Business Intelligence (BI)

Business Intelligence

Business Intelligence (BI) is an intelligent business analyst, which is a process for analyzing data and presenting information about customer activity and operational activities, helping company managers, business managers, business owners to make the most effective management decisions. The BI system includes applications, infrastructure and tools, as well as advanced technologies that allow access to data and with high performance and analyze information to improve and optimize solutions that play a key role in the strategic planning process within the company. Business Intelligence is a comprehensive business solution that includes Statistics, Predictive Analytics and Intellectual Analytics proper.

Business Intelligence (BI) is an intelligent business analyst, which is a process for analyzing data and presenting information about customer activity and operational activities, helping company managers, business managers, business owners to make the most effective management decisions. The BI system includes applications, infrastructure and tools, as well as advanced technologies that allow access to data and with high performance and analyze information to improve and optimize solutions that play a key role in the strategic planning process within the company. Business Intelligence is a comprehensive business solution that includes Statistics, Predictive Analytics and Intellectual Analytics proper.

Statistics

describes the past or the current state, she does not say what to do, stating only what was and what is. In most cases, the statistics are presented in sets of reports in Excel, approved and adopted by the company. We share statistics on internal (periodic reports of operating activities) and external (exchange rates, consumer prices, political events, development of urban infrastructure, etc.). Comparing internal and external statistics without the use of technology is a rather laborious process, especially if you do not know that need to find.

Predictive analytics

built by modeling future periods, based on data from past periods. As a result, the Predictive Analytics predicts the future with varying likelihood, provided that the market situation and the company's strategy do not change significantly. Forecast analytics is often used for planning financial activities, building budgets, and calculating the effectiveness of investments.

Intellectual analytics

Unfortunately, in real life, neither statistics nor prognostic analytics in their entirety provide for business challenges, especially in the digital era of big data. They are replaced by Intellectual analytics, based on statistics and mathematical modeling. Intellectual analytics answer the questions “what to do”, “what will happen” and with what probability or “what will happen” with this or that management decision. Intelligent analytics are often referred to as advanced analytics, since its use requires the availability of a sufficient amount of statistical data, technical resources and computing power.

Statistics

describes the past or the current state, she does not say what to do, stating only what was and what is. In most cases, the statistics are presented in sets of reports in Excel, approved and adopted by the company. We share statistics on internal (periodic reports of operating activities) and external (exchange rates, consumer prices, political events, development of urban infrastructure, etc.). Comparing internal and external statistics without the use of technology is a rather laborious process, especially if you do not know that need to find.

Predictive analytics

built by modeling future periods, based on data from past periods. As a result, the Predictive Analytics predicts the future with varying likelihood, provided that the market situation and the company's strategy do not change significantly. Forecast analytics is often used for planning financial activities, building budgets, and calculating the effectiveness of investments.

Intellectual analytics

Unfortunately, in real life, neither statistics nor prognostic analytics in their entirety provide for business challenges, especially in the digital era of big data. They are replaced by Intellectual analytics, based on statistics and mathematical modeling. Intellectual analytics answer the questions “what to do”, “what will happen” and with what probability or “what will happen” with this or that management decision. Intelligent analytics are often referred to as advanced analytics, since its use requires the availability of a sufficient amount of statistical data, technical resources and computing power.

The process of implementing business intelligence

1

The customer forms the task

which needs to be addressed. There is no need to prescribe complex technical tasks with a variety of parameters and business processes. You just need to fill out a short form and write in human language what problem you need to solve. Fill out the form
2

We get access to the Company's operating data.

In the course of operating activities, each company accumulates a certain amount of data. This may be information about the personal data of customers, employees, payment transactions, marketing campaigns, prices, etc. Existing information is an array of statistical data stored in a database. In order to maintain confidentiality, we use for business analytics "impersonal" data, without a full name, phone number, passport data, etc. We also sign a confidentiality agreement with the Customer and connect to its database.
3

We expand the service for working with data

We use reliable and proven resources offered by Amazon. Learn more about services by clicking on the link . It is also possible to use our own resources, resources of the Customer or other services, for example, Microsoft or Google.
4

Assessment of the reliability, completeness, relevance and quality of data

The accuracy of the data is determined by the quality of the business processes through which data is collected and processed, how accurate the data is and reflects the reality of the processes. The completeness of the data is determined by their presence in certain periods (with or without spaces). Relevance implies that the available data corresponds to the company's current business model (there may be outdated data, for example, when a company sold vacuum cleaners and there were 3 employees, and currently already engaged in construction with several branches with a complex organizational structure). Data quality is the format and method of data storage, the degree of their integrity, in which database or databases are stored, how deeply the data exchange is integrated and to what extent, security policies, data transfer protocols, etc.
5

Preparatory stage

As a rule, this period lasts up to 2 weeks and during its length we work with data to achieve the results of the task. At this stage, our experts determine the data and signs that most affect the expected result. For example, to determine the optimal price for a new product, solvency and customer loyalty are measured, measured by average check and purchase frequency, and to increase the effectiveness of targeted marketing (buy, not buy, how much buy and with what probability) personal data of the client (gender, age, location of residence, marital status, etc.), as well as the history of his activities (when he bought, what he bought, how much he spent, etc.).
6

Predictive and probabilistic models

At this stage, we use our own algorithms, classes (neural networks, decision trees) and metrics for assessing the quality of model operation. The model building process itself looks like this: First, we use one of two approaches - either we evaluate the model when dividing the available data into training and test samples randomly, or we make an assessment of the model's predictive capabilities when testing on new data, at a time later than the training sample. Preliminary data are divided into several time slices, for each of which the model is trained on older data and tested on newer ones.
принцип моделирования
принцип моделирования
model_3 Business intelligence S2B Amazon QuickSight

To assess the accuracy of the models, we use the following metrics:

Metric # 1: Precision (model selectivity) - the percentage of correctly identified positive cases among all classified by the model as positive

Metric # 2: Recall (model sensitivity) - the percentage of correctly detected positive cases among all existing positive cases.

The integral indicator of the accuracy of the model F1 = 2 / (1 / Precision + 1 / Recall)

Unlike accuracy, expressed in percentages, the integral index does not allow the model to ignore either rare positive or rare negative cases when they are unequally distributed.

To assess the accuracy of the models, we use the following metrics:
The model building stage is completed by estimating the probability of obtaining the expected results for the given existing parameters. For example, if we take into account all the specifics of the company's operational activities for the past periods (seasonal fluctuations in the number of sales or prices, sales channels and marketing companies) and impose them on external factors (fluctuations in the dollar rate, gasoline prices, products, utilities, etc.) A prediction of the expected results will be obtained with a certain degree of error. (the result may be the number of sales in the context of products, revenue for the period, price trends, etc.). The error is determined by the standard deviation and fully depends on the completeness, relevance and reliability of the data.

Buy, buy, what and when buy, where to buy and at what price?

7

Trial operation.

If the Customer is satisfied with the forecast error, the one obtained at the modeling stage is transferred to the period of trial operation - this is the period of model calibration in real conditions. As a rule, this period lasts for 3-4 months and during this time we need to achieve positive dynamics in the results. During the period of trial operation, we compare the company’s performance results obtained using forecast models with the results without using them. To do this, we form test groups divided by employees and (or) locations (if the business is geo-dependent). The main goal of the stage is to improve the existing results and assess the extent of the improvements by calibrating the model, thanks to which the Customer receives answers to the questions - what prices, what discounts, what promotions and when is it better to send, who to send SMS, and who should call, who to offer a discount and to whom it is not necessary, to whom and which product or service is better to offer, in which place it is better to open a sales point or close, in what way is it better to deliver the goods, etc.
8

Industrial exploitation

If the pilot operation phase ends with a positive dynamic that suits the customer, the BI system can be put into commercial operation, scaling to all departments of the company, introducing appropriate interfaces for automating processes and monitoring indicators. At this stage, we fully develop the system of intellectual analytics, we establish processes for self-learning of models, we integrate with geographic information services, which allows us to correct forecasts according to a continuous process of changing data. Such technologies are used by Google giants (for example, calculating the route and rebuilding it depending on the traffic situation) Facebook (a good example of direct marketing is when it’s worth searching for something, after which the corresponding advertising appears).

Interfaces and Technologies

9

Calculations

All calculations, building predictive and probabilistic models require large technical resources. At the stage of development and testing, we usually use our own equipment - a pool of servers with large resources. After the system is transferred to commercial operation, the computing part can be transferred to the cloud, or to the production facilities of the Customer. The decisions made are negotiated with the Customer during the development process and depend on the nature of the calculations and the amount of data.
10

Amazon QuickSight

For the final presentation of the results of the work done, we use a cloud service from Amazon - QuickSight . The service allows you to connect to most popular databases, flexible in settings and stable in operation. SPICE technology based on an extremely fast parallel computing mechanism allows you to visualize big data in clear charts and diagrams, create dashboards for monitoring, and share analytics with other users. Learn more about the service here
© 2015 Solutions to business. All Rights Reserved.