Customer information: Server log file and clickstream analysis; data mining
MARK 430Week 3
During this class we will be looking at:
Technololgy tools for online market researchers Web analytics - server log file analysis and
Clickstream analysis static (historical data) realtime analysis personalization
Data mining - including “buzz” research Customer relationship management (CRM)
Technology-Enabled Approaches The Web provides marketers with huge amounts of
information about users This data is collected automatically It is unmediated
Server-side data collection Log file analysis - historical data Real-time profiling (tracking user Clickstream analysis)
Client-side data collection (cookies) Data Mining These techniques did not exist prior to the Internet.
They allow marketers to make quick and responsive changes in Web pages, promotions, and pricing.
The main challenge is analysis and interpretation
Web server log files All web servers automatically log (record)
each http request
Log file basics (from Stanford)
Most log file formats can be extended to include “cookie” information
This allows you to identify a user at the “visitor” level
What log files can record includes:
Number of requests to the server (hits) Number of page views Total unique visitors (using “cookies”) The referring web site Number of repeat visits Time spent on a page Route through the site (click path) Search terms used Most/least popular pages
Software for log file analysis (web analytics)
Market leader is Webtrends
Many other software packages available often made available by an ASP (outsourced
solution) can purchase and manage the software inhouse
How to select a web metrics package (from Webtrends)
How do you use log files effectively?
1. Identify leading indicators of business success
2. Identify the key performance metrics with which to measure them
3. Establish benchmarks to track changes over time
4. Configure software and use settings consistently
Shortcomings of log file analysis
Cannot identify individual people. The log file records the computer IP address and/or the “cookie”, not the user.
Information may be incomplete because of caching.
Assumptions made in defining “user sessions” may be incorrect.
This is why benchmarking is so important trends rather than absolute numbers
Log file analysis is a useful tool to:
identify what visitors are looking for what content they find most interesting which search and navigation tools they find most
useful whether promotions are being successful identify normal volatility in usage levels measure growth in site usage as compared to
overall web usage
Enhancing marketing tactics using web analytics - some examples
Identify point of drop-off in registration or purchasing process. Pinpoint problem and concentrate efforts on the apparent
trouble spot to improve conversion rates. Maximize cross-selling opportunities in an on-line
store Identify the top non-purchased products that customers
also looked at before completing the purchasing process. Add these products in as suggestions
Refine search engine placements by implementing keyword strategy Use referrer files to identify commonly used search terms
and the search engine or directory that sent the customer.
Improve web site structure using web analytics - some examples
Analysis of search logs to improve findability on the web site. Do people search by “category” rather than “uniquely
identifying” search terms? Redesign home page to enhance visibility of most
commonly used links and therefore promote usability. Demote least used items to “below the fold”
Analyze “click paths”, entry and exit points to trace most common routes around the site. Identify areas where navigation seems unclear or confusing Improve navigation to match demonstrated user
preferences.
Server log reports Format of reports depends on software used
In lab next week we will look at Webtrends reports
This is a demo from a competitor, showing typical reports
Clicktracks reports demo
Real-time profiling: building relationships with customers
Uses real-time Clickstream Monitoring - page by page tracking of people as they move through a website
Uses server log files, plus additional data from cookies, plus sometimes information supplied by user
Real time profiling entails monitoring the moves of a visitor on a website starting immediately after he/she entered it.
By analyzing their “online behavior” the potential customer can be classified into a pre-defined profiles. eg. stylish bargain-hunter etc
Clickstream monitoring and personalization
How does Amazon.com do that?
This type of personalization is very complex and expensive to achieve Existing customers and order databases must be mined for
buying patterns People who bought a Nora Jones CD also bought a John
Grisham novel Called collaborative filtering
Real-time monitoring of customers on your site needed, so you can make recommendations or special offers at the right time
Becomes even more complex when combined with information actually provided by the customer
Data Analysis and Distribution Data collected from all customer touch points are:
Stored in the data warehouse, Available for analysis and distribution to marketing
decision makers.
Analysis for marketing decision making:
Data mining Customer profiling RFM analysis (recency, frequency, monetary
Data mining Data mining = extraction of hidden predictive
information in large databases through statistical analysis.
Marketers are looking for patterns in the data such as: Do more people buy in particular months Are there any purchases that tend to be made
after a particular life event
Refine marketing mix strategies, Identify new product opportunities, Predict consumer behavior.
Real-Space Approaches
Real-space primary data collection occurs at offline points of purchase with: Smart card and credit card readers, interactive point
of sale machines (iPOS), and bar code scanners are mechanisms for collecting real-space consumer data.
Offline data, when combined with online data, paint a complete picture of consumer behavior for individual retail firms.
Customer profiling Customer profiling = uses data warehouse information to help
marketers understand the characteristics and behavior of specific target groups.
Understand who buys particular products,
How customers react to promotional offers and pricing changes,
Select target groups for promotional appeals,
Find and keep customers with a higher lifetime value to the firm,
Understand the important characteristics of heavy product users,
Direct cross-selling activities to appropriate customers;
Reduce direct mailing costs by targeting high-response customers.
RFM analysis
RFM analysis (recency, frequency, monetary) = scans the database for three criteria.
When did the customer last purchase (recency)? How often has the customer purchased products
(frequency)? How much has the customer spent on product
purchases (monetary value)?
=> Allows firms to target offers to the customers who are most responsive, saving promotional costs and increasing sales.
Data mining - including “internet buzz” research
“deploying technology that mines data for insights—nuggets of consumer opinion and real-time trends to aid and sharpen market research, advertising campaigns, product development, product testing, launch timetables, promotional outreach, target marketing and more”. (Intelliseek Marketing)
Intelliseek and firms like it use a variety of tools for data mining
A typical site that might be scanned for marketing intelligence is Planet Feedback
Customer relationship management (CRM)
Traditionally marketers have focused on acquiring new customers
CRM reflects a change in focus toward building one-to-one relationships with existing customers to increase retention Significant benefits in terms of cost effectiveness and
efficiency - it costs 5 times more to acquire a new customer than to retain one
Move toward a customer-centric focus However, just implementing CRM software cannot change
the nature of an organization to be customer facing Selling CRM software is big business - one Canadian
example is OnPath