Media

Consumer preference model

The model allows you to determine the consumer preferences and predict their consumer behavior based on the analysis of historical data, taking into account accompanying events, the geolocation of a person.

 The group behavior of similar people is analyzed at the same time. Knowledge of patterns of consumer behavior helps to build effective marketing strategies.

Building recommendation systems is also an important direction. Such systems carry out automatic selection of content according to the interests of a particular user, content ranking, optimization of digital targeting. Based on the models, you can build forecasts of coverage, ratings, content views, site traffic.

Search for anomalies

We use a number of methods to detect anomalies, and their application depends primarily on whether we are dealing with separated or unseparated data.

 Anomalies are rare combinations of data, such as non-standard user behavior, or abnormally large purchases.

It is important to distinguish anomalies that indicate errors in the data and anomalies that are determined by hidden relationships between data.

In the first case, the search for anomalies is part of the data processing and cleaning process, in the second case it can be the overall goal of data analysis.

Analysis of media consumption

When studying media consumption, a number of parameters characterizing user behavior are used. These options include:

• ATV - minutes, the average time spent watching a media event per day among the population.
• Rch – coverage, the number of people who watched the media event for at least 1 minute.
• Rtg – rating, the average number of people who watched the media event throughout its duration, expressed as % from the total audience.
• Share – the audience share that had contact with the media event for at least 1 minute, in % from the total audience.

Panel construction methodology

In panel studies, the information obtained for a research participant is projected onto the entire population. Therefore, it is important to maintain the representativeness of the sample throughout the study period in conditions of a high level of rotation of research objects.

To maintain a stable and realistic research result, when building a panel, dynamic weights are calculated that are determined for each user based on the sample bias at each moment of the research time.

Customer clustering and user profiling

Based on purchasing activity data, media consumption data, it is possible to identify typical consumer behavior patterns and identify behavioral patterns.

The user profiling system makes it possible to formalize a list of data characterizing people's behavior and create typical user profiles.

User profiling is the basis for clustering consumers based on their preferences, time and place of purchase, and additional external data.

Analysis of geopositions of users

Determining the location of a person and its connection with consumer behavior. Such analysis can help predict user preferences. Since consumer behavior of a person depends on external factors. For example, whether he is near the house, traveling or visiting. For example, geolocation data on mobile devices can be used to analyze the user geolocation.

Data processing and visualization systems

Data processing includes operations such as aggregation, obtaining summary characteristics for a certain period, determining the characteristics of a data set depending on the requirements. This allows for dynamic tracking of metrics of interest.

We develop and implement methodologies for processing the systematization and organization of databases. We develop the logistics of information flows.

 We visualize the data in a user-friendly form:

• immersive BI dashboards,
• heat maps,
• interactive sites and applications.

We can integrate the application into the client's CRM*/ERP system or set up email notifications about significant market events.

*customer relationship management system

Employee Profile Record

R&D

We have vast experience and a growing knowledge base in building a large number of ad-hoc research and products for end users. Thanks to this, we can choose the most effective solution for any client's tasks.

Outsourcing

What products do we offer:

• development and implementation of a methodology for processing the systematization and organization of databases,
• management and systematization of Big Data,
• calibration and verification of data,
• provision of data scientists teams.

Text analytics

Semantic analysis of messages from online media and social networks, determines the most relevant topics. It is possible to conduct a study of leaders’ opinion texts tone. Moreover, this can be done for certain categories. For example, news topics, persons, geographical indication, searches, neutralization of negative reviews , etc.

The dynamics of the occurrence of entities in media content gives an idea of current trends and user interests. Extracting entities from text is also used to create automatic description and tagging of audio and video clips, images, for following inclusion in a search system or an analytical system.

Product analytics

Product analytics defines how users interact with a product. It helps to identify the most popular product features, retain customers, find product growth points.

We have over 15 years of experience in generating almost all types of analytical reports:

• Excel uploads and Ad-hoc research at the request of the client,
• dynamic analytical dashboards in BI and web interfaces that provide real-time analytics,
• building user profiling models based on preferences and consumer behavior