// product engineering

Software & Platform Engineering

Custom software development for enterprise — from MVP to production-ready platforms. With architectural discipline instead of fast hacks, and business value over feature factory.

// about the practice

What it is and who needs it

Software development is not "hire 10 developers for 3 months". It is an engineering discipline with its own methodology: domain-driven design, architectural decisions (monolith vs. microservices), TDD/CI/CD, code review as a mandatory part of the process.

In 2026 the key advantage is not delivery speed, but long-term quality: whether the project remains maintainable in 2 years, or becomes legacy that is "easier to rewrite than support".

Signals that engineering discipline is needed

  • You need a custom product not available off the shelf.
  • Existing system requires reengineering — old code became a blocker of business agility.
  • Development team grows but delivery speed falls — time for architecture consulting.
  • New product launch with unknown risks ahead — you need from-scratch design experience.
Business promise

A product should survive year two, not only the first release

The software page sells not an outsourced team, but architectural accountability: a codebase that can be maintained, scaled and handed over.

Who feels it

The product owner wants speed, the CTO sees debt

When the roadmap grows faster than architecture, delivery first accelerates and then drops sharply. We sell control over that inflection point.

First move

Domain model and ADR before "let us just code"

We start with bounded contexts, risks and architectural decision records. This saves more time than a rushed sprint zero without decisions.

// our position

Why we do this differently

01

Delivery speed without architectural discipline is not speed — it is debt.

The first year of a feature factory gives an illusion of speed. In year two, the team spends 60% of time on refactoring and fixes. This is not a bug, it is a pattern — without architectural reviews, codebase degrades linearly with time.

02

Microservices without operational maturity = a distributed monolith, which is worse than a monolith.

Microservices pay off for teams of 50+ developers with own DevOps. For teams of 5–10, it is overengineering that doubles complexity without benefits.

03

Code review is not quality control — it is knowledge sharing and peer training.

A team where code review is a formality ("LGTM") becomes n separate experts in a year, none of whom understands the others' code. Bus factor = 1 per component. This is a first-order organizational risk.

// honest filter

When you need this — and when you don't

It is more honest to say "you do not need this yet" than to sell an engagement that will not deliver ROI.

✓ Need it

  • Custom product needed, not available on the market
  • Existing system blocks business initiatives
  • Engineering culture ready (code review, tests, CI)
  • Strategic sponsor ready on client side

✗ Not yet

  • Solvable with off-the-shelf SaaS — do not build from scratch
  • Team without testing culture — engineering maturity first
  • You want an "MVP in 2 weeks" — better path is no-code/low-code
// process

How we run the engagement

01

Discovery & domain modeling · 3–4 weeks

Interviews with process owners, event-storming for bounded contexts, risk assessment. Output: domain model and architectural decision records (ADR).

02

Architectural design · 3–4 weeks

Monolith vs. microservices choice, tech stack, API design, data schema, IaC approach.

03

MVP/Pilot · 3–4 months

First version for one use case with real users. Proves technical architecture and business assumptions at production quality.

04

Scale-out · 6–18 months

Phased functionality expansion, performance optimization, adding new features by business value priority.

05

Continuous engineering · ongoing

Regular architectural reviews, refactoring sprints, dependency updates, security audits. Without this, code degrades in 12–18 months.

Project lead: Softengi (custom software, AI/ML systems) and InBase (UnityBase low-code, enterprise platforms).
Brought in when needed: SL Global Service (cloud-native architecture), Data Management IG (data layer design).

// anti-patterns

Typical mistakes we have seen projects fail on

Feature factory without architectural oversight

Team delivered 50 features in year one. In year two, 30% of time goes to bug-fixing, another 30% to refactoring. Business asks why speed dropped.

What we do instead: mandatory architectural reviews per epic, refactoring as part of sprint capacity (not "later").

Microservices "to grow into"

A team of 6 developers builds 12 microservices "because it is modern". DevOps overhead exceeds business logic.

What we do instead: we start with a monolith and extract services only when there is evidence-based reason: different scaling profiles, different release cycles, or teams >50 engineers.

Tests "when there is time"

Test coverage <10%. Every release is a week of manual QA. After 18 months, no refactoring is possible.

What we do instead: tests as part of definition-of-done (not optional). Target coverage >70% for business logic.

// experience

Typical scenarios from our practice

Each example shows the client-side story: what was getting in the way, what changed, and what result the team got.

Bank

A workspace built around real roles

Problem
The off-the-shelf solution did not fit internal rules, so the team worked through workaround spreadsheets.
What we did
We designed the workspace around roles, permissions, and daily user scenarios.
Result
Fewer manual workarounds and more control over product development.
Energy group

Replacing spreadsheets with a managed process

Problem
Critical processes depended on hundreds of files that were hard to check and update.
What we did
We moved processes in phases, starting with the most frequent operations.
Result
Teams work with shared data, and errors are visible before they affect the business.
Public service

Stable work with external requests

Problem
External requests kept growing, and the old routine could not handle the load.
What we did
We created a controlled entry point for requests and processing rules for different participants.
Result
The service became more stable, and onboarding new participants became easier.
// deep dives

Articles on this topic

Nine recent expert articles — from thematic overviews to specific architectural decisions.

// toolkit

Technologies we work with

Backend

Java · Python · .NET · Node.js · Go · PHP (UnityBase)

Frontend

React · Vue · Angular · Next.js · TypeScript · Svelte

Databases

PostgreSQL · MS SQL · MongoDB · Redis · ClickHouse · Elasticsearch

Architecture

Domain-Driven Design · CQRS · Event Sourcing · API-first · Hexagonal architecture

DevOps & QA

Docker · Kubernetes · GitLab CI · GitHub Actions · Playwright · Cypress · JUnit · pytest

Standards

ISO/IEC 27001 · OWASP · SOLID · 12-factor app · Clean Architecture

// answers

Frequently asked questions

Monolith or microservices for a new project?

Almost always — modular monolith. Microservices pay off only for teams of 50+ with established DevOps culture. For everyone else — overengineering with worse operational profile.

How long does a custom product take?

MVP — 3–4 months. Production-ready — 6–12 months. If you are promised "product in 2 months" — it is either a templated solution (then why custom?) or a product debt to be paid later.

How to measure development team quality?

Three metrics: (1) lead time from commit to production (target <1 day); (2) deployment frequency (target several times per week); (3) change failure rate (target <15%). These are DORA metrics — the industry standard.

Is TDD still justified in 2026?

For business logic — yes, mandatory. For UI and scratch prototypes — opt-in. Classic test-first → green → refactor cycle catches 80% of bugs before code review.

What is Domain-Driven Design?

An approach to modeling complex business domains via ubiquitous language with business stakeholders and explicit bounded contexts. Pays off for projects >6 months with non-trivial business logic. For simple CRUD — overhead.

// adjacent areas

Related competencies

Real projects rarely fit in one competency. See which other areas we work in.

Tell us about the product you want to build

30-minute discovery call with an architect. We will discuss your idea, technical risks and realistic expectations.

Alliance model

Intecracy Group does not force a single delivery team. We clarify the task, identify the required competencies and help involve the relevant alliance members.

Contact directly