WO2001027760A2 - Method for the analysis of user behavior in a computer network in order to optimize web presence - Google Patents

Abstract

The invention relates to a method for the analyis of user behavior in computer networks in order to optimize web presence. Said method can in particular be used to obtain qualified and future-orientated data on user behavior on the internet and intranet in order to influence one's own internet presence as a means for planning production and turnover. Said method provides a feedback circuit comprising an actual status analysis, designation of target functions, constant observation of user behavior, actual/desired comparison, performance of a trend calculation in addition to content and structure optimization of said web presence.

Description

Process for analyzing user behavior in computer networks to optimize the web presence

The invention relates to a method for analyzing user behavior in computer networks to optimize the web presence, which can be used in particular to obtain qualified and future-oriented data about user behavior on the Internet and intranet in order to be able to influence one's own Internet Presence as a means of production and sales planning.

Since the introduction of the Internet (world wide web or www), the number of online users worldwide has increased enormously. This development has made the Internet an important platform for business on the international level. With increasing commercialization, the network is becoming more and more multimedia; Graphics, videos, music and speech complete the pure text information. The communication culture, which was originally geared solely to the exchange of information, now enables the dissemination of advertising messages. Other media such as mobile telephones or TV sets are directly integrated into the communication process provided by the Internet. The Internet has thus matured into a fully-fledged communication tool in the business-to-business (B2B) and also in the business-to-customer (B2C) market. This leads to a drastic change in consumer habits and behavior. In order to meet the changed requirements, it is It is imperative on the company side to adapt the offer to the changed need structures of the consumers. Innovative companies in the consumer goods and services sector recognize the enormous market potential and invest in these integrative communication platforms. In order to proactively help shape the future of this medium, detailed information about the permanently changing user profiles is required. The chances of creating exact user profiles using a controlling instrument are greater than ever, but they also harbor the risk of misinterpretation of the data received.

In all known related procedures, the user behavior is only quantitatively logged, based on the so-called log file, a continuous log book in which all requests to the server are logged. Also, and this is usually neglected, the systems show an accuracy that has been proven to be wrong. The method of the proxy cash server used on the Internet, which is used to increase the overall performance of the Internet, does not lead to an entry in the log file, for example, if the page is loaded via the cash server. In this case, actual visitors to the corresponding page will not be registered. Conversely, a visit to "ROBOTS", that is, spies of the search engines, leads to entries that are logged as visitors to the website in the log file, but are not in the sense of the analysis. As soon as these situations are not taken into account, the interpretation is subject to errors. The explanation for this condition is quite simple. The log file is a format for the temporal documentation of the web server. It is not for the analysis of user behavior directly conceived. For this reason, the result of the log file is not wrong, but is rather interpreted incorrectly.

The invention is therefore based on the object of developing a method with which the deficiencies of the known methods are avoided, which in particular provides data corrected for errors, a separate representation of data which have an error tolerance, including the representation of the Tolerance range, a distinction between new and old customers as well as clustering in customer segments allows, - supplies future-oriented data and planning data, which can be determined by analysis intervals of your choice, and also provides the user with an advanced technical performance such as Language, screen resolution, color depth, etc. and through the historical documentation of the website gives the user a control instrument for the communication process, which enables him to quickly analyze the user behavior on the Internet and to use it to successfully carry out planning measures.

This object is achieved by the features in the characterizing part of claim 1 in cooperation with the features in the preamble. Appropriate embodiments of the invention are contained in the subclaims. A particular advantage of the invention is that the method analyzes the user behavior in computer networks and, based on this, optimizes one's own web presence by optimizing the source code of the domain after implementation of the method, recording and logging user activities from at least one server computer, the recorded activities are corrected from interpretation errors, the data available after the cleansing are analyzed and compared with data from an experience database and the results of the comparison are used for optimization by actively influencing the server computer by means of a feedback and thereby by closing one Control loop is realized.

A method for optimizing the web presence in computer networks is provided, by means of which the user behavior is recorded and analyzed, faulty influencing factors are eliminated and the web presence is optimized with the aim of increasing the economic effectiveness of the web presence in such a way that a rule value for the realization of a closed control loop is formed from the recorded, cleaned user activities and statements from an experience database, which influences the design of the web presence.

The invention is to be explained in more detail below on the basis of exemplary embodiments shown at least in part in the figures. Show it:

Fig. 1 sequence diagram m of the method;

2 shows the components involved in the method; 3 shows a graphical representation of the analysis data.

The method according to the invention allows an operator of a server for Internet domains (home pages) to automatically analyze all visits, activities and environmental parameters of the visitors (users) of these domains, to compare and evaluate this information with other data stored in databases and thereby automatically optimizing the internet presence of its domains.

This is how it works:

All requests from users to the server on which the homepage is located are logged. These requests and the subsequent server responses are recorded in the so-called log file. This recording is designed for the technical structure of the server, since the server activity is logged. The log file is used by various conventional "analysis" programs as the basis for measuring the success of the Internet. The incorrectness of these results has already been discussed. In the case of the pure log file analysis, only historical data are recorded and displayed.

In the following sections, the relevant processes in the communication between visitor and server, problems that arise and their solution by the process are described.

HIT problem: In the early days of controlling, the lines in the log file were simply counted. Every request from a visitor triggers a hit. However, a hit count is unsuitable at the present time. A screen page that has been created in HTML format can contain multimedia elements such as images, sound files or video in addition to the text body. If this page is requested, the associated multimedia elements are loaded in parallel, which are also logged in the log file and trigger a hit. A page with many multimedia elements generates a higher number of hits and therefore more hits than a page that is designed purely on a text basis. In the method according to the invention, only a subordinate status is assigned to the heat count, since this value only allows a clue about the traffic. In the process, the hits of a domain are therefore still counted. However, no great importance is attached to them. In contrast to the counts based on the usual analysis methods, the hits will be more precise. The more precise statements are made by temporarily switching off the proxy cache server and then calculating the unregistered accesses.

Logfile problem:

Due to the falsifications caused by incorrectly evaluated hits, uncontrolled recording of the hits cannot be considered. So things have to be captured that are independent of the graphic opulence of the website. In this case, it makes sense to count the requirements of individual "frame files" (HTML documents). The request for an HTML page is counted in PageViews. In contrast to PageViews, only the page impressions introduced by the method are Pages counted as content leading.

The procedure allows the exact determination of PageViews. The "PageView" unit is of course not presentative for an overall analysis, but today provides the basis for calculating the 1000 contact price for advertising space on the Internet. This unit gives the advertiser the impression that his advertising message was displayed more often than was actually the case.

The "PageView" unit is of course also counted by the process. Two problems meet here: on the one hand, too many accesses are counted, which the method corrects in a later step, on the other hand, the proxy cache server technology described in more detail below has the opposite effect: page requests are served from cache memories and lead to no entry in the log file. A complex procedure ensures, however, that all requested HTML documents are counted. A CGI image of just a few BIT forces access to the domain server. The actual pages can be loaded from the cache server. This method maintains good performance, but it triggers a hit on the domain server. A subsequent calculation determines the necessary number of PageViews for this CGI pulse. In a further step, of course, all search engine access is removed from the analysis of the PageViews, because these are not visitors in the sense of the analysis. As a result, the operator of a domain receives a 100% statement about the number of PageViews. As already mentioned, this value is not representative but is still used when determining advertising medium performance. The method will therefore support all key figures and also display them more precisely, even if these values are of little or no significance. Frame pro lejπatiJt:

Today, PageView and Pagelmpression are used as the basis for calculating the 1,000 contact price, but these analyzes are still not suitable for efficient controlling. With the introduction of "frame technology", it was possible to build a screen page from several HTML pages. However, when this screen page is loaded, each frame is logged in the log file with an associated HTML file and the linked pictures.

Depending on the frame structure and graphic opulence of a website, the quotient differs between the number of accesses (hit) and the screen pages viewed (content view). For this reason, the measurement of success using this method is inadequate.

Since programming in frame technology has become more and more common over the years, the localization of the content-leading pages for evaluation purposes is the first imperative and one of the decisive goals of the process.

Of course, each created domain corresponds to certain expectations. This is also the reason for the varied and different forms of presentation of websites. A page layout of 2 frames or a page layout of 7 frames are nothing unusual and can be useful. The procedure takes into account the individual scope for design of domain operators. The method enables those pages to be found which contain definitions of the page structure. Exactly these pages contain neither navigation, advertising messages nor content pages. So you are solely responsible for the technical functioning of the site and uninteresting for the assessment of the appearance. The- Finding the "technical" HTML documents reduces the total number of PageViews by approx. 30%. The remaining HTML documents are stored in a table according to their purpose and provided with attributes, such as content-leading, navigation or advertising-leading This assignment table makes it possible to correctly assign and count all HTML requirements from the log file, in combination with the CGI image method described above, which ensures the exact number of accesses to the individual HTML documents This means that for the first time it is possible to determine ContentViews 100%.

CGI problem: The dynamically generated pages are another challenge. Here, the page is generated on the server side from databases according to the preselection of the user and sent back to the user. For this reason, no HTML page is entered in the log file. In the event that dynamically generated pages are built or the frame technology is used, all the units and methods explained so far are unsuitable for controlling. On the other hand, the pages generated in this way are particularly interesting for product and financial planning, since they usually contain a detailed statement about the wishes and interests of the visitor or customer through the pre-selection of the visitor. More and more HTML documents, especially in the e-commerce sector, generate your content dynamically. This means that information from a database is generated and displayed in an HTML frame. For this reason, it is not possible to draw a direct conclusion from the log file between the called HTML document and the dynamic one generated information to close. However, in addition to the log file, the method enables another self-generated file to be created in parallel, where the request parameters are logged from the HTML document. In this way it is possible, in addition to the HTML document frame, to also log the information content and to relate it. The Contentview unit, which also corresponds to the facts, is correctly recorded and displayed quantitatively, ie numerically and qualitatively, ie content-related, about the information content of dynamically generated HTML documents.

Proxy - cache server - problem:

Proxy cache servers serve to make better use of the existing bandwidth of the Internet. This is achieved through the use of cache servers by temporarily storing these Internet pages. If a user connected to the proxy server requests a website, the server checks whether the page is already available locally. Only if the page is not available, is the server of the page provider accessed and an entry made in the log file. In most cases, however, this entry contains the IP address of the proxy cache server and not that of the user who requested the page. In the other case, the user receives the page from the cache server, which means that no entry is made in the log file of the page provider. For precise evaluation, it is necessary that a page that the user requests is recorded on the server side and is not sent to the user by the proxy cache without being noticed and unregistered.

The overall performance improved by the proxy cache server method also has some disadvantages. Frequently updated content pages must not be in one Cache exist, otherwise it cannot be ensured that the data contained therein is actually up to date. Normally the command "NOCACHE", ie "caching not allowed" is entered in the programming. However, there are providers who deliberately suppress this command, otherwise the service limit of the provider would be exceeded. The insertion of the "NOCACHE" command does not guarantee 100% that the information is retrieved directly from the domain server and thus logged in the log file.

The method counters this problem by introducing forced access to the domain server for content or advertising pages. A command set generated by CGI of only 50 kb forces the transfer of a 1-pixel image from the domain server each time a page is accessed by the browser, even if this page already exists in the local cache of the user or that of the proxy server. If desired, the page itself is still subject to the advantages of the cache mechanism. The network load is only increased by that 50 kb per counted page access. This value is irrelevant in today's network bandwidth and does not have a negative impact on the overall performance of the site.

local cache problems:

A comparable effect is triggered by the local cache memory that is present on the user's computer. All pages that have already been transferred once are saved in the cache directory on the user's computer. This leads to an accelerated construction of the page and is supported by all common browsers today. This process also leads to a reduced counting of visits to the website provider's server.

Example (timing): A user requests page 1 and decides to go to another page when reading a cross-reference. After a short time, he switches back to page 1. This page is now restored from the local cache directory. The measurement of the length of stay in the log file is now incorrectly assigned to page 2 on the premise that a page that is not yet in the cache is requested by the server. It should be noted here that the local cache problem with the proxy cache server problem can exacerbate the misinterpretation. For precise evaluation it is necessary that a page requested by the user is recorded on the server side and is not sent to the user from the local browser cache without being noticed and unregistered. The browsers used today allow the user to individually set the management of the cache memory. Since there are countless variations of settings for this reason, the procedure described in the discussion of the proxy cache server problem also applies to the local cache problem. A command set of only 50 kb generated by CGI forces the transfer of a 1-pixel image from the domain server each time a page is accessed by the browser, even if this page already exists in the local cache of the users. In addition to the proxy cache problem, the method also solves the local cache problem.

dynamic IP problems. The assignment of dynamic IP addresses represents another challenge in determining visitors. Many Internet service providers assign so-called dynamic IP addresses. This IP address is mandatory in the Internet world. The Internet works on the client-server principle. The client makes a request to the server. The server processes this request and sends the result back to the IP address of the request. For clear assignment of the message, the IP address is mandatory from both sides. The IP address is described in a standardized format. (www.musterfirma.de or 195.128.102.64). There are now more internet users worldwide than IP addresses. For this reason, several Internet users have to share a few IP addresses. Large service providers in particular must use this method. Two groups are formed here in dynamic IP allocation.

Type 1: When dialing into the Internet, the user receives a dynamic IP address which he keeps for the entire online session. Type 2: The user receives a new IP address each time a screen element (picture, text, etc.) is requested. In the event of heavy data traffic at the service provider, this IP is released again if the element has been answered successfully.

If no special additional procedures are used, it is impossible to determine the number of visits. At the last internet congress it was decided to expand the format by 1 * 3 or 2 * 3 digits. It will probably take about 2 years until this is put into practice worldwide Problem is exacerbated by the strong increase in users.

Unfortunately, correcting misinterpretations of dynamic IP addresses can only be achieved to a limited extent. One way to compensate for this problem is to identify a user regardless of his IP. Once this user has been identified, he can be recognized as a single visitor, even if he appears with different IPs within one session. If the user does not want to identify himself, there is still the possibility of determining the probability that the different IPs will be assigned to a visitor using an assignment logic. The number of visitors is then shown in bandwidths according to the laws of probability of statements. In addition to the actual number of visitors, there will be another graphic with the maximum number of visitors, the lowest number of visitors and the calculated number of visitors. Unfortunately, it must be said that the problem cannot be solved 100%. However, the advantages of the method are based on the knowledge of differentiating between safe and unsafe data, filtering it and presenting this data in a bandwidth. This procedure can even solve this problem up to 98% of the time. The results of the procedure are therefore so precise that planning data conforming to standards can be calculated. A deviation of 2% from the real value does not play a decisive role for the strategic development. The solution is based on two fundamentally different user behavior patterns.

Case 1: The user identifies himself (optionally also under a nickname) Case 2: The user definitely wants to remain undetected

Let us now consider case 1: In this case, the user has activated the cookies in his browser. In this case, a variable is entered in the cookie. If the visitor now visits the homepage, he can be identified during the session by means of his cookie entry. If he now visits the domain again at a later time, this visitor is assigned a new visit, even if he appears under a new IP address. With this method (under the premise that 100% of the users have activated cookies) a clear statement, which could look like this, can be made.

“In September 1999 you had 650 visits to your homepage. These 650 visits were triggered by 120 different visitors. You won 40 new visitors last month. ... etc. "

Experience shows, however, that only 50% of users have activated cookies. As a consequence, this means that an unknown mass of 50% remains, which cannot be identified beyond doubt. One way of assigning this mass to individual specific visitors despite different IPs is to keep track of which pages are called up one after the other, the so-called clickstream. Additional information that is not logged via the normal log file is helpful for this. Another command for assigning different IPs to a visitor is "VIA". Via contains the entire chain of proxy forwardings. Based on the forwarding, ie the Way, and the logical connection in the clickstream of the log file, an extremely high accuracy of statements can be made after the probability calculation. The overall security of this sub-process amounts to more than 98% and can therefore be classified as very safe.

Router problem:

Many companies use firewall computers for security reasons. A firewall computer is used for security in the company. A "bottleneck" is created through which all information must flow inwards and outwards. This bottleneck is much easier to monitor than many open interfaces to the outside. This firewall sets all the IP used in the company to a single external IP In the event that this company does not use a proxy cache server, only this firewall IP will appear in the log file, which means that different users in the company will appear under the same IP address and make a differentiation without additional information Procedure impossible.

Example: 3 visitors to a company simultaneously visit an Internet domain. Through the

Assigning an IP to all three users gives the impression that a person jumps quickly through many chapters and subchapters.

However, if the company uses a proxy cache system, which is recommended, the situation is similar to that of the large Internet service providers. This is a challenge similar to solving dynamic IP. A 100% statement can only be made if the user can be identified. If this is not the case, then statistical models can only be used to determine how high the probability is that a certain user will make the requests to the server. Investigations and tests that have been carried out show that around 50% of all users can be identified, so that a high degree of accuracy can be given about the number of remaining visitors. With these approaches, up to 98% accurate information about the number of visitors can be made. The router or firewall problem can be solved using the same method as described in the case of dynamic IPs (clickstream, tracking of proxy forwardings). However, it should be noted that the accuracy of statements when this problem arises is somewhat lower. It is around 92%. However, this value only flows into the overall evaluation to a small extent and depending on the industry and should be taken into account individually depending on the information requirement. Another option for identification can be achieved by "session tracking", provided that PHP is installed on the server. All document types are then renamed from * .htm or * .html to * .php (usually index. php) a unique variable is assigned to the user, this variable remains constant over the entire session and is appended to all called documents of the session.This attachment is visible in the standard log file and can be selected for evaluation is definitely the "royal road". In In combination with activation of the cookies or unlocked Pentium III ID, almost 100% identification of user behavior is possible. It can also be determined with no doubt how often a user has visited the page. (Differentiation between new and old customers)

Robot problem:

Search engines visit the website at regular intervals to look for new features and updates. These visits are done fully automatically by so-called search robots. Search robots are the small software spies that search the net and write all innovations in the database system of the search engine operator. Of course, these visits also trigger access to the server log, which could incorrectly be considered a visit. These visits are definitely the easiest to filter out, but should not go unnoticed. The aim must be to filter the visits of search engines on a domain from the logile so that they do not falsify the evaluation data. Furthermore, these visits are to be shown in a separate evaluation in order to obtain information about the search engines used and the associated search words.

Almost all common search engines access the "ROBOT.TXT" file. A user will never access this file in the directory structure. For this reason, the visit to search engines can be filtered in the log file and deducted from the user data. The visit to search engines can now be displayed in a separate evaluation, which can then be used to initiate further development stages for an optimized meta tag. Search engine and link problems with frame-oriented pages:

This problem is triggered by a lack of knowledge on the part of the manufacturers of the internet. If a search term is entered into a search engine and the search was successful, the results of the search are displayed to the requester in the result list. If the user clicks on the specified URL, then usually only the content page appears without the associated frame structure.

Unfortunately, you can find this error on most of the pages that are frame-oriented. This error is exacerbated by the comfort and performance of modern search engines. If a URL is registered with a search engine, the search robots will visit this URL within the next few days. With various parameters with METATAGS or parameters in ROBOT.TXT, it is permitted or prevented that certain subdirectories are searched. If searching through subdirectories is permitted, which is also sensible, care must be taken to ensure that the complete frame structure is built up in the event of user access via the search engine in order to enable navigation. This is the only way to present the complete internet domain to the user. This programming is mostly forgotten. From relevant studies it is known that 40% of all search engine results lead to a sub-branch of the homepage. This large number of "lateral entrants" must be operated properly. A programming-technical revision, especially of the homepages created by laymen, is unavoidable, since this Pages are also displayed x times in the search engine results list.

The goal must be that when a content page is called up directly, the entire frame structure is built up and the user is looped through to this page.

Experience shows that with a not inconsiderable number of domains, there are serious errors in the source code, or meta tags have been generated incorrectly or not at all. Another potentially present error, especially with frame-oriented Internet structures, is the incorrect structure of the page if it is called up in a subordinate content and not from the homepage. In this case, the navigation is no longer displayed. These problems are dealt with in the process using suitable methods that automatically and semi-automatically optimize and correct the code in a suitable manner.

The measure described above is of great importance for the overall process. Firstly, quality assurance is satisfied and secondly, the data of the meta tag and thus the information for the search engine are available in a suitable form in a database. It is now possible for the first time to carry out a target / actual comparison between the most searched keywords, the search engine brands and the information actually available on a website. The results of this comparative study enable the targeted optimization of the meta tags for customer-oriented search results and thus for the success of a domain.

Interprefcatioπs-Pro.blematiit:

The only way to get data about user behavior (unless the user identifies himself) is the log file. However, the log file is on for logging server activity and not necessarily for further interpretations. Some manufacturers of interpretation software read the data and pass it on unfiltered and not adapted to various analysis tools. Missing information about user behavior and technical performance is simply accepted or adopted. The problems mentioned above are largely ignored. The consequence is pseudo-precise basic data that is neither acceptable for controlling nor for planning. On the contrary: incorrect data are even listed as the assessment basis, which cause planning in the wrong direction and thus cause damage.

The aim is that the site provider receives only the data that is valid and meaningful for planning and assessment purposes. The data, which are based solely on presumption, are to be eliminated in the analysis so that they cannot falsify the evaluation.

Example: Some analysis programs state that 99% of users use cookies. This statement is based on the percentage of browsers capable of cookies. In reality, about 50% to 60% of the users have switched off "accept cookies", so that the value "99% of the users have cookies" is wrong. Such numbers can lead to dramatic wrong decisions in the e-commerce sector. A company that builds its shop system cookie-based cuts its customer base by 50% -60% due to the false statements. Should this child who has already fallen into the well be late sales are recognized, which would not have arisen with a standard-compliant tool. It is essential to re-program the shop system under Java, CGI, etc.

Uncertain data or empirically determined data are to be checked for their accuracy and probability and to be presented in the resulting tolerance ranges.

The challenges presented here are compensated for by the individual modules of the process. The inventors of the method did not presume to obtain all data with 100% certainty. In this method, however, a distinction is made between statements that are 100% safe and checked and the data that are inaccurate. These uncertain data are checked for their meaningfulness using recognized statistical methods and displayed in a tolerance field.

The user can choose between three variants of the display.

1. Representation of the data that is safe and separate from it 2. Representation of the data that lies within a tolerance field, and 3. the mixed and weighted representation of both data with the tolerance field also included. In this way, it can be ensured that no data is interpreted incorrectly and incorrect decisions can be largely excluded.

Forma t problem em ti k: More than 30 different log file formats are currently in use. Depending on the operating system, server architecture and provider, there are not only 30 different basic formats, but also a wide variety of combinations. Another problem is the management of this data. Many providers delete historical data or compress it and store it in appropriate subdirectories. In this case, of course, the knowledge of the internet administrator about handling the log files is of great importance. There can no longer be any question of independent management decision-making. This large number of basic formats of the log file, different combinations and inconsistent management of the historical data on the provider side have so far made it impossible to measure the performance of Internet activities in a uniform, complete and comparable manner. The goal of compensating for the format problem can only be achieved if the data is logged in a separate, independent database system. In particular, it should be borne in mind that traditionally many providers overwrite the historical data at the end of the month with the new month. Long-term planning and freely configurable analysis periods are therefore only possible in the rarest of cases.

The process completely supports the different formats. At the start of the procedure, a log file is downloaded and analyzed, among other things. The information content is determined from this analysis and compared with the controller's requirement profile. Special additional information such as screen resolutions, "Cookies active / not active" and the determination of new / old customers and the elimination of proxy cache Problems are saved in an extension of the log file. Probably the biggest challenge, the elimination of the different handling of log files on the provider side, is solved by the automatic downloading of the log file. The adjusted results are then saved locally in a specially created database system on the computer of the controller. All other displays are generated from this data management system. With this method it is possible to select different time intervals for the analysis or to compare different time intervals.

This procedure therefore offers several decisive advantages: - The database is independent of the log file;

The database can be logically linked to other databases (target / actual comparison); Individual additional information (e.g. updates) can be saved; - Independence from the handling of the provider is guaranteed;

Interfaces to other applications (e.g. merchandise management systems) are provided, arbitrary determination of the analysis periods is permitted;

It represents a basis for extensions and flexibility in subsequent expansion stages (mail flow, comm flow, etc.).

Presentation problems:

All log data is saved chronologically in the log file. In order to make this data "readable", it has to be filtered, sorted and interpreted. This effort is rare or only with high costs operate. It is not easy to draw a conclusion from the technical directory structure that can be seen in the log file and the presentation structure. All files, including the data intended for navigation, are also displayed unstructured. Another challenge is to find a relationship between the file name and the content presented. The result is: Only the group of people who developed the website is able to interpret the log file and the evaluation. However, a great deal of effort is required to interpret the data, which is rarely carried out. The sales and marketing department has no way of making strategic decisions about the future of the domain, regardless of the website's development department. The presentation must therefore be prepared in such a way that a group of people without excellent knowledge of the Internet can understand the results. The basis is the clear structuring of the content, which is independent of the technical structuring. Likewise, the form of representation "text" as a carrier of information should not be underestimated and should work in addition to the graphic representation. The procedure therefore provides freely configurable measurement periods and an individual reduction of the data to the information needs of the interested party.

A prerequisite for an easily understandable presentation is the exact representation of the presentation structure of a website. This means that the structure must correspond to what is shown to the user on the monitor. This In most cases, the structure corresponds to the technical structure on the server. For this reason, it is imperative to create a "translation" between the technical and the representation structure. This "translation" also brings the names of the files into a language that laypersons can understand. As already mentioned, the download of the complete internet structure is available as an offline version at the beginning of the process. The process now filters out all images, since these are irrelevant to user behavior. All HTML pages that form the framework for the content, ie call a frameset, are also irrelevant. The files now remaining only meet the following criteria: navigation, content, advertising space, executive page, download.

The controller assigns the files to the navigation, content and advertising space items by displaying them one after the other. After this procedure, the content-bearing pages are fixed and are structured in chapters and sub-chapters. This information is also stored in the database system and serves as the basis for the evaluation and presentation of the analysis. A further query systematics analyzes the individual information needs of the controller. This data is also stored in the database system. With this method, a strong degree of visualization is achieved, which serves the required requirements and captures the essentials.

The procedure includes the classic log file analysis to document the chronological processes on the server side, such as the documentation of the hits and the errors of the individual HTML pages in a time context Representation of various page access statistics in the form of bar and pie charts and the determination of the browser type used by the visitor. The classic analysis is expanded by a web analysis system for meaningful documentation of the actual user behavior on the Internet. This web analysis system includes the program optimization of the Internet presence, the presentation of corrected information about user behavior, such as. For example: content views, downloads, page views; the advertising medium performance according to the IVW method, the representation of the exact temporal user behavior, information about the user such as the operating system used, browser, color depth, resolution of the monitor, actual cookie activation, JAVA capability, VB capability; Representation of the traceability of the user: used search engines and search words, origin of the user via links, direct dial-in or dial-in via subdomain, analysis of the typical user paths: entry and exit page, dwell times, old / new customers; the analysis of the dwell times as well as an alarm system for the analysis of low-traffic pages. Another advantage of the method compared to conventional methods is the fact that the method according to the invention provides a tool for analyzing any period of time, taking into account the relative and absolute success. By accessing an external database of experience, in which the user behavior is documented, meaningful trend analyzes of future developments and an evaluation of user acceptance are possible.

Another part of the process is the tool for optimizing existing websites, which automatically creates a Code optimization and frame correction as well as keyword and meta tag optimization. There is also a tool for updating the search engine entries and the entry itself. Another component is made available in a training tool. This enables web-based training, analysis for learning success measurement and evaluation of the efficiency of the training program. A training program for effective web controlling taking legal aspects and technical requirements into account conveys the basics and standard methods and offers a strategically structured controlling system to optimize the interactive communication chain. The process is supplemented by an extension to include the typical functionalities of e-commerce (PAP programming, CGI programming)

The functionality of the method just described is supplemented by the possibility of establishing target control functions and alarm systems based on the data analysis, which enable efficient controlling, because these data are for the strategic development and optimization of the Internet domain under the aspect of customer orientation most relevant. This creates a link between analysis and optimization in order to implement a continuous improvement process (CIP). Additional functionalities provided by the process consist in verifying the effectiveness of marketing measures through a special analysis of the website, and verifying the company's response activities to inquiries about the in- ternet domain and an assessment of communicative activities also outside the web.

The process as a whole thus represents a control loop that can be described as follows using the example of an Internet presence with a homepage. Starting from the presentation of the website, the user activities are then recorded and analyzed, which are corrected for errors in recording and interpretation. The subsequent comparison with information provided by an experience database is then used to optimize and adapt the website with the aim of achieving the specifications of the target control functions.

The process can be described in detail as follows:

All requests from users to a server are logged. These requests and the subsequent server responses are recorded in the so-called log file. This recording is designed for the technical structure of the server, since the server activity is logged. The log file is used by various programs as the basis for measuring the success of the Internet. The incorrectness of these results has already been discussed in detail. In the pure log file analysis, only historical data is recorded and displayed. Future-oriented information cannot be calculated because the input information is already subject to errors.

The following information is shown in the classic log file: Hits: display of all hits incl. Loading images and building the frame-oriented pages;

PageViews: Display of all HTML pages - the loading of images is not counted. This value is used today as a parameter for calculating advertising medium output;

Errors: Errors when pages are requested that are not available on the server or cannot be found;

Bytes Transferred: Amount of data transferred from the server;

Daily behavior: temporal representation of information requests; Indication of the time when information was requested

Weekly behavior: temporal representation of information requests in the daily overview;

Monthly behavior: temporal representation of information requests in the weekly overview; General temporal behavior: length of stay on a page, with the background of the intensity check and checking the communication strength of the page;

Operating system: numerical recording of which operating system the user works with;

Browser: numerical recording of which browser the user works with;

Search engines: Statistical survey with which search engine the domain was found; Keywords: Statistical survey with which keywords the domain was found;

Origin of the user: recording the country from which the visitor of the domain comes; User paths: recording the click sequence within the

Domain; Entry / exit side: Entry of the entry

/ Exit page users within a domain

As already stated, the above However, data for the implementation of efficient controlling is not sufficient. Based on the classic log file analysis, the method offers the following additional benefits:

Transport of the technically oriented information into the real representation form of the domain. Check whether the data were correctly recorded. Correcting technically caused misinterpretations

Empirical determination of uncertain information with the representation in tolerance values representation of the data according to the need structure of the controller

According to the latest studies, the structure of needs can be represented as follows:

What happened on the domain from a time perspective?

What about the performance of the user's computer performance?

How did the user find the domain, what was he looking for and where does he come from? - What was the intention of the user, has he visited the domain before and how long has he requested which information?

Additional administrative information for checking communication. If all emails have been answered, the targets have been reached, what influence do advertisements and advertising, for example in print media, have on the access of the domain etc.

The method can generate further information that is of interest to the controller. Here you can see a clear extension of the performance compared to the previously known products. The R.T.F.I. Procedure using the following additional functions:

Contentviews: Number of page contacts that can be described as content-leading, i.e. actually have relevant content, and are not used for navigation. Another secondary side effect is the analysis of banner advertising and the resulting billing of advertising costs. These are currently being determined from the incorrectly recorded PageViews; Pages statistics: How is the visitor frequency distributed on the individual pages to the entire content? What does the page look like and how long was it visited on average? Downloads: Which and how often were certain files downloaded?

Information request: How often was the page visited with an information request and how often was the information actually requested afterwards? Update: When and what was changed or updated on a certain page? Color depth: What color depth has the user set on his computer? Resolution: What resolution has the user set on his computer?

Processor: What processor type does the user have?

Cookies: Has the user actually activated the cookies or does he only have a cookie-enabled browser?

Java: Which version of Java can be run on the user's browser?

VB: Is the user's performance able to perform VB?

Left: How are the links used? The measurement is bidirectional, i.e. outgoing and incoming links are measured and compared. Direct dial-in: How many users dial the domain directly and what is the percentage distribution compared to the other dial-in options?

Subdomain: How many users come to the main domain via a possibly subordinate domain? This option is often used to evaluate the success of an action.

New / old customer: Has the visitor been to the domain before, how often does he come back at what time intervals? What profile does he have and what are his information needs?

Dwell times: Which user group stays with which information for how long? How long does the visitor stay on certain pages and how long does his entire visit last? Number of visits: How many visits did the domain have and how many visitors triggered the visits?

Actions: The evaluation of actions is an essential factor of the procedure. With this evaluation, All actions that are supposed to trigger an improvement process are logged and then checked for success. Mail listings: logging of all incoming mails

Alarm system: The intervention limits are defined in this module. It is to be avoided here that pages refer and do not receive the necessary attention in the evaluation process. Target control: This module is used for target control. Here, the desired targets are compared with the achieved goals and adjusted for the normally expected rate of increase. The result is an over- or under-coverage of the targets. e-commerce: This comparative tool measures the relationship between viewing information and actual orders. This ratio is a decisive planning factor for further optimization measures in the communication mix. Picanalysis: Picanalysis shows all processes that are outside of a trend. By examining unusual rashes, conclusions can be drawn about advertising measures outside the web.

This already very detailed documentation of the historical data, which of course has been corrected from interpretation errors, is not sufficient for goal-oriented planning. This is where the comparative analysis comes into play, which differentiates between relative and absolute success. Of course, the periods to be compared can be chosen freely. Likewise, possibly Adjustments that could distort the evaluation are shown adjusted.

The process is therefore characterized by the following additional functionalities:

Comparison of any time period: When comparing any time period, all overview data are displayed in the corresponding time intervals. The resulting results show the success of various measures. The results have been cleared of the updates and possible misinterpretations associated with them. Rate of increase compensation: Every internet domain experiences a natural rate of increase. This is often due to the increasing number of old customers who visit the domain again and again or the increasing number of Internet users. This natural rate of increase, which does not necessarily reflect the increase in success, is determined and compensated for in the process. This is how the exact success is determined, which is cleared of possible misinterpretations. Analysis of the users 'information needs: The comparative analysis makes it possible to determine changes in the users' information needs. With this knowledge, the content can be adapted to the user profile and thus customer satisfaction can be positively influenced. This is particularly important for the retention of old customers. Analysis of the relative balance: At the start of an Internet domain you will find that users do not need the full range of information. The first visit lasts on average one minute. For this reason, it is important to create user profiles in order to create a balance of the entire domain. The symmetrical distribution of visits to content-generating sites should be aimed at.

Analysis of changes in computing skills:

Would you also like to know whether your clientele has activated cookies or not? Or whether you can finally require a resolution other than standard? With the comparative analysis you can see at a glance the changes in the performance of computers. This offers enough space to be able to react to the changed conditions in good time. Search engines: Submitting to search engines is very important. This is often the only way to get information. It is not enough to place advertisements with www. Address or mention it on the business card. An entry in the front places is of immense importance. To do this, the success of the search engines and the search words must be analyzed and the changes saved. Links: How many new visitors to your site result from a link? Is the advertising medium suitable for placing advertising messages? This and other questions are answered with this module. Old / new customer: The changes in the new / old customer relationship are very interesting. Is it possible to permanently bind customers to the domain through a special service? A balanced relationship between new / old customers is very important ness. It shows how much trust you can build in your customer over the Internet. Dwell times: How do dwell times change, particularly in the legacy customer segment? The answer to this question expresses relaxation and confidence in the information content. Changes in the user paths are also important for customer-oriented design of the domain. Rise rates in e-commerce: The changes between obtaining information and actually ordering services, information or products are important key figures for sales planning. The resulting results can be used directly in the corresponding planning tools.

Comparative picanal analysis: deviations in normal access behavior, e.g. triggered by ads or other actions can be followed closely here. An example is a series of advertisements that appear unchanged in the same newspaper every 14 days. After a certain period of time, the rashes on the Internet decrease. This can be seen as a clear sign of the "wear" of the advertisement. The consequence: Have the ad action paused or design a new ad.

After the analysis process has determined correct and corrected data and this leads to a comparative analysis, the logical consequence is the determination and calculation of the trends. This trend calculation is carried out according to DIN ISO. Statistical models of quality assurance are used as a basis. The- These calculations were made in accordance with the DGQ and TÜV cert.

The following questions are answered: - How are the pageviews developing?

How will time-related user behavior change in the future?

How is computer performance developing?

What influence will search engines and links have on the success of a domain in the future?

How will the relationship between new and

/ Change old customers and what effects will this have?

How will the relationship between information needs and orders in e-commerce change in the future?

Change sector?

The webmaster also needs a reference point in order to determine his own position. For example, the question arises: "Is the ratio of 36: 1 of obtaining information to an actual order to be evaluated positively or negatively? This calibration is carried out by an external experience database. The know-how of many domains is shown in compressed form in the external experience database Data has been cleaned up from identifying features of the companies. This data is made available to the users of the method online. Information about trends on the Internet, experiences in the field of e-commerce, typical user profiles and information about user behavior are stored. From these experiences result in various optimization options for the existing Internet domain. Various tools are available for optimizing the domain, which can be divided into three general categories:

Category 1 contains static modules. These modules optimize the code of the domain in two ways. On the one hand, the modules are integrated for exact analysis. On the other hand, possible errors are compensated for and corrected. A "translator" is integrated, who transfers the technical information into a descriptive form of presentation that corresponds to the content. In this section, the parameters and targets are set, questions are asked about the limits of intervention and the process is thus initialized. The static modules in Category 1 are: - Code optimization

Frame correction

Proxy Optimization

Structure converter general correction modules - parameter module

Category 2 contains dynamic modules that provide information during the operation of the Internet domain. These modules include the implemented statistical modules and a so-called error messenger. If an error occurs on the server, the latter sends an email to the webmaster stating which error has occurred with the user. In return, the user receives the message that this error will be eliminated as soon as possible. This measure is a decisive step towards customer orientation.

Category 3 contains dynamic administrative modules. The search word optimizer analyzes from the content ten words that have a high frequency and compiles a list of suggestions from them. This list of suggestions is accepted or improved by the webmaster. These words are then programmed into the Meta Optimization module. The result is a meta tag that meets all search engine requirements and guarantees the greatest possible success. The changes are saved in the search engine entry module. This module automatically contacts the selected search engines and prompts the search robots to visit the updated domain. This measure makes it possible for the first time to partially automate this task. This makes it possible to significantly improve the ranking and to permanently maintain one of the top positions. Additional modules generate suggestion lists for all actions that have to be carried out offline. These include, for example, a suggestion system for advertising.

The internal control loop is now closed: measurement, comparison, trend calculation, optimization, transfer and renewed success measurement. This process enables a continuous improvement process to be initiated step by step, which is required by DIN ISO 9000.

Another part of the process is a tutorial for effective web controlling and a configuration tool. This tool conveys all the know-how that is relevant for the successive development and marketing of content on the Internet. The following information can be called up here:

What can be analyzed?

How safe are the statements and the data? Which data protection laws must be observed? Which points need special attention? How can a continuous improvement process be achieved? How is effective web controlling best structured?

In the "Configuration Tool" module, the webmaster is asked specific questions about his need for information. This is where the results are presented. The content structure is replicated and the process is presented as the basis for effective web controlling. This tool modifies and optimizes the source code Furthermore, structural changes are carried out and certain features are added.These modifications are saved so that the original state can be restored at any time.For this reason, the central data pool is the heart of the process.

In a further module, the central data pool, a backup of the domain is created against any changes. In addition to the old structure of the data, the now modified structure is created there. The parameters and the historical changes are saved. This makes it possible to integrate further developments in the comparative analysis. Misinterpretations are almost impossible. A second log file is established in parallel to the classic log file. This log file is saved next to the external log file. All data are now to be kept independently of the provider; Automatic downloading of log files enables the seamless recording of user data. Here All analysis evaluations are saved and displayed in an understandable form in the analysis area in a way that is webmaster-friendly. This information can also be integrated into existing planning tools via the universal interface. Of course, this data can also be used in individual expansion modules.

The open platform that represents the process allows further tools to be created. These modules are manufactured individually according to customer requirements. These possible products include mail controlling, communication mix controlling or administration controlling.

The rapid progress in modern communication technologies, such as ADSL, HDSL and power line communication enable unimagined multimedia capabilities on the Internet. The content is geared more towards target group-specific aspects. Interactivity promotes dialogue with the individual customer segments. Service and service features can thus be better emphasized. The Internet enables a cost-effective, continuous communication chain to the customer and forms a broad information base without immediate buying pressure. The calculation of the success or return on investment (ROI) of a website is crucial for the strategic and content-related further development of content, customer approach and customer proximity.

The aim of the process - in the sense of a high-quality reporting system - is to increase the interactivity with the customer. The highlighting of product or service advantages and the optimization of logistics and cost management are also of great relevance.

Simple website statistics can be generated through standard analyzes. However, these generally remain unspecific, since they do not allow any identification of the user and the user paths and no analysis of the depth of the query, the pauses, the length of stay, etc. However, attractive content and communication tools can only be realized through a personalized analysis of user needs. Therefore, the target group analysis, i.e. the subdivision according to typical behavioral characteristics, is of immense importance.

The standard analysis is not sufficient to generate new, innovative dialog concepts. Only a combination of different methods enables an ideal-typical dialogue concept and ensures that all essential influencing factors of a forward-looking and qualified customer orientation are taken into account.

The benefits for the user that can be realized in this way can be specified as follows: - Flexibility and optimization of the customer-oriented communication mix Documentation of the website Presence measurement and disclosure of strengths and potential challenges of communication - Establishment of integrated communication concepts through print media, TV, radio, direct Mailing and electronic media Immediate derivation of market and demand trends and immediate adjustment of the communication mix

Reduction of media occupancy, logistics and print costs optimized target group approach.

The scope of functionality of the procedure opens up possible uses that range from private providers to local and regional providers to national / multinational providers.

The private provider provides personal information on a virtual web space. Like the private provider, the local provider often does not have its own server. His profile of requirements thus largely corresponds to that of the private provider. Both will only need a few of the modules described to meet their requirements. The regional provider provides a much more extensive range of information, since it usually has several branches in one region. It can be assumed that this customer has more sophisticated technical options to obtain verified data about user behavior. He is therefore a potential user of the method in full. The national / multinational companies mostly have their own server and are therefore able to use the entire Internet spectrum. They are potential customers of the entire process, including the training tools. Depending on the size of the company, it is also advisable to individually program interfaces to other commercially available ERP programs and planning tools. The service concept begins with the as-is analysis of the customer's website. Here, a strengths / weaknesses analysis is carried out, which shows customer improvement potential.

Since not all features of the controlling tool can be automated, some tools are adapted manually according to the customer's requirement profile.

This results in various advantages for the customer. Thus, the "outsourcing" of software specialists ensures high-quality support at all times and at the same time realizes a cost reduction in the area of the customer's human resources. Specialist know-how can be called up "on demand" or, alternatively, on-site support by a local specialist - e.g. a franchisee possible.

An optimal service concept includes the following components:

Actual analysis

Workshop for the determination of improvement potential - determination of the target state

Goal definition: short term - long term

Implementation with the method according to the invention

control

The current state of the domain is determined by an actual analysis and all usable information is taken from the log file. A potential is then determined in a workshop in cooperation with the Customers. Furthermore, the target status as well as the short and long-term goals are defined.

The implementation of the individual modules of the process, meta-optimization and structural optimization enables the first error-corrected analysis of user behavior. Based on historical data, trend calculations are carried out, which serve as the basis for further planning. Then the data is prepared in a customer-friendly manner and the resulting optimization suggestions. These results are compared and evaluated with the data from the external database and with the data from the data pool created by the method. After presenting this information, the short and long-term goals and the resulting measures are compared again in cooperation with the customer. In accordance with these goals, the content and structure optimization as well as the development of the corporate identity and the optimization of the workflow now take place. This closes the control loop, which is repeated at any intervals of 1 - 3 months.

The procedure includes the following special services:

Optimization of the communication mix through targeted advertising effectiveness analyzes - establishment of sponsor concepts through exact analysis of visitor numbers and target groups online advertising incl. Success measurement and cost / benefit analysis Support of "star alliances" through interface programming to ERP systems and establishment of logistics concepts. Optimization of the workflow from ordering to delivery and invoicing in the respective partner company

The following characteristics underline the uniqueness of the service concept for the customer: - Outsourcing, resource-saving personnel and cost structure, innovative, "state-of-the-art" solutions by permanently trained specialist personnel, rapid, measurable success through the timely establishment of concepts independent and value-free analysis of user behavior through quality-assured measures, clear presentation of the results for decision-makers - company-independent evaluation of the analysis results as a "staff function"

Investment minimization and protection through successive investment activity with parallel success measurement - networks through cooperation with market leaders and know-how carriers to solve individual challenges

The invention is not restricted to the exemplary embodiments shown here. Rather, it is possible to implement and implement further embodiment variants by combining and modifying the means and features mentioned, without leaving the scope of the invention.

Claims

claims

1. A method for analyzing user behavior in computer networks to optimize the web presence, characterized in that

- after implementing the procedure, the source code of the domain is optimized,

- user activities are recorded and logged by at least one server computer,

- the recorded activities are corrected from interpretation errors,

- the data available after the cleanup are analyzed and compared with data from an experience database and

- The results of the comparison are used for optimization by actively influencing the server computer by means of feedback and thereby realizing a closed control loop.

2. A method for analyzing user behavior in computer networks to optimize the web presence according to claim 1, characterized in that the user activities are logged in a log file and / or stored in a second database.

3. A method for analyzing user behavior in computer networks to optimize the web presence according to claim 1 or 2, characterized in that in the log file and / or the second database parameters of controlling and / or data on the structure of the HTML documents and / or the actual analysis and / or the historical data and / or the trend analysis are stored.

4. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 3, characterized in that the method provides a universal interface.

5. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 4, characterized in that the logged user activities are stored over a period of several months.

6. A method for analyzing user behavior in computer networks to optimize the web presence according to claim 1, characterized in that the experience database contains information about all HTML documents, which are provided with attributes, listed in assignment tables, and / or about parameters of controlling.

7. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 6, characterized in that the attributes assigned to the HTML documents the HTML documents as technical documents,

Mark documents for navigation, content-bearing documents or advertising documents.

8. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 7, characterized in that by the structure of the website and / or by the use of

Proxy cache servers and / or local cache memories and / or routers and / or firewalls and / or dynamically assigned IP addresses, interpretation errors caused are cleaned up.

9. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 8, characterized in that each request to a dynamically generated page including the generated content is logged.

10. A method for analyzing user behavior in computer networks nd for optimizing the web presence according to one of claims 1 to 9, characterized in that each request to a website leads to an entry in the log file by a request by means of a CGI image always is communicated to the server.

11. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 10, characterized in that the interpretation errors are corrected with the aid of assignment tables.

12. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 11, characterized in that - the automatic correction of programming errors in the HTML documents and / or the automatic adaptation of the HTML documents to the technical requirements of the users and / or the automatic adaptation of the content of the websites to the user behavior and / or the requirements of the users are part of the optimization.

13. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 12, characterized in that the optimization by

Comparison with the typical target group behavior and / or - made according to methods of customer-related quality management.

14. A method for analyzing user behavior in computer networks to optimize the web presence according to one of claims 1 to 13, characterized in that the results of the analysis are presented according to the specification of individual parameters.

15. Method for optimizing the web presence in computer networks, characterized in that the user behavior is recorded and analyzed, with erroneous influencing factors being eliminated and an optimization of the web presence with the aim of increasing the economic effectiveness of the web presence takes place in such a way that a rule value is formed from the recorded, cleaned user activities and statements from an experience database for the implementation of a closed control loop, which influence the design of the Web presence.