Tech Radar

Technology choices with practical criteria: strengths, limitations, MVP suitability and production responsibilities. Informed by experience on both sides.

Adopt

Evaluate

Caution

Avoid

MVP / Production (1-5)

Databases

Database choices are expensive to reverse. Evaluate the model, access patterns and migration cost before committing.

PostgreSQL

A relational database that grows with you

Adopt

MVP

Production

Strengths

+Full ACID support
+Native JSON (JSONB) alongside relational data
+Extensible: PostGIS, pgvector, pg_trgm
+Open source with mature enterprise support
+Strong OLTP and lightweight analytical queries

Limitations

−Multi-region replication requires careful design
−High-concurrency configurations require tuning
−BI integration may require more setup than SQL Server

Use for

A strong default for SaaS, e-commerce, APIs and transactional systems.

Avoid when

The team already operates another relational database well and migration offers no clear benefit.

SQL Server

Mature enterprise tooling in the Microsoft ecosystem

Evaluate

MVP

Production

Strengths

+Native integration with Power BI and Azure
+SQL Server Agent for scheduled jobs
+SSRS and SSIS analytics tooling
+Established enterprise support

Limitations

−Production licensing can be significant
−Dependence on the Microsoft ecosystem
−Operational overhead for small projects
−Fewer hosting choices than PostgreSQL

Use for

Organizations using Microsoft/Azure, teams with T-SQL expertise, and strong Power BI integration needs.

Avoid when

Limited licensing budgets or cloud-agnostic teams where PostgreSQL meets the requirements.

MongoDB

Flexible for exploration, demanding at scale

Caution

MVP

Production

Strengths

+Flexible schema while the model evolves
+Fast MVP development
+Built-in horizontal scaling through sharding
+Suited to document-shaped logs, events and catalogs

Limitations

−Joins through $lookup require attention to modeling and cost
−Multi-document transactions add complexity
−A modeling learning curve for relational teams
−Atlas production costs need monitoring

Use for

Evolving document models, variable product catalogs, logs, events and content systems.

Avoid when

The domain is strongly relational and consistency requirements are easier to meet with a relational model.

Redis

Caching and ephemeral state

Adopt

MVP

Production

Strengths

+Low-latency in-memory access
+Rich data structures: sorted sets, streams and pub/sub
+Native per-key TTL
+Established clustering and replication

Limitations

−Memory capacity and persistence choices need planning
−Durability depends on RDB/AOF configuration and recovery design
−Large values increase memory, network and eviction costs

Use for

Sessions, rate limiting, lightweight queues, pub/sub, counters and API caches with TTLs.

Avoid when

Transactional source-of-truth requirements without a validated durability design, or large binary payloads.

MySQL / MariaDB

Established and widely supported

Evaluate

MVP

Production

Strengths

+Available on most hosting platforms
+Large community
+Strong simple OLTP read workloads
+MariaDB offers its own extensions and features

Limitations

−Compare analytical and extension needs with PostgreSQL
−JSON features differ from PostgreSQL JSONB
−A different extension ecosystem
−Replication settings and failover need explicit planning

Use for

Existing hosting requirements, WordPress and LAMP applications, or an experienced operations team.

Avoid when

New analytical workloads where PostgreSQL better matches the requirements.

DynamoDB / CosmosDB

Large-scale access with demanding modeling

Caution

MVP

Production

Strengths

+Managed horizontal scaling
+Managed database infrastructure
+Designed for predictable access latency
+Sem schema: flexível na estrutura

Limitations

−Access patterns must drive the data model
−A substantial modeling learning curve
−Inefficient access patterns can increase costs
−A query model distinct from SQL

Use for

Large workloads with well-understood access patterns, such as IoT, gaming and global sessions.

Avoid when

Rapidly changing models, ad-hoc queries or teams unfamiliar with key-value NoSQL design.

Business Intelligence & Analytics

Choose BI around the users and questions it must answer. Adoption by the business team matters more than popularity.

Power BI

Accessible BI in the Microsoft ecosystem

Adopt

MVP

Production

Strengths

+Integration with Excel, Azure and SQL Server
+Evaluate licensing against the existing Microsoft plan
+An approachable learning curve for business users
+DAX supports complex measures
+Integrated publishing and sharing workflows

Limitations

−Large models require capacity and licensing planning
−Visual customization differs from Tableau
−Microsoft ecosystem dependence
−Web and desktop capabilities differ

Use for

Microsoft-oriented organizations, business self-service, executive reports and operational dashboards.

Avoid when

A different cloud stack or highly specialized visualizations make another tool a better fit.

Tableau

Advanced visualization with substantial cost

Evaluate

MVP

Production

Strengths

+Rich, flexible visualization capabilities
+Extracts support large analytical workloads
+Strong community and learning resources
+Tableau Prep for self-service data preparation

Limitations

−Significant per-user licensing costs
−Product direction is part of the Salesforce ecosystem
−Complex reports require deeper technical skills
−Cost can be difficult to justify for small operations

Use for

Larger organizations with dedicated analytics teams and demanding visualization needs.

Avoid when

Budget-constrained startups, small businesses and MVPs without dedicated analysts.

Looker / Looker Studio

Semantic modeling and BI governance

Evaluate

MVP

Production

Strengths

+LookML provides reusable semantic definitions
+Looker Studio offers an accessible dashboarding option
+Integration with BigQuery and Google Cloud
+Governed self-service exploration

Limitations

−Enterprise implementation can be costly and complex
−LookML has its own learning curve
−A smaller ecosystem than Power BI or Tableau

Use for

Google Cloud organizations and teams needing centralized metric governance. Evaluate Looker Studio for simpler dashboards.

Avoid when

A Microsoft-oriented stack or insufficient resources to maintain a semantic layer.

Apache Superset

Powerful open source with operational responsibility

Evaluate

MVP

Production

Strengths

+Open-source software without a software license fee
+Supports many data sources
+Customizable and extensible
+Suited to technical teams

Limitations

−You operate and maintain the infrastructure
−User experience is less polished than some commercial tools
−Support options depend on the deployment and provider
−Performance depends on your infrastructure

Use for

Technical teams seeking control, self-hosting and predictable licensing, including restricted data environments.

Avoid when

Nontechnical teams or organizations unable to maintain the platform.

Qlik Sense

An associative engine for specific analytical needs

Caution

MVP

Production

Strengths

+Associative exploration across data relationships
+Useful for correlation analysis and complex drill-down
+An established QlikView installed base

Limitations

−High cost
−A smaller, more proprietary ecosystem
−Significant learning curve
−Assess the available talent and ecosystem locally

Use for

Organizations with existing Qlik investment and expertise, or specific associative exploration needs.

Avoid when

New projects without Qlik requirements where a simpler BI tool suffices.

Messaging & Queues

Understand the difference between work queues and event streams. Their retention, replay and delivery semantics are different.

RabbitMQ

Established, familiar messaging

Adopt

MVP

Production

Strengths

+Mature, widely supported AMQP protocol
+Roteamento flexível de mensagens
+Useful management interface
+Can run locally or in the cloud

Limitations

−Queue and stream semantics differ; choose the appropriate mode
−Scaling design differs from Kafka
−Replay depends on using streams or a separate retention design

Use for

Asynchronous processing, notifications, service decoupling and background tasks.

Avoid when

A durable event log with replay is the primary requirement and a streaming platform fits better.

Apache Kafka

Event streaming at scale with real complexity

Caution

MVP

Production

Strengths

+Retained events with replay within the retention window
+High throughput with suitable infrastructure
+Can underpin an event-sourcing design
+Streaming integrations with Flink and Spark

Limitations

−Operating it requires expertise or a managed service
−Can add unnecessary complexity to simple workloads
−Steep learning curve
−Latency trade-offs differ from a simple work queue

Use for

Event histories, real-time analytics and data pipelines with independent consumers.

Avoid when

Small workloads without replay needs or teams lacking streaming experience.

Cloud Queues (SQS/Pub-Sub)

Managed messaging with usage-based pricing

Adopt

MVP

Production

Strengths

+Provider-managed infrastructure
+Automatic scaling
+Usage-based costs rather than an idle cluster
+Native dead-letter handling

Limitations

−Cloud-provider dependence
−Routing capabilities differ from RabbitMQ
−High-volume costs need monitoring

Use for

Cloud-native MVPs and production systems. Evaluate SQS on AWS, Pub/Sub on GCP or Service Bus on Azure.

Avoid when

Vendor lock-in is a firm project constraint.

AI & Emerging Technologies

The AI market evolves quickly. The decision criteria remain: which problem it solves, operating cost, and how provider changes affect you.

Vertex AI / Gemini

Google models integrated with GCP

Adopt

MVP

Production

Strengths

+Gemini models for reasoning and multimodal tasks
+Integration with BigQuery, Cloud Functions and GKE
+Google Search grounding options
+Compare model tiers, token prices and task quality

Limitations

−GCP ecosystem dependence
−APIs and model versions evolve quickly
−Validate documentation against the selected version

Use for

GCP teams, multimodal applications, BigQuery-backed retrieval and production assistants.

Avoid when

An AWS-only or Azure-only stack already has a suitable managed AI platform.

AWS Bedrock

Managed multi-model access within AWS

Evaluate

MVP

Production

Strengths

+Multiple model providers through a common AWS interface
+Knowledge Bases for managed retrieval
+Configurable policy guardrails
+Managed model infrastructure

Limitations

−High-volume costs need monitoring
−Console workflows can require familiarity
−New model availability varies by provider and region

Use for

AWS organizations needing managed access to several AI model providers.

Avoid when

A direct provider API better fits a simple prototype.

Ollama / vLLM (IA Local)

Local models for controlled data handling and latency

Adopt

MVP

Production

Strengths

+Data can remain local when the full pipeline is configured accordingly
+Inference without a remote network round trip
+Hardware-based costs with capacity planning
+A choice of open-weight models for different hardware budgets

Limitations

−Smaller models may be less capable on complex reasoning tasks
−Suitable GPUs or dedicated hardware may be needed
−Tools and integrations require your own operational work

Use for

Classification, extraction, moderation, controlled log analysis and request routing.

Avoid when

The available hardware and model cannot meet quality, context-window or latency requirements.

pgvector (PostgreSQL)

Vector search in the database you already use

Adopt

MVP

Production

Strengths

+No separate vector service if you already run PostgreSQL
+SQL for hybrid vector queries and filters
+Reuses the existing database platform
+HNSW and IVFFlat indexing options

Limitations

−Large vector collections require benchmarking and tuning
−Fewer specialized features than some dedicated vector platforms
−Globally distributed search requires additional design

Use for

RAG, semantic search alongside transactional data, and projects already using PostgreSQL.

Avoid when

Very large vector collections or global distribution requirements exceed the tested setup.

Edge AI (TensorFlow Lite / ONNX)

On-device AI without a cloud inference round trip

Evaluate

MVP

Production

Strengths

+Offline inference on mobile and IoT devices
+Can reduce external data transfer with appropriate controls
+No network latency for local inference
+Can work offline in the field or in factories

Limitations

−Smaller models with capability trade-offs
−Model development and optimization can be complex
−Device hardware limits model size and performance

Use for

Factory visual inspection, field-device classification, offline OCR and edge-camera detection.

Avoid when

Model quality or size exceeds the device's capabilities and a permitted cloud service is acceptable.

Frontend Frameworks

Frontend architecture affects perceived speed, accessibility, SEO and team productivity. Choose against your product's real requirements.

React / Next.js

A broad ecosystem for complex applications

Adopt

MVP

Production

Strengths

+Large ecosystem
+App Router and Server Components support server rendering and performance work
+Straightforward deployment options on Vercel
+A broad hiring pool in many markets

Limitations

−App Router introduces a learning curve
−Can add unnecessary complexity to simple static sites

Use for

Interactive SaaS products, e-commerce and complex dashboards with a React-oriented team.

Avoid when

Simple content pages where minimal JavaScript is the main goal; evaluate Astro.

Astro

Minimal JavaScript by default for content sites

Adopt

MVP

Production

Strengths

+Island architecture with selective client hydration
+A strong starting point for fast content delivery
+Supports React, Vue, Svelte and other UI integrations

Limitations

−Complex application-wide state may need a different architecture
−A smaller integration ecosystem than some application frameworks

Use for

Marketing pages, blogs, documentation and content-led commerce.

Avoid when

Highly interactive applications with complex shared client state.

Vue / Nuxt

An expressive reactive alternative

Evaluate

MVP

Production

Strengths

+Expressive Composition API
+Built-in reactive data model
+Nuxt supports SSR and static generation

Limitations

−A smaller enterprise ecosystem than React in some markets
−Legacy Vue 2 migrations need planning

Use for

Teams already experienced with Vue or products well matched to its reactive model.

Avoid when

Local hiring availability and existing React expertise are decisive constraints.

SvelteKit

Compiled reactivity without a virtual DOM

Evaluate

MVP

Production

Strengths

+Compiler-based optimizations reduce runtime work
+Fine-grained reactivity with runes
+Productive development experience
+Built-in animation features

Limitations

−Smaller community
−Fewer ready-made UI libraries for specialized needs

Use for

Teams prioritizing lightweight interactions and comfortable with a smaller ecosystem.

Avoid when

Large projects relying on a broad catalog of established UI libraries.

Angular

An integrated enterprise toolkit

Evaluate

MVP

Production

Strengths

+Integrated router, HTTP client and forms
+RxJS for complex asynchronous flows
+Standalone components simplify application structure
+Strong TypeScript integration

Limitations

−A substantial learning curve
−Significant framework overhead for small projects
−Reactive flows can become complex

Use for

Large enterprise teams, Java/C# backgrounds and applications needing shared conventions.

Avoid when

Small MVP teams with limited scope and a need for rapid UI iteration.

Backend Frameworks

Architecture matters more than language alone. Each ecosystem offers different strengths and operating costs.

Node.js (TypeScript)

Fast iteration with a shared ecosystem

Adopt

MVP

Production

Strengths

+TypeScript across frontend and backend
+Large npm package ecosystem
+Strong fit for I/O-heavy services

Limitations

−CPU-heavy work needs workers, processes or dedicated services
−Memory behavior requires monitoring
−External dependencies require careful selection

Use for

MVPs, GraphQL APIs, I/O-bound services and backends for frontends.

Avoid when

CPU-heavy mathematical or data-processing workloads without a suitable offloading design.

Python (FastAPI)

A strong data and AI ecosystem with modern APIs

Adopt

MVP

Production

Strengths

+Readable, productive syntax
+FastAPI provides typing, async support and validation
+Extensive AI and data libraries

Limitations

−Threading behavior depends on the Python build and libraries
−CPU-intensive Python code may need compiled libraries or other runtimes

Use for

AI-backed APIs, data processing, machine learning and AI-oriented services.

Avoid when

Very high connection or CPU demands without validated runtime and deployment choices.

C# / .NET

A modern, strongly typed enterprise platform

Adopt

MVP

Production

Strengths

+Expressive C# features, including LINQ and pattern matching
+Strong performance in modern .NET
+Minimal APIs reduce boilerplate

Limitations

−Outdated tutorials can obscure modern practices
−Cross-platform support includes Linux

Use for

Complex enterprise domains and high-performance typed services with an experienced team.

Avoid when

Teams without platform experience that need immediate iteration speed.

Java / Spring Boot

An established choice in enterprise systems

Evaluate

MVP

Production

Strengths

+Mature enterprise ecosystem
+Established JVM production tooling
+Broad Spring integrations and conventions

Limitations

−Boilerplate can be substantial
−Memory overhead matters for small services
−Startup time requires attention in short-lived deployments

Use for

Banks, insurers and enterprises with JVM expertise and established operational requirements.

Avoid when

Small teams or ephemeral services where startup and resource costs dominate.

Go

Simplicity, concurrency and efficient execution

Adopt

MVP

Production

Strengths

+Lightweight goroutines for concurrent work
+Single-binary deployment options
+Efficient memory and CPU use
+A deliberately small language

Limitations

−Explicit code can be repetitive
−Verbose explicit error handling (`if err != nil`)

Use for

High-scale services, gateways, networking infrastructure and concurrent processing.

Avoid when

Domains that benefit more from richer language abstractions than from a minimal runtime.

Rust

High performance with strong memory-safety guarantees

Caution

MVP

Production

Strengths

+Safe Rust prevents many memory errors and data races
+Performance comparable to C++ in suitable workloads
+Low resource overhead
+No tracing garbage collector

Limitations

−A demanding ownership and borrow-checking model
−Initial development can be slower
−A smaller available hiring pool

Use for

Systems programming, WebAssembly, CLI tools, cryptographic systems and latency-sensitive services.

Avoid when

Routine CRUD or MVPs where immediate delivery speed outweighs systems-level control.

Elixir

High concurrency and fault-tolerant design

Evaluate

MVP

Production

Strengths

+OTP supervision supports failure recovery
+Concurrency through lightweight processes
+Well suited to many persistent WebSocket connections
+Expressive functional code

Limitations

−Functional programming requires a mindset shift
−Smaller ecosystem than TypeScript or Python
−Hiring may be more specialized

Use for

Real-time chat, games, IoT, distributed services and large WebSocket workloads.

Avoid when

Heavy numerical work without native integrations, or teams unwilling to adopt functional programming.

Mobile

Mobile has its own platform constraints. Compare cross-platform and native options against the required hardware, UX and operating model.

React Native (Expo)

A bridge between web and mobile teams

Adopt

MVP

Production

Strengths

+Expo simplifies common native project setup
+OTA updates within platform policies and runtime compatibility
+Builds on the React ecosystem
+Expo Router supports shared web and mobile navigation

Limitations

−Complex animations and native interactions still need profiling
−Custom native modules add maintenance work

Use for

Web teams building mobile products, B2C apps and MVPs with frequent iteration.

Avoid when

Required hardware integrations are unsupported or demand extensive custom native work.

Flutter

Consistent rendering with detailed UI control

Adopt

MVP

Production

Strengths

+Its own rendering pipeline using Skia/Impeller
+Consistent UI across iOS and Android
+Smooth animations when frame budgets are met
+Productive hot reload

Limitations

−Dart is a more specialized language choice
−Application binaries can be larger
−Code-update options need third-party tooling and policy checks

Use for

B2B apps, point-of-sale systems, mobile dashboards and visually demanding interfaces.

Avoid when

The team prefers to reuse TypeScript expertise and has no capacity to adopt Dart.

Technology is a tool. Choose what the team can operate, what fits the budget, and what solves today's problem without blocking future growth.

Czanix Tech Reference

Tech Reference Boilerplates on GitHub

Choosing a stack for a new product and want a second technical opinion?

Talk to Cesar Zanis