WO2017144982A1

WO2017144982A1 - The method of identifying users who view information and advertising websites through various devices

Info

Publication number: WO2017144982A1
Application number: PCT/IB2017/050131
Authority: WO
Inventors: Piotr RYBAK
Original assignee: Gemius Spółka Akcyjna
Priority date: 2016-02-24
Filing date: 2017-01-11
Publication date: 2017-08-31
Also published as: PL416257A1

Abstract

The subject of the invention is a method of users identification, who view information and advertising websites through devices, consisting of hardware, software and web browser. It is characterized by the following steps: assigning the users (1) with their unique identifiers (2) in site-centric environment (3), where the services are available on website (4), followed by the stage of transforming the user with identifier (2) into real user (5), as well as creating a panel (6), construction of calibration panel (7) and connecting panels in order to create a total panel (8). This is followed by assigning weights to real users (5) and publication of the number of real users (5) for the reach of the website (4) in a specified period of time and for at least one device (9), as well as for the total reach of the website generated by all devices (9).

Description

DESCRIPTION

The method of identifying users who view information and advertising

websites through various devices

This invention concerns the method of identifying users who view information and advertising websites through various devices. It's been implemented by javascript scripts, which register data from cookies and/or browser identifiers (BID).

There is a market demand, especially from advertising and media companies, for the results of websites' views measurement, as this is a crucial factor for deciding on broadcast time of commercials represented by specific industries and branches. The process aimed at estimation of the number of real users, in other words, a thorough measurement of viewing a specific website via all the devices, such as PCs, mobiles, smartphones, tablets and audio-video, is highly important.

Modern companies, providing services of internet content viewing measurement, are required to provide the results that reflect only real users behaviors on specific websites and that are given in the shortest possible time after the measurement period is finished. At the same time, users' behaviors are measured in relation to the information and advertising material on the website, in order to adjust the advertising content to the interests of specific user and to avoid automatic display of advertisements that are not adjusted to the users' needs.

The internet uses small text files called cookies to store specific fragments of information by a browser. The information can be then used for different purposes. For example, they serve as indicators of website viewings estimation.

Each cookie has the following components: cookiename=cookievalue; expires; domain. The first part of the cookie syntax is the name and the value of the cookie. These are the only cookie parameters required. If a browser does not find a cookie with a specific name in a current domain, it will create a new cookie. If a cookie with specific name already exists, its current value will be replaced with a new value. The value of the cookie is given as a string of letters, without spaces. In order to apply space in a cookie value, the cookie needs to be transformed using a special function.

The parameter 'expires' specifies expiration date. In cookie files, the expiration date is expressed in milliseconds. Once the expiration date is reached, the cookie is removed from the system. If this parameter is not provided, the cookie will be valid during the session, which lasts as long as the browser is open. Once the browser is closed, the cookies without the parameter 'expires' are deleted. If a cookie expiration date is earlier than the current date, the cookie is deleted.

The parameter 'domain' describes cookie domain. Lack of this parameter will make the browser will set its name the same as the name of the server, on which the cookie was created. In case of displaying all cookies out of a cookie document, only those with the 'domain' parameter compatible with the one from which the request for cookies has been sent will be displayed.

There are two kinds of cookies: first party cookies and third party cookies.

'First party' cookie is assigned by the server that contains a viewed website. 'Third party' cookie is assigned by a different server from the one containing the viewed website. This different ("third") server may be an ad-server. It can be responsible for displaying adverts on many websites, for example belonging to one advertising network. While showing the advertising materia to the user it also assigns him with a cookie (a 3^rd party cookie). In such a case, the user might receive two cookies - one from the server where the viewed website is located (1 ^st party cookie), and the other from the server responsible for the displaying adverts on this website (3^rd party cookie). This is how the ad-server recognizes users and tells if the user has already seen a particular advertisement. Therefore it will not keep displaying the ads, but will show it only a number of times during the campaign. The matter of the invention is the method of identifying users who view information and advertising websites through various devices that consist of a hardware, a software and a web browser, including the following stages: a) The users are given their identifiers in a site-centric environment, where services content is available on a website.

b) The stage of transforming the users with the identifiers into real users.

c) Creation of a panel

d) Creation of a calibration panel

e) Connecting both panels into total panel

f) Assigning weights to the real users

g) Publication of the number of real users per reach of the website in a specified period of time and for at least one device, as well as for the total reach of the website generated by each device.

The method of users identification is characterized by users' visits on the website at least once within 4 weeks.

Profitably, when in stage a) the identifier is a 3rd party cookie or an individual browser identifier (BID).

Profitably, when in stage a) the identifier is applied using Javascript and the problem of 3^rd party cookie deletion and using the same device by multiple users is being eliminated.

Profitably, when stage b) takes place in site-centric environment and the data being used are from 28 days before.

Profitably, when the panel constructed in stage c) consists of real users.

Profitably, when the calibration panel, constructed in stage d), and also if the calibration panel is constructed from panelists connected via e-mails, mailings or phones.

Profitably, when the complete set of information on the panelists is collected in stage d). Profitably, when in stage e) the panel connects with the calibration panel, including the ones from each device, into the total panel.

Profitably, when correction in the total panel is being conducted in stage f).

Profitably, the device in stage g) is a PC, a mobile, a smartphone, a tablet or audio- video stream material. Profitably, when the publication of the number of real users measured in a specified 24-hour period occurs directly after the measurements are performed, based on one device as well as on all devices.

The invention has been presented in figures and examples.

The 1 st party cookie, the 3rd party cookie and the BID (browser identifier) are assigned to the internet users.

As understood by the method, according to this invention, a user is an internet user visiting a specific website, identifiable and recognizable by a cookie and/or the BID assigned to the browser.

The method of installation of the BID is shown in fig 1 .

The BID may be installed on browser or applications, if a device such as a smartphone, a mobile, a tablet or an audio-video material utilized by the internet users is available. In case of applications, the identification of users is performed by application identifiers, which are being saved in an analogical environment such as the cookies or localstorage (separate from the browsers).

By user's (1 ) consent, a measuring device, a so-called 'hit', is installed. It is equipped with auditing javascript scripts that enable registration of data from cookies and/or from the browser identifier (BID). Then, the script then sends information on the hit to a server called 'hitcollector'. Each situation, once the user (1 ) connects with the website, is called a display. In other words, display happens when the website is viewed by a user (intentional display of the website). Duration is the time difference between events. Collected hits are located in the data center, where they are distinguished between the source they come from and specific devices This is performed by using computer application called 'user agent'.

The stage a) of the invention has been shown in fig 2.

Due to the fact, that the 3rd party cookies allow users to be recognized not only within one website, as it is in case of the 1 st party cookies, but on many websites, has risen justified concerns regarding privacy protection of the internet users. This issue is being solved by defining the privacy policy by the owners of the servers that assign 3^rd party cookies. The information on the policy having been defined is located in the cookie header.

The BID is an improved solution, as it contains security measures that protect and respect the users' privacy, whilst thoroughly and reliably identifying them in the Internet.

The BID eliminates the problem of non-accepting cookie files by the internet users. Installed in applications, it is characterized by the fact that the phenomenon of deletion doesn't practically exist.

The following situations commonly occur in the internet: the cookie deletion, using the same user name on a device (so one cookie file is assigned to more than one internet user), using more than one browser by one user (so more than one cookie is assigned to one user). As a result, the measurement of the internet viewing, including the assessment of the real internet users number, is impossible to be established.

The longer the analyzed period, the bigger the difference between the number of cookie files and the number of observed internet users are. The number of all cookie files observed within a month may be even tenfold higher than the number of internet users from a given country.

The presented problem is being solved in the site-centric environment with the use of system for functions and algorithms that assess the number of website visitors. As a result, according to this invention, the number of real users applies to the number of people who visited a certain website at least once during the analyzed period, regardless of the duration or the frequency of visits.

Site-centric data is received from scripted websites, players and applications. User- centric data is obtained by questionnaires filled in by online respondents as well as by the function installed on the respondents' devices, which monitors their activity. Then, the data is being combined, which allows to achieve a unified set of user-centric information. In addition, the user-centric data are weighed, considering their representation within population. Sets of site-centric and user-centric information are then combined to create full source of information on the ways of using websites' content, material streaming and the applications being used as well as on social and demographic structure and users behavior. The method of measuring the internet viewing involves many verification processes, including the ones that eliminate disruptions (such as, for example, an automatic refreshing of the websites). Also, the quality of the measurement results is being tested.

The algorithms used in order to calculate the indicators of real users and their reach has been designed to eliminate the influence of many unfavorable factors, such as sharing a computer, using many computers by one user, using many types of devices by one user.

As it is common that one user accesses internet using different types of devices, there is a need for de-duplication of data and for presenting them as a single, collective result.

The US patent description 8,996,696 ('Measurements based on panel and census data') outlines the access to the first set of usage data for the first set of resources on network and the second set of usage data for the second set of resources on network. The first group of client systems accessed the first set of resources and the first set of usage data is being determined on the basis of the information received from the first group of client systems sent as a result of beacon instructions included with the first set of resources. The second set of usage data is determined based on information received from monitoring applications installed as second group of client systems that accessed the second set of resources. Users of the second group of client systems are a sample of a larger group of users that use resources in the network. The initial usage of measurement data for the third set of resources in the network is determined on the basis of the first set of usage data. The third set included one or more common resources that are included in the first set of resources and in the second set of resources. One or more adjustment factors are determined on the basis of the second set of usage data and applied to the initial usage measurement data to generate adjusted usage measurement data. One or more reports are generated based on the adjusted usage measurement data.

On the other hand, the US patent 8,874,652 describes the access to panel and census data representing accesses by sets of users with multiple types of media platforms to media content associated with multiple media entities. An overlap in the accessed panel that represents users who have accessed media content associated with media entity with more than one of the multiple types of media platforms is determined. Based on the accessed panel data, the determined overlap in the accessed panel data, and the accessed census data, an overlap function that estimates an overlap in the accessed census data is derived. The derived overlap function is applied to census data associated with a media entity to estimate an overlap in the census data associated with the media entity. The overlap in the census data represents users who have accessed media content associated with the media entity with more than one of the multiple types of media platforms.

However, the above solutions do not apply to the identification of the real users that use various devices to view information and advertising websites, where the measurement result is given directly after the specified measurement period is finished and it concerns both the reach of the website accessed from at least one device and the total reach generated by all devices.

In the current state of technology, all site-centric and adserver systems present the number of cookies registered on a certain website as an estimator of a number of real users visiting this website. Unfortunately, this doesn't include the occurrence of cookie deletion, which means that once in a while some users, either deliberately or otherwise, delete the registered cookies from their computers or mobile devices. Because of that, the site-centric systems see these cookies more than once within the specified period. Cookie deletion is the major reason why the number the registered cookies is much larger than the real number of a measured website users. As an example, a user who deletes cookies from his computer every day will be recognized by site-centric systems and the adserver systems as thirty different users within one month. In order to identify the number of users for the measured website, some additional factors, that contribute to the fact that the number of registered cookies is not equal to the real number of users, need to be taken into consideration. These include sharing one computer, and therefore one cookie, by many users, as well as using internet by the same people on many computers.

The commonly known method of the real users' number estimation consists of the following stages:

1 . Calculation of the number of page views generated by all users at the website - this number is labeled Ow. 2. Estimation of the number of cookies for which an assumption that they existed throughout the entire researched period is being made (the cookie that is presumed to have existed throughout the whole researched period is the one that has existed both before and after the end of that period) ;this number is labeled Cd.

3. Calculation of the number of page views generated by the cookies defined in point 2 above - this number is labeled Od.

4. Calculation of the number of cookies that would be registered for a researched website if there was no cookie deletion, in accordance with the formula below:

Cw = (Ow/Od) * Cd

In the same manner, but by replacing the researched website with the set of all websites enrolled in the study, calculate the number of cookies that would be registered for all websites taking part in the site-centric research, assuming that there is no cookie deletion - labeled Cp. a) Calculation of the relative reach for the researched cookies' website in a given time period, in accordance with the following formula Zw = Cw/Cp b) If P is the size of the total population of the internet users in the research period, the number of users of the particular website in the given time period can be calculated in accordance with the following formula:

Uw = Zw * P

The above calculations are based on the following assumptions:

I) It is possible to select a group of cookies as described in point 2 above.

This means that it is possible to monitor cookies activity, not only on the website included in research, but on the highest possible number of websites, and not only within the study period but also before and after (for a study period of one month it is best to monitor the cookies activity at least 4 weeks prior to it and at least 4 weeks after). As a consequence, the publication of research results is delayed. Certain amount of time needs to elapse in order to verify whether the cookies do belong to the group described in point 2 above. The concern of this study is a technical possibility to monitor the activity of cookies.

II) The group of cookies selected in point 2 is a representative group for all the cookies, especially when it comes to the average number of generated views. The research includes verification through comparative analyses of all the possible behavioral characteristics of the selected group of cookies in relation with the remaining cookies.

III) The proportion of the number of views generated on all selected websites by the group of cookies described in point 2 to the number of all views on all selected websites is identical as the proportion of the views generated by the same group of cookies of all the websites (including the ones that are not included in research) to the number of all the views of all websites. The relevance of this study has been confirmed by the panel research results, that took into consideration the views of both monitored websites and the ones not monitored by the site-centric system, by users who visited at least one of the studied websites at least once, including both the ones who delete cookies and the ones who do not.

IV) The relative reach of the studied websites calculated within the study period based on the selected group of cookies is a good estimator of the reach amongst the users of studied websites.

V) The size of the internet users population (P) is a good estimator of all the internet users that visited all the studied websites within the study period.

Thanks to applying the method of identifying users according to this invention, the number of real users is given daily for a specific website. In the method covered by this invention, the 'real users' are defined as the number of the internet users that accessed a specific website within a specified time frame. This represents people who accessed the websites from the territory of the country of analysis, as confirmed by the IP address analysis.

The invention solves the technical problem of the necessity to wait a certain period, e.g. one month, after the study period, in order to verify which of the registered cookies belonged to the users that meet the definitions of the internet user and were not deleted within the study period. Thanks to the analysis of the activity of users according to the cookies assigned to them, particularly of the time that elapsed from the last registered activity and of the scale of the registered activity prior to the last registered activity, the invention allows to set the probability of deletion within the study period for each of the registered cookie or BID. As a result, it allows to estimate the size of the registered cookies group, that belonged to the users meeting the criteria of the internet users and being not deleted within the study period, immediately after the period, for which the number of real users is calculated, is finished.

The method of identifying users according to this invention assumes, that the number of real users of a website is not equal to the number of cookies measured on that website. The number of cookie files exceeds the number of real users for many reasons mentioned above, such as cookie deletion, using many devices and sharing web content with other users.

The method of identifying real users according to the invention uses an algorithm in order to analyze and create a panel of real users. The algorithm uses two sets of input data:

1 . Information on the movement on all websites covered by research.

2. The number of internet users in the country, as defined by the structural studies.

The method of users identification according to this invention is based on the fact, that it is possible to define a group of cookie files that represents all the cookie files, which can be applied in the calculation of a website's reach. Knowing the average number of website displays per cookie file in the selected group and the total number of the website display, the number of real users can be defined. In order to do that, the following steps need to be performed:

1 . Counting the number of browsers registered for the analyzed website, provided that there was no cookie deletion.

2. Measurement of the relative reach of the website in comparison with all the examined websites.

3. Calculation of the number of real users of the analyzed website, based on its reach and the number of internet users in the country of research. The method of users identification according to this invention assumes that the movement generated by the selected group of cookies should have the same characteristic features as the movement generated by all the cookies. In order to include cookie to be included in the study group, it needs to exist throughout the whole study period, which is both before and after that period. To shorten the time of calculation of the real users by skipping additional waiting time after the study period, an advance stage was applied. It assumes the probability, that the users that were active on a specific website at least once, will still be active after the study period ends.

In the method covered by this invention, each cookie file is described using two parameters:

1 . The number of days between the last noted activity within the study period and the end of that period.

2. The indicator of activity of cookie file within four weeks preceding the last noted activity.

Based on the parameters 1 and 2, all the cookies are divided into strata. For each stratus, the probability of being a representative cookie is estimated on the basis of historical data, as a proportion of the representative cookie files in each stratum.

The period of 4 weeks - 28 days is best for calculating the level of cookie activity. Due to the fact, that the behavior of cookie users differs throughout weekdays and on weekends, the 28-day period always contains the same number of particular days, for example, there are 4 Mondays, 4 Tuesdays etc.

The stage b) according to the invention contains the following steps:

1 . Estimation of the number of cookie files registered for the analyzed websites, provided that there was no cookie deletion.

2. Measurement of the reach of the analyzed website.

3. Calculation of the number of real users.

Stage b) of the solution brought by this invention has been presented in fig 3. EXAMPLE 1 : stage b) All the data for a PC device (platform) are analyzed on the basis of the movement of the internet users (2), measured by the BID identifiers and the number of views of the information and advertising websites. Then, historical data (from the past) for the BID users, preferably from 28 days prior, is being verified regarding their credibility, i.e. considering cookie or BID deletion or using the same PC by multiple users. The selected group of users is true number of the real users (5) in a given period.

In other words, for the purpose of the example of applying stage b), we need to: a) Calculate the number of views generated in a given period by all the cookie files registered on the website. b) Estimate the number of cookies, for which it is assumed that they existed for the whole study period. In order to do so, it is required to:

I. Divide cookie files into groups, following the below criteria:

- the time that elapsed from the moment of the last noted activity in the analyzed month until the end of that month. Here, it is given in calendar days and denoted as T1 .

- the level of cookie files activity within four weeks prior to the last noted activity. Here, it is given in calendar days and denoted as T2.

II. Calculate probability of existence of the representative cookie files for each stratum, depending on T1 and T2 and the movement within 6 preceding months, as follows:

Probability (T1 , T2) = representative cookies (internet T1 , T2)

All cookies (internet T1 , T2)

III The estimated number of cookie files, for which it is assumed, that they existed throughout the entire study period equals:

c) Calculate the number of website displays generated by cookie files defined in point b of this example. To obtain this indicator, it is necessary to:

I Divide website displays into W, T1 , T2, where each group contains the displays generated by cookie files from the stratum defined by T1 and T2

II Apply the same probability as the one used in point b of this example.

The number of website displays generated by cookie files defined in point b is:

d) Calculate the number of cookies registered on the studied website, provided that there was no cookie deletion, according to the following formula:

e) Analogically, by replacing studied websites with the group of all the studied websites, we can calculate the number of cookie files registered on all the websites covered by the site-centric study, providing there was no cookie deletion

f) Calculate relative reach of the website linked to the studied cookies in a given period, according to the following formula:

The number of real users that visited the studied website in the specified period has been defined as follows:

BeolUsersiW) = Re ch(W) - InternetPopulation

The method of identifying real users per stage b) applies to the measurements concerning websites users amongst users of PCs, smartphones, tablets and audio / video streaming, divided into the PC, smartphone and tablet users, estimation of the users of foreign websites and of audio/video streaming among the users of PCs, smartphones and tablets.

For each of the above situations, the method of users identification in stage b) is identical and incudes filtering of movement coming from a specific device (9), based on a User Agent entry, followed by the identification of browsers implemented by cookie files or BID, counting the number of browsers from each website (4) using functions and algorithms, and finally identifying real users (5) for each website (4).

The method of users identification according to this invention, provides a possibility to obtain official results for the real users (5) of a specific, single website (4) directly after the study period is finished.

After identifying real users (5) for every single website (4), a panel (6) is created in stages b) and c) for each of the analyzed devices (9). The panel (6) consists of the real users (5). The next step is to build a calibration panel (7) in stage d. The calibration panel (7) is a set of pairs of identifiers of users representing different devices (9), and the condition for being a part of this panel is being the same person. As an example of performing stage d), this group can be constructed as follows:

Example 2, stage d)

Connecting panelists based on e-mail addresses.

A panelist fills in a survey and provides his e-mail address. If the panelist fills in the survey on more than one device (9), his activity via different devices (9) can be associated with help of the provided e-mail address. In order to validate, whether it is the same person, a consistency of answers for the VQ questions need to be confirmed. VQ = {q1 , ... .,qd} - the list of identifiers of the social and demographic variables, for which the answers must be consistent. As a default, it is the sex and the age group.

Panel (pit) - a set of panelists on a pit panel

Calibration Panel (pit 1 , pit 2) for pltl , plt2 - a set of users from two platforms out of all of the analyzed ones, as to which we are certain, that they represent the same person.

Formally, calibration panel for platforms pit 1 , plt2 can be defined as:

Calibration Panel (pltl , plt2) = {(panelistl £ Panel (pltl ), panelist2 e Panel (plt2)) | panelistl .mail = panelist2.mail Λ VqeV Q panelistl .q.a = panelist2.q.a} (1 ) where:

• panelist. mail - e-mail address provided by panelist,

• panelist.q.a - an answer to the question q

Example 3 - stage d

Calibration panel (7) may also be built by sending mailings to the current panelists. The mailings contain requests to the panelists for clicking on a provided link using all available devices (9). The transmitted URL address contains the panelist's identifier. Thanks to the identifiers, the activity of two or more users can be associated. In this case, there are no questionnaires to be filled, so there is no possibility to validate the social and demographic information.

Example 4 - stage d

Calibration panel (7) may be constructed by associating the panelists by the phone numbers.

Connection of the panels into a so-called total panel (8) is performed in stage e). This is presented in fig 4.

Example 5 - stage e

Constructing of a predictive model with the following parameters: - number of Models - a number of constructed models, 50 by default

- maxCoef - the maximum value of coefficient, the default value is 1

- minCoef - the minimum value of coefficient, the default value is 0.

Input data:

- panel (pit)

- calibration panel (pit 1 , plt2) Construction of the training dataset

For each pair of the panelists - user 1 , user 2 and a node, function d) can be defined as:

The panelist.node.pv is a number of views performed by a specific user on a node in study period. Less formally, the function d equals 1 , if each user visited the analyzed node, and 0 in the opposite case.

The training dataset is a matrix defined as:

These all are two-element combinations of users that belong to the calibration panel, originating from two different devices.

The decision vector yn should be defined as:

it equals 1 , when a pair of users represents the same person and 0 in the opposite case. Panelists who have not been paired are assigned to a single device, the one that they were using.

The aim of the model presented in this example is to assess the probability of a chosen pair of panelists, who use different devices, is in fact the same person. In order to do that, a training dataset and a logistic regression model have been applied. The Coordinate Descent algorithm has been used to obtain the final result.

Example 6 - stage e

The model teaching scheme is as follows: 1. The training dataset is randomly divided into two sub-sets, containing 80% and 20 % of all observations.

2. Logistic regression model with L2 regularization for the series of parameters λ defining the validity of regularization factor in relations with the sum of residues squares is learning from a larger subset.

3. For each of the constructed models (for different λ), an F-score measurement is calculated, based on this subset. The model that maximizes metrics is chosen as a target model.

4. Steps 1 -3 are then repeated and the number of repetitions equals the number of models. The coefficients for the final model are an arithmetic average of the coefficients for each of the base models

5. In order to increase the stability of model, the final coefficients of the model are weighed against the ones from the previous two months. The weights for the coefficients are: ½,⅓,1 /6. In case of coefficient being not presented in the previous months, it is not averaged, but the newest value is taken into consideration.

Example 7 - stage e)

The detailed description of teaching the model

The logistic regression model takes into consideration a number of parameters that were omitted in the previous example

• minCoef, maxCoef - the minimum and the maximum value of the coefficient for a dependent variable

• a weight of observation - equals 1 if the observation contains a pair of users that are not the same person and 1 + fracNiiCalibrationPanelii for the same person. N is the number of all the observations and iiCalibrationPanel n is the number of pairs of panelists.

• the model intercept - always equals 0.

In the glmnet packet, the teaching process of the model is as follows:

ml = glmnet(X,y,family="binomial", alpha=0, weights=y^*(1 + length(y)/sum(y)), intercept=FALSE, lower.limit=0, upper.limit=1 )

Output data:

CK - a vector of coefficients for each behavioral variable.

Example 8 - stage e) Metric segmentation

Parameters

• SQ = {q1 , . . . , qd} - the list of identifiers for social and demographic variables.

As a default, these are: sex, age group and binary variables representing the use of a particular device.

Input data

P anel(plt) - the set of panelists on the platform - pit.

Description

In the first stage, the study panels from each device are divided into independent segments of panelists of the same social and demographic profiles (the classes of abstraction are set based on the vector of answers for the defined social and demographic variables). Formally, for each device, the set of panelists is divided into:

where A- is a set of all the combination of answers to the questions included in the SQ, It is possible that there exists an A or a pit, where S(A, pit) = 0. For example, the panelists using tablets, who claim that they only use PCs.

Output data

For each of the analyzed pit device, the division of the panel into S(A, pit).

Example 9 - stage e)

Combining panelists through the following parameters:

- minSimilarity - a minimal similarity needed to associate two panelists, by default it is 0.2

- maxUsage - the parameter defining how many times a panelist can be used in the association process

- panel (pit) - a set of panelists on particular pit device

- S (A, pit) - the division of panel or each of the analyzed pit devices

- Ck - vector or coefficients for each of behavioral variables.

Based on the defined coefficients, a function is derived, which estimates the probability of a pair of panelists being in fact the same person.

For the model defined above, the probability for each pair of panelists can be calculated, and, as a result, study panels can be connected. This process is conducted as follows:

2. Repeat point 1 with Mobile as a main device and Tablet as an additional device. In case there is no BPS function for Tablet and Mobile, use the function for PC and Mobile. This point only refers to the division S (A, pit), where the panelists only use Mobile and Tablet devices.

As shown in the above example, each of the PC panelists can be also assigned with Mobile or Tablet activity and will show up in the target panel exactly once. On the other hand, the Mobile and Tablet panelists may be connected with many PC panelists. Some panelists, (PC, Mobile, Tablet) will not be associated with any other panelist. They will be classified in the final panel as unmatched.

3. For each of the newly created panelist, his demographics is defined by the demographics of the PC panelist. If he does not use a PC, then the demographic of the most current panelist is taken into consideration.

Example 10: stage f)

In the process of assigning weights, the number of real users (5) for each website (4) on each PC/Mobile/Tablet/audio - video device (9), the data concerning population structure for each device (9) and the size of population of the common parts for each set of devices are combined.

Sylwia Owczarek

Patent Agent

Claims

1 . The method of identifying users who view information and advertising websites through devices that consist of hardware, software and web browser, characterized by consisting of the following stages:

a) users (1 ) are assigned with their identifiers (2) in site centric environment (3), where services are available on website (4),

b) the stage of transforming a user with identifier into a real user (5)

c) creation of a panel (6)

d) construction of a calibration panel (7)

e) connection of the panels in order to create a total panel (8)

f) assigning weights to real users (5)

g) publication of the number of real users (5) for the reach of website (4) in a specified period of time and for at least one device (9), as well as for the total reach of website, generated by all the devices (9)

2. The method of identifying users as described in claim 1 , characterized by user's (1 ) visits on the website (4) at least once within 4 weeks.

3. The method of identifying users as described in claim 1 , characterized by identifier's (2) being in stage a) profitably a 3^rd party cookie or an individual browser identifier (BID).

4. The method of identifying users as described in claim 1 or 3, characterized by the identifier's (2) being in stage a) applied with the use of Javascript.

5. The method of identifying users as described in claim 1 , characterized by the problems of the 3^rd party cookies deletion and of using the same device (9) by multiple users are being eliminated in site-centric environment in stage a).

6. The method of identifying users as described in claim 1 or 3 or 4 or 5, characterized stage b) taking place in site-centric environment (3).

7. The method of identifying users as described in claim 1 or 6, characterized by using data from 28 days prior in stage b).

8. The method of identifying users as described in claim 1 , characterized by inclusion of real users (5) in the panel constructed in stage c).

9. The method of identifying users as described in claim 1 , characterized by inclusion of at least 150 panelists in the calibration panel (7) constructed in stage d).

10. The method of identifying users as described in claim 1 or 9, characterized by inclusion of the panelists, connected via e-mail, in the calibration panel (7) constructed in stage d).

1 1 . The method of identifying users as described in claim 1 or 9, characterized inclusion of the panelists, connected via mailings, in the calibration panel (7) constructed in stage d) .

12. The method of identifying users as described in claim 1 or 9, characterized by inclusion of the panelists, connected via phones, in calibration panel (7) constructed in stage d).

13. The method of identifying users as described in claim 1 or 9, characterized by the complete set of information on the panelists being collected in stage d).

14. The method of identifying users as described in claim 1 or 8 or 9 or 10 or 1 1 or 12 or 13 characterized by the panel (6) and the calibration panel (7) being connected in stage e) in order to create the total panel (8).

15. The method of identifying users as described in claim 1 or 8 or 9 or 10 or 1 1 or 12 or 13 or 14 characterized by the panel (6) and the calibration panel (7) from each device (9) being connected in stage e) in order to create the total panel (8).

16. The method of identifying users as described in claim 1 or 15 characterized by the corrections in the total panel being applied(8) in stage f).

17. The method of identifying users as described in claim 1 or 2 or 3 or 4 or 5 or 6 or 7 or 15 or 16 characterized by the probability of a device being in stage g) a PC.

18. The method of identifying users as described in claim 1 or 17 characterized by the probability of a device being in stage g) a mobile or a smartphone.

19. The method of identifying users as described in claim 1 or 17 or 18 characterized by the probability of a device being in stage g) a tablet.

20. The method of identifying users as described in claim 1 or 17 or 18 or 19 characterized by the probability of a device being in stage g) an audio or a video streaming material.

21 . The method of identifying users as described in claim 1 or 17 or 18 or 19 or 20 characterized by the occurrence of the publication of real users (5) number, measured in a specified period, in stage g) directly after the measurements are performed, based on one device (9) as well as on all devices (9).

22. The method of identifying users as described in claim 1 or 17 or 18 or 19 or 20 or 21 characterized by 24 hours being the specified period in stage g).

Sylwia Owczarek

Patent Agent