Self-Service – Business Intelligence

Self-Service

The responsibility for delivery has been shifting from the IT area to the business area. Driven by business’ desire for more control and ownership, business teams responsible for digital delivery often have little or no link to the IT department. While this is a positive trend, there is a risk that organizations will become fragmented and highly inefficient. The right method would be to formalize the right combination of business stewardship and IT stewardship to maintain efficiency and at the same time give ownership to the business department. This way, IT could focus on being an enabler of business.

Benefits to this arrangement are as follows:

•    Increase focus on the specific and most important business goals and outcomes

•    Have better understanding of their customers and their needs

•    Increase ability to change quickly and easily

•    Increase innovation and collaboration across organization

•    Develop optimal solutions and remove duplicate efforts, resources, and solutions

•    Integrate Local build black box solution with the enterprise-wide architecture using best working practices, standards, and methods.

•    Self-service analytics require advance tools that necessitate building descriptive, predictive, and prescriptive capabilities. However, data management with drop-and-drag capabilities is limited to descriptive capabilities.

•    Data science and machine learning platforms require governance of data science models and tools.

Advanced BI Analytics

Text analytics: Derives business insights from structured/unstructured text data. Determines, classifies, and extracts key entities, and summarizes text and identifies the tone or sentiment of texts.

Most organizations have large quantities of unstructured text data in the form of memos, company documents, emails, communications, websites, social media posts, and blogs. Most businesses don’t know the value of data, what to extract from where, or where, when, or how to extract value from text data. Some of the use cases for text analytics are text categorization, text clustering, concept extraction of the most relevant text, assessment of opinions and sentiment, and summarization of documents. These use cases may analyze the contents of both internal and external documents, including emails, social media posts, etc.

Sentiment/opinion analytics: Seeks to extract data from text, video, or audio data to understand opinion, attitude, or sentiment of internal or external consumers. The advance polarity sentiment analytics can also go further by classifying the emotional state of person by using facial expression. This type of BI analytics is popular for social media data where people are showing a variety of emotions. Health care, insurance, finance, legal, retail, marketing, law enforcement, and digital publications are some examples of industries, but not all.

Image analytics: Seeks to extract information for images and processes images for the purpose of finding patterns and metadata. After processing, needs to recognize these images or graphics.

Image analytics can be used in facial recognition in the security industry. Another use case is recognizing brand or product in photographs shared on public social media platforms for the retail industry. Casinos use face recognition to identify high rollers and provide special treatment. Image recognition in medical CT scanning is a value addition. Some other use cases can come from security services/forces. An important point here is to identify value through answering strategic questions and delivering on long-term goals.

Voice analytics: Another important case is voice analytics. Gathering, storing, and processing files, and having voice analytics in the customer relations management area, can lead to insights. These insights could help you to spot these potential pitfalls before customers take to social media to complain. Another popular use case is related to the security industry.

Some of the use cases related to voice analytics are as follows:

•    Proactively identify upset customers by analyzing the pitch and intonation of conversation.

•    Proactively intervene in cases before escalation.

•    Help to identify underperforming customer service representatives for additional training or coaching in call center.

•    Secure the devices by implementing voice recognition.

•    Other cases in security industry

Stream analytics: This is the stream processing of events for the purpose of stream data integration. This is applied to data in motion to enable real-time situational awareness and near-real-time responses to threats and opportunities, and they merge, or it stores data streams for use in applications downstream. This enables faster and more precise decision-making. Sources can include IOT sensors, digital control systems, social computing platforms, news and weather feeds, data brokers, and so on.

Natural language analytics: Business users utilize an interface for the creation/ consumption of analytics content. They search for information via a query/search/chat box using business terms, whether by typing or through their voice. These queries are translated into natural language questions using NL processing technologies and/or by using a keyword search. These are supported by supporting the querying of structured data or a semantic search of multiple structured bits of information.

Foundation models like BERT (bidirectional encoder representations from transformers) and generative pre-trained transformer (GPT) techniques, advanced text analytics, deep learning, natural language generation (NLG), and natural language query techniques.

NLQ can interpret geospatial questions and immediately deliver location-based answers and business insights; e.g., relevant food, hospital, and businesses nearby.

Graph analytics: Exploration and discovery of trends and relationships between entities, people, and transactions can determine any connections across data points. This is a multi-context visualization tool that can inform insights and decision-making by using path analysis, network coordination, clustering, outlier detection, Markov chains, and more. Knowledge graph insights can be used for analyzing optimized supply tracing, disease tracking (not forgetting COVID-19), fighting fraud, supply chain tracing, etc., by identifying outliers and unusual patterns in relationship data. This consists of

models that connect the dots across data points in the form of visualization for better insights and decision-making. Established AI techniques can increase the power of knowledge graphs.

ModelOps: Model operationalization is primarily focused on the life-cycle management of ML (machine learning), AI (artificial intelligence), and NLP (natural language processing) models. This includes creating policies, procedures, and rules optimizations and dependencies associated with models, such as the following:

•    ModelOps helps in enabling standardizing, scaling, AI, and analytics by combining statistical and machine learning models.

•    Move models from the lab environment to production environment.

•    Operationalize and scale these models.

•    Monitor and govern machine learning model.

There is a wide range of risk management concerns across different models, like drifting, bias, integrity, and so on. Comprehensive governance also includes data, application, and infrastructure.

As the number of analytics, AI, and decision models at an organization increase, and as projects become more complex and complicated in needing to manage different type of analytics, AI, and decision models, governance policies and procedures for development, testing, automation, and maintenance are required. Governance ensures collaboration among all business departments regarding development of analytics models and associated KPIs and deployment of these models in production.

Extend the skills of an organization’s ML experts to operationalize a wide range of models. Recruit/upskill additional AI to cover graph analytics, optimization, or other required techniques for composite AI. Skills for knowledge engineering should also be available.

DataOps: Data operations is a part of data management focused on improving interoperability, automation, observability, and operations. It involves analyzing and developing dataflows—through process-oriented methodologies—for designing, developing, and delivering analytics in an agile and collaborative way.

DataOps is a response to friction around the consumption and use of data across the organization. Some of the metrics for DataOps are as follows:

1. How much time is taken to deliver data pipelines

2. How many usable datasets are delivered

 3.  How often codes, tables, etc. move to production

 4.  How many request tickets related to data are resolved

 5.  Error rate in production (code errors, data integrity errors)

6. Measuring qualitatively, how the delivery process has become predictable, observable, and repeatable

7. Self-service: what is the adoption rate of self-service users with enabled tools?

8. How many data features are reused or foster reusability and standards