In English language, we often come across similar sounding words (homophones) with different meanings. e.g. stationary vs. stationery; bear vs. beer; their vs. there and many more.

In today’s world of digital transformation, we all use multiple technology buzzwords, without necessarily knowing the differences across many of them. I personally have come across multiple concepts which seem similar but are different when you understand them properly.

To be able to articulate the right digital transformation story, it is important to understand the nuances of those concepts well, so that you personally feel confident and your audience finds you credible. I am sure there are many such technology concepts, but here are the top ones which I have personally found confusing in the past:

1.Artificial Intelligence vs. Machine Learning vs. Automation vs. Natural Language Processing vs. OCR

Artificial Intelligence: As the name suggests, AI is a system which mimics human intelligence. You can imagine AI enabled machines doing similar tasks as a human would do (e.g. self-driven cars, playing chess, conversational bots). AI is a broader concept which incorporates concepts of ML, Automation, NLP etc.

Machine Learning: Machine Learning systems can be simply viewed as auto-classification systems. Given a set of training data sets, the system tries to classify new data in one of the categories present in training data set. Richer the training data, better the ability of ML system to correctly classify new data. Typical examples of ML are in Email spam filtering, demand forecasting etc.

Automation: Automation refers to executing a well-defined sequence of steps using a machine instead of a human. Examples include downloading data from a website, moving files from one folder to another, automation in manufacturing assembly lines etc.

Natural Language Processing: It is the ability of a system to interpret natural language (which humans talk). Classic examples of NLP enabled systems are Chatbots, which can interact with humans over a Web interface in natural language.

OCR (Optical Character recognition): It is a very specific use case (usually part of Automation) which enables a system to take typed / hand-written texts as input and convert them to digital entities for further processing by humans/downstream systems.

2.Omni channel vs. Multi-channel

Omni channel: Customer is at the centre. This involves creation of a single customer view and context aggregated across all the channels through which customer interacts with an organization.

As a simple example, consider a customer registering a service request for a product replacement over a call centre. If the line gets disconnected and the customer decides to close the request over chat, the person (or bot) taking the request should already have context of previous interaction which customer did over phone.

This is a very powerful concept which is being used by digitally advanced enterprises to create Wow customer service experiences and drive retention/cross-sell/upsell.

Multi-channel: This revolves around an organization’s product and allows customers to engage with the organization through multiple (and often siloed) channels, such as branches, website, mobile app. Once these multiple channels are integrated to create customer 360-degree view, it becomes omni-channel

3. Data lake vs. Data warehouse vs. Database

Data Lake:

Storage: Can store unstructured, semi structured or structured data
Processing: ELT methodology (Extract, Load, Transform). Ingests data first and transformation done later depending on use case
End use case of Advanced Analytics

Data Warehouse

Storage: Can store only structured data
Processing: ETL methodology (Extract, Transform, Load). Transformation is needed to a structured format before data can be stored
End use case: Store large volume of historical data and enable fast queries across data, typically using Online Analytical Processing (OLAP). This enables comprehensive and granular business reporting / insights, without incurring a lot of manual effort and time.

Database

Storage: Can store only structured data (graph, relational etc.)
Data processing: ETL methodology (Extract, Transform, Load). Transformation is needed to a structured format before data can be stored
Purpose: Store current transactions and enable fast access to certain transactions for ongoing business, known as Online Transaction Processing (OLTP)

4. Data Scientist vs. Data Engineer

Data Scientist: Works on developing the end use cases of Data Science which can answer business problems. A Data Scientist builds sophisticated analytics programs and statistical models (predictive, prescriptive) etc. using the languages such as SPSS, R, Python etc.

Data Engineer: Works towards making sure that data is available for the Data Scientists to use for their models. Usually the data used in statistical models is from different sources and may not always be complete/clean. A data engineer maintains architectures and large scale processing systems to make sure that the data from different sources is usable by data scientists

5. Master Data Management vs. Metadata Management (both MDMs)

Master Data Management: Master data refers to full 360-degree data (including meta information and actual values) which can be a treated as a Master / Single source of truth

Meta Data Management: Meta data refers to data about data; which means all the attributes associated with a given data set. It helps understand the details of what, when, where, how etc. for the data (e.g. when the data was acquired, source of data, end use of this data, changes which happened to data over time)

6.Structured vs. Semi-structured vs. unstructured data

Structured: e.g. excel files / CSV files. The fields are properly defined and the corresponding values in those fields usually are of a particular type (e.g. numeric, text)

Semi-structured: E.g. log files. Data is not fully segmented across different fields like in excels. However, data still follows a pre-defined pattern and it is easy to extract the right information with some rules

Unstructured: E.g. audio, video, transcripts, click-stream. Data does not have a defined format. Usually the power of Data Lakes and Advanced Analytics comes to life for this type of data and can do wonders for organizations if done properly

7.DevOps vs. Agile

DevOps: It is a practice of bringing Development and Operations together. In a traditional setup, developers develop a piece of functionality, which is then passed on to an Operations team to deploy for end user. In DevOps, the development and operations are integrated in a single team. DevOps bridges the gap between Developers and IT Operations

Agile: Agile refers to an iterative approach of product development, where there is a strong collaboration between customers, developers and testers. In a traditional setup (waterfall), the requirements are gathered at the beginning of a project, a detailed Business Requirement Document is created, and final product comes out at the end of a long development cycle.

This poses risk of not capturing changing / evolving needs of customers. In an agile world, an incremental version of end-product is continuously developed, tested and aligned with the customer. Agile bridges the gap between customers, developers and testers

8. IaaS vs. PaaS vs. SaaS

IaaS: Infrastructure as a Service means using something like a virtual data centre where all the core infrastructure elements are on Cloud (E.g. Servers, Storage, Networking). An example is EC2 instance from AWS

PaaS: Platform as a Service. In addition to making core infrastructure available on Cloud (as described in IaaS), PaaS offers a platform (e.g. Operating systems, databases, run-time environments) where developers can build their own software. An example is Google Anthos

SaaS: This includes a set of applications which are deployed on Cloud and can be directly used by end users (e.g. through a web browser). A common example is Gmail.