Legacy System Modernization: Replace, Refactor or Migrate.
Legacy system modernization means moving business software that still works, but is hard to change, hard to secure or known by one person, onto something your team can safely build on. We start with an assessment, pick the path with you, move the data and run old and new side by side until the cutover is boring.
Three paths, and how to tell which is yours
There are three honest options. You can replace the system with an existing product, refactor the code you have, or rewrite it in small pieces. The right one depends on how unusual your process is and how much of the old logic is worth keeping. Devvista is a software development and growth agency; the assessment below is how we decide, and sometimes the answer is to leave a system alone.
| Path | Fits when | Main risk | Price lean |
|---|---|---|---|
| Replace with a product | Your process is ordinary and a product covers most of it, for example moving accounting to Odoo or a CRM to an existing platform | Fitting the process to the product, and data mapping | Lower build cost, plus licences |
| Refactor | The logic is sound and the platform is supported, but the code is tangled or untested | Hidden dependencies surface while you change things | Hourly often fits, scope shifts as you learn |
| Rewrite in pieces | The logic is unique to you and the platform or data model is the problem | Running both systems for longer | Higher total, spread over stages |
The price column is our judgement from scoping work. Your estimate comes from the assessment.
Not sure which path? Describe the system and we will tell you what we would check first.
Get an EstimateSigns the system has become a risk
- The platform is out of support. Microsoft's lifecycle page lists extended support for Access 2019 ending in October 2025. As of November 2026, the PHP project lists 8.2 as the oldest branch with security fixes, until 31 December 2026; the list changes each year.
- One person understands it. If that person leaves, so does the knowledge of how invoices, stock or schedules are calculated.
- Every change breaks something else. A one-field request takes weeks because nobody can say what depends on it.
- It cannot be reached from outside. Remote staff, customers or a mobile app need a way in the system was never built to give.
- Reports are rebuilt by hand. Someone exports to Excel each month to answer a question the system should answer.
- It does not connect. Orders, payments or CRM records are retyped between the old system and everything else.
Support dates are from Microsoft Lifecycle and php.net supported versions, opened for this page. An out-of-support product is a risk, not a failure: check what you run before deciding.
What we inventory before recommending anything
The assessment is a short, separate piece of work, so you can take its output to another vendor if you prefer. It ends with a written document, not a sales call.
| Area | Question we answer | What you receive |
|---|---|---|
| Business rules | What does the system calculate, and which of those rules still apply? | A list of rules in plain language, marked keep, change or drop |
| Data | Where is it, how clean is it, what must be kept for legal or audit reasons? | A data map with problem areas flagged |
| Integrations | What feeds it and what reads from it, including spreadsheets and manual exports? | A diagram of inbound and outbound connections |
| Users and access | Who uses which screens, and who must keep access during the move? | A role list and a migration order by team |
| Platform and security | Is it supported, patched and backed up, and who can restore it? | A risk list ranked by how bad and how likely |
| Options | What would replace, refactor and rewrite cost and risk for this system? | A recommendation, a staged plan and a proposal |
What we are usually asked to replace
These are common sources, not a promise that we have seen yours. If it is not listed, describe it and we will say whether we can read it.
Moving off a database one team built
A database that started as one person's tool often ends up running payroll, orders or inventory. We read its tables, queries and forms, list the rules hidden in them, and rebuild the screens people really use as a web application with a proper database behind it. Our guide to replacing a Microsoft Access database compares the options and cost in detail.
If the real need is accounting or inventory, a product may beat a rebuild. Our guides on QuickBooks Desktop to Odoo and Dynamics GP end of life cover those cases, and our Odoo implementation page explains how a move runs.
Applications that still run but cannot change
For older .NET and PHP applications the first question is whether the code can be upgraded in place. If it can, an upgrade with tests added is the cheapest path. If the framework, the data model and the hosting are all aging, rewriting the parts that change most is safer than a single big rewrite. Our .NET and PHP developers work on both, and our custom software development API development: connect an old system to new ones step by step page covers the build side.
Moving the data, running both systems, then cutting over
The strangler fig pattern, described by Martin Fowler, grows the new system beside the old one and shifts functions across until the old one can be retired. It reduces the risk of each change, and it needs temporary scaffolding so both systems can run together. We use it whenever the system is too important to switch off for a weekend.
| Stage | What happens | How you know it is done |
|---|---|---|
| Map and clean | Fields are matched to the new model; duplicates and broken records are listed for your decision | You sign off the mapping sheet |
| Trial loads | The data is loaded into staging several times, each load checked against the old system | Totals and sample records match |
| Parallel running | Users work in the new system for one function while the old one stays available for the rest | A full cycle, such as a month-end, closes on the new system with matching numbers |
| Cutover | Final load, old system made read-only, users switched, rollback plan ready | You confirm and the old system is kept read-only for an agreed period |
| Retire | Old system archived and access removed once reports no longer need it | An archive you can open exists, and the old licences or servers end |
Source for the pattern: Martin Fowler, Strangler Fig Application.
Six steps from first look to retired system
Each step ends with something you can read, so you can stop or change course at any of them.
Assess
Read the code, data and screens; interview the people who use it; write the inventory above.
Choose the path
Replace, refactor or rewrite in pieces, with a staged plan and a proposal you can compare.
Build the first slice
One function goes live on staging first, shown to real users within the first sprints.
Migrate data
Trial loads, comparison checks and your sign-off before any real switch.
Run in parallel
Old and new run together for a full business cycle, with a daily list of differences.
Cut over and retire
Final load, rollback plan, training, a read-only archive and an optional support retainer.
What drives the cost of modernization
We quote to scope after a short call, in writing, and we ask what budget range you are working with so the proposal fits it. The total depends mainly on how complex the work is, how many systems it touches and the level of developer involved. The assessment can be priced on its own, then the build is quoted from its findings.
| Driver | Pushes the price up when | Pushes it down when |
|---|---|---|
| Data volume and quality | Years of duplicates, open text fields, missing keys | Clean tables with clear keys |
| Hidden rules | Logic lives in macros, queries and one person's habits | Rules are documented or simple |
| Integrations | Many systems read or write the data | The system stands alone |
| Users and roles | Many roles, approvals and audit needs | A small team with one role |
| Staging | Parallel running over several cycles | One function moves at a time |
The hours and the rate are assumptions for the example. Yours comes from the assessment.
When to leave the old system alone
- It does one job, with one user, and backups work. Document it and move on.
- The process itself is wrong. New software will repeat a broken process faster. Fix the process first.
- A product would do the job for less. If an existing tool covers what you do and you can leave it with your data, adopt it and handle any gap with a small add-on.
- Nobody can decide what the new system should do. The assessment can help, but the decision stays with you.
Related projects
We have no published case study of a legacy database migration, and we will not present one. These two projects are the nearest work we can show, and neither is a legacy database migration. At DAligners, manual clinic onboarding, case intake and status updates became two custom portals. At TradingSmarter, manually generated and sent licence keys became an automated licensing platform, as each case study describes.
A Custom Dental Aligner Platform for DAligners Member Portal and Automated MT5 Licensing for TradingSmarterWhere the facts on this page come from
- Microsoft Lifecycle, Access 2019, for the extended support end date.
- PHP, Supported Versions, for the oldest supported branch and its security support date.
- Martin Fowler, Strangler Fig Application, for the incremental replacement pattern.
Straight
Answers.
Questions we hear before a modernization starts.
Tell Us About Your Old System.
We will review it and come back with an honest recommendation and a clear estimate.
- 30-minute consultation
- NDA signed on request
- Reply within 24 hours
- Book a call30 minutes, no pitch. Tell us what you want to build.
- Get a straight answerHonest feedback on fit, scope and timeline.
- Decide with no pressureMove ahead only if it feels right. No commitment required.