Introducing G2’s New Vector Database Category

June 17, 2023

With the major driving factor behind machine learning being vector databases, it is the latest buzzword in the world of AI. Vector databases are the base for technologies like machine learning and large language models (LLMs). 

Vector databases help in extending the long-term memory of LLMs. They also enable distance metrics, indexing, and similarity search. Another use of these databases is high-dimensional data management. Vector databases store high-dimensional data in the form of vectors.

Identifying the growing demand for vector database platforms by buyers, we decided to launch this brand new category to cater to their needs. G2 launched the Vector Database Software category in April to keep up with the growing AI and machine learning space. 

Vector databases: An answer to unstructured data management 

A vector database is a new way to represent data. Vector databases store data in the form of vector embeddings. Embeddings are mathematical representations of words or sentences. They are used in natural language processing (NLP) and help efficiently analyze and manipulate text data. Multi-dimensional embeddings help in giving a better understanding of what a piece of data could be. 

Thus a vector database can be called a subset of machine learning, and embeddings are a subset of vector databases.


Vector databases have a significant role to play in various applications, such as recommendation systems, search engines, video and image analysis, anomaly detection, and so on.

Godard Abel, CEO, G2 says, "Vector databases turbocharge LLM, enhancing response speed and quality while underpinning the development of AI's long-term memory."

“G2's exciting new category leverages peer reviews, delivering unprecedented insights to all AI creators. Utilizing vector databases, we're fast-tracking Monty AI, the world's pioneering software buying assistant!”

Godard Abel

With scalability, agility, and speed, vector databases have revolutionized data analysis.

How does a vector database work?

80% of the data generated is unstructured. The sources of unstructured data usually are social media posts, audio, video, and so on. It is difficult to fit this data into a relational database because relational databases can handle only predefined schemas. Structured or semi-structured data usually have predefined formats or schemas.

Unstructured data does not have a pre-defined format and this is where vector databases stand out. It is usually stored in NoSQL databases, time series databases, and vector databases.

Now how do vector databases make this process more efficient? Let's take a look.

Data undergoes three processes:

  • Embedding: Based on the type of data, a vector embedding model is generated for the embeddings to be indexed. Embedding models turn images, text, and audio into numbers or embeddings. For example, if there are words like “dog”, “fish” and “bicycle”, these words can be represented in a 2D space where X-axis represents “animal-ness” and Y-axis represents “land-ness”. The coordinates for “dog” can be [0.9, 0.8] because a dog is an animal and very much a land creature. Similarly, for the bicycle, the coordinates could be [0.1, 0.8] because it is not an animal but used on land.
  • Indexing: The vector embeddings are now stored on vector database software using algorithms like local sensitive hashing (LSH), product quantization (PQ), etc. Indexing helps in the effective storing and retrieval of data. Referring to the example above, the embeddings for the particular vector are stored so that they help in faster relevant or similarity search.
  • Querying: When any application issues a query, it passes through the same vector embedding to generate the stored data on the vector database. This query is placed on the vector database, and the closest matching embedding is retrieved as the most suitable answer to the query.

In a traditional database, we always query for the rows where the value matches the query. In contrast, vector databases use algorithms to obtain the approximate nearest neighbor (ANN) search. Traditional methods of querying databases are based on exact matches or pre-defined criteria. ANN helps in finding the most similar vectors associated with the query from a library of vectors.

Vector Database Software ➜

Interest in vector database platforms is growing

Various startups in the vector database space have raised funds in April, such as Weaviate raising $50 million, Pinecone changing its valuation to $750 million, and Chroma raising an $18 million seed round

This is plenty of evidence that buyers are curious about vector database platforms.

Supporting this is the rise in category page views for the Natural Language Processing (NLP) Software category in February, March, and April with a whopping 32% increase from February to March in category page views and continued growth by 5% in April. 

The overall percentage rise from February to April was around 37%. With the buyers looking for NLP software, we sensed the need for creating a category for the base technology.


What is in store for vector databases?

Vector databases are closely tied up with generative AI models, which further help in anomaly detection and similarity search. With the rise of high-dimensional data, vector databases will be required to improve current AI models.

The vector database category on G2 and, in general, is bound to grow exponentially. Investors fascinated by the generative AI bug have already started investing in all interconnected areas of the AI industry, and vector databases are invaluable here. So, it should be no wonder if we see almost every LLM app integrating them.

Learn more about machine learning, how it works, and different machine learning methods. 

Edited by Shanti S Nair

Introducing G2’s New Vector Database Category G2 launched the Vector Database Software category in April. Vector database is the technology behind machine learning.
Shalaka Joshi Shalaka is a Market Research Analyst at G2, with a focus on data and design. Prior to joining G2, she has worked as a merchandiser in the apparel industry and also had a stint as a content writer. She loves reading and writing in her leisure.