grown, doesn't mean we're getting smarter. Access to knowledge requires a framework, which in turn provides the prerequisite environment for all data collection and subsequent analytical actions. Without such a framework, even with massive amounts of data, one would not understand their value.

Originally reprinted with permission from The Model Customer, January 30, 2001, Intelligent Enterprise magazine (www.intelligententerprise.com), author copyrighted 2001. Translated by Lian Qingsong.