azyware
Technology

Application modernization vs rewrite: the honest comparison

EZ
Eazyware
· Updated · 6 min read
Quick answer

Application modernization vs rewrite: which should you choose?

Modernize in slices unless the assessment proves a rewrite is unavoidable. Rewrites take years, stall on migration risk and rebuild behaviour nobody documented; modernization keeps the business running while the system changes underneath it.

Every company with software older than a decade eventually hears the same pitch: throw it away and start again. Sometimes it is right. Usually it is a two-year bet that fails on the thing nobody planned for, migrating a business that never stopped running. This comparison sets out what each path really costs, when a rewrite is genuinely the answer, how incremental modernization works in practice, and how to decide with an assessment rather than a sales deck.

What the two paths actually involve

RewriteModernization (strangler pattern)
ApproachBuild a new system in parallel, then cut overWrap the existing system with an API, replace modules one at a time behind it
Timeline12–36 months before any valueValue from the first phase; 8–16 weeks per phase
Risk profileOne large cutover; all-or-nothingMany small cutovers, each with rollback
Undocumented behaviourRediscovered during migration, expensivelyCaptured by characterisation tests before change
Business disruptionParallel running, retraining, data migration in one eventScheduled around the calendar; zero-downtime slices
Cost shapeLarge, front-loaded, hard to stopPhased, priced per slice, can stop after any phase
AI readinessOnly after the rewrite shipsAPI façade enables AI features within the first phase

Why rewrites fail

Not for lack of skill. They fail because the old system encodes years of decisions nobody wrote down: the fee calculation with a special case for one customer, the report finance depends on, the cron job that fixes data every night. A rewrite rediscovers each of these as a bug report after cutover. Meanwhile the old system keeps changing, because the business does, so the new one chases a moving target. Joel Spolsky's warning against rewrites is old and still correct.

When a rewrite is genuinely the answer

  • The system is small enough to rebuild in weeks, not years
  • The platform is dead: no runtime, no security patches, no engineers who can read it, and no way to wrap it
  • The business process it encodes is being replaced, not modernised
  • It is scheduled for decommissioning and only needs to survive until then
  • The assessment shows more than half the code must change anyway

If none of these apply, and they usually do not, modernization is the safer path.

How modernization works

1. Assess

Two weeks to map what is actually there: modules, dependencies, data, integrations, scheduled jobs, and a risk-and-effort rating for each part. The output is a modernization map and a phased plan with prices.

2. Characterise

Write tests against the live system's real behaviour before touching it, so refactoring is checked against what the system does, not what people remember. This is the safety net for everything that follows.

3. Wrap

Put an API façade over the legacy application and database. Every module now has a clean interface, and replacing one becomes an independent choice rather than a dependency chain. The façade also lets AI features, partner integrations and new front ends connect immediately.

4. Replace or migrate, one slice at a time

Upgrade the runtime, migrate the database, or rebuild a module, each behind the façade with a switch, each scheduled outside business-critical periods, each with a rehearsed rollback. The university ERP case study shows four such phases scheduled around an academic calendar with no term-time downtime.

5. Embed AI on the stable core

Document extraction, search, copilots and natural-language reporting are added on the modernised foundation, where they finally have data to work with. See Legacy-to-AI Modernization.

Cost, honestly

Modernization programmes at Eazyware run $31,500–105,000+ (₹22.4–72 lakh+) over 8–16 weeks per programme, priced after the assessment; each phase is a fixed price and the business can stop after any of them with a working system. A rewrite of a comparable system typically costs several times that and delivers nothing until the end. The comparison that matters is not total price but value per month and risk per cutover. Ranges are on the pricing page.

Questions to ask any vendor proposing a rewrite

  • What did your assessment find that makes incremental replacement impossible?
  • How will you capture behaviour nobody documented?
  • What runs in parallel, for how long, and who reconciles the two systems?
  • What is the rollback if cutover fails?
  • What value does the business see before month twelve?

The decision, in one paragraph

Assess first, for two weeks, with characterisation tests and a module map. If the assessment shows a small system, a dead platform or a process being replaced, rewrite. Otherwise wrap, replace in slices, schedule around the business, and add AI on the stable core. It is slower to describe and faster to deliver, and the business keeps running throughout. A Sprint Zero is the assessment in its fixed-price form.

A worked comparison on a real system

A fifteen-year-old PHP ERP running admissions, fees and records for a university group. Two vendors proposed a rewrite: two years, a large budget, a single cutover during a growing intake. The assessment mapped fourteen modules, three critical cron jobs nobody remembered, and a fee calculation the finance office corrected by hand every month. Modernization ran in four phases over sixteen weeks: characterisation tests and an API façade, runtime migration for low-risk modules during a term break, fees and records between collection windows, admissions rebuilt with document extraction ahead of the season, and reporting last. No term-time downtime; the group's IT team now ships their own modules on the façade. The case study has the detail.

Signs your system is a modernization candidate

  • It runs on an unsupported runtime or framework version
  • It has no API, so every integration is a manual export
  • Nobody on staff can confidently change it
  • Reports take days because only IT can query it
  • You want AI features and the data is locked inside
  • A previous vendor said "rewrite" without an assessment

What you need to bring

Access to the codebase, database and infrastructure; someone who knows the system's history; the business calendar with the periods when nothing may change; and an IT lead who will own the system afterwards and pair with the team during the work. With those four, the assessment takes two weeks and the first phase can start immediately after. The programme is described on the Legacy-to-AI Modernization page, related work on Enterprise Platform Implementation, and prices on the pricing page.

Team and timeline

A modernization programme is led by an architect who runs the assessment, with two engineers who can read the legacy stack as well as build the new one, an AI engineer for the capabilities added on the stable core, and a delivery lead who manages the calendar with your business owners. Phases run eight to sixteen weeks each and are scheduled around your calendar: no changes during month-end, exams, fee windows or peak season. Your IT team pairs with ours throughout so the system has an owner when we leave.

Before you start: a checklist

  • Grant access to code, database and infrastructure for the assessment
  • Find the person who knows the system's history
  • Write down the periods when nothing may change
  • Nominate the IT lead who will own the system afterwards
  • List the integrations, including the manual exports
  • Decide which AI capabilities you want on the modernized core
  • Agree stop points: the business can pause after any phase

What the API façade makes possible

The façade is the quiet hero of incremental modernization. Once every module is reachable through a clean, versioned interface, the order of replacement becomes a business choice rather than a dependency chain. New front ends, mobile apps and partner integrations connect immediately. Document extraction, search and copilots run on data that used to be locked in the monolith. Reporting moves out of the IT queue into a natural-language layer. And when a module is finally replaced, the switch is a configuration change with a rollback, not a cutover weekend. Clients who complete a modernization programme typically keep building on the façade with their own teams, which is the real measure of whether the work was done properly.

Related reading: Multi-tenant SaaS architecture for the migration from single-tenant, AI readiness assessment for how legacy systems block AI, and Why basic RAG fails in production for what to build once the data is reachable.

The safest way to test everything in this article is the two-week assessment. If it recommends a rewrite, you will have the evidence; if it recommends slices, you will have the plan and the first fixed-price phase. Either way, the decision is made on what the system actually is rather than on what a proposal says it should be.

Frequently asked questions

How long does an assessment take?

▾

Two weeks for most business systems: module map, dependencies, data, integrations and a phased plan with prices.

Can we modernize a PHP 5 or .NET Framework system?

▾

Yes. Runtime upgrades happen slice by slice behind the façade with characterisation tests as the gate.

Will our own team be able to continue afterwards?

▾

That is the point of the façade and the tests; we pair with your IT team so the system has an owner.