Quick Answer: What questions should I ask a legacy software modernization company before hiring?
Ask how the vendor will surface undocumented business rules, how they will cut over from old to new, and how they will prove the new system behaves the same as the old one. Those three answers tell you more than any sales deck, because they reveal whether the company plans a risky big-bang rewrite or a controlled, incremental migration.
Vendors differ in how much discovery happens before code is written. Some model the existing system first; others start refactoring immediately and learn as they go. Both can work, but only one usually comes with a clear audit trail.
Before signing, compare discovery method, cutover strategy, testing approach, data migration and reconciliation plan, security controls, IP ownership, knowledge transfer terms, and measurable references. This guide walks through each one.
Why These Questions Matter More Than the Proposal

Legacy modernization is one of the few IT projects where the demo looks flawless and the delivery still fails. The reason is simple: nobody fully understands the old system until someone tries to change it. Business rules live in code comments, stored procedures, spreadsheet workarounds, and the memory of one person who retired two years ago.
That is why the questions below focus on method rather than technology. A vendor’s stack choice matters, but their approach to discovery, testing, and cutover determines whether the project survives contact with reality. Readers weighing broader platform decisions may also find value in this look at what application modernization really means for legacy systems, which frames the difference between a rewrite and a genuine modernization.
There is also a practical trap worth naming early. Many buyers ask excellent questions during evaluation and then accept vague answers because the vendor is likable and the timeline is tight. Push for specifics. “We use agile best practices” is not an answer. “We extract rules into a decision table, review them with your subject matter experts, and freeze them before sprint one” is.
How Will You Find the Business Rules Nobody Documented?
This is the first question because it is the one most often skipped. A legacy system encodes decades of decisions: pricing exceptions, tax logic, eligibility checks, approval thresholds. Much of it was never written down.
Ask whether the vendor uses a structured discovery method. One established approach is architecture-driven modernization, a family of standards that includes the Knowledge Discovery Metamodel, adopted in the mid-2000s to model existing systems before changing them. The point is not the acronym. The point is whether the vendor has a repeatable way to extract rules from code and data and re-specify them before rewriting.
Practical tip: ask them to walk through one real rule from your system. If they can trace it from a screen to a database column to a code branch, they know what they are doing. If they talk only about frameworks, they are guessing.
Common mistake: letting the vendor interview only managers. The people who know the exceptions are usually the ones who have run the system daily for fifteen years. Insist that those users are in the discovery sessions.
Will You Use a Strangler Pattern or a Big-Bang Rewrite?
Ask directly: “Do you replace the whole system at once, or route traffic piece by piece?” The incremental approach, often called the strangler fig pattern, was described by Martin Fowler in 2004. A facade sits in front of the legacy system, and calls gradually move to new services until the old system can be retired.
The difference in risk is enormous.
| Approach | Typical risk | What buyers should watch |
|---|---|---|
| Big-bang rewrite | High; all failures arrive at once | No rollback path, long freeze on new features |
| Incremental strangler | Lower; failures stay contained | Requires strong routing and monitoring skills |
| Hybrid (phased modules) | Medium | Depends on clean module boundaries |
Neither is automatically right. A small internal tool with ten users may be fine to replace in one go. A billing engine that touches revenue cannot be. What matters is that the vendor explains why their choice fits your system and what happens if a phase slips.
Ask a follow-up: “What is your rollback plan for the first cutover?” A vendor without one has not thought seriously about failure.
How Will You Prove the New System Behaves Like the Old One?
Functionality that looks identical can behave differently in edge cases, and edge cases are where legacy systems earn their keep. The standard safety net is characterization testing, sometimes called golden master testing: capture the current system’s real outputs for thousands of inputs, then compare the new system against those captures. The technique is closely associated with Michael Feathers’ 2004 book on working effectively with legacy code.
Ask the vendor how many characterization tests they typically build, who approves the expected results, and how they handle cases where the old behavior is actually a bug. That last question is revealing. Good vendors have a process for logging “preserve or fix” decisions with business owners. Weak vendors quietly fix things and create surprises later.
Watch for teams that plan to test only against written requirements. If the requirements were incomplete enough to cause the modernization project, they are incomplete enough to miss defects.
What Is Your Data Migration and Reconciliation Plan?
Code is half the job. Data is the other half, and it is usually messier. Ask how the vendor will map old schemas to new ones, how they will handle duplicates and orphaned records, and how they will prove nothing was lost.
A credible answer includes row counts, checksums, and business-level reconciliation: total balances, open orders, active contracts. It also includes a rehearsal. Ask whether they will run the migration at least twice in a test environment before touching production, and how long the final cutover window will be.
Public-sector guidance on working with legacy systems treats migration as a distinct discipline with its own planning, which is a useful benchmark even for private companies. If a vendor treats data migration as a footnote in the code plan, expect the schedule to break there.
How Do You Handle Security, Compliance, and Technical Debt?
Legacy systems often carry regulatory obligations: payment rules, health data protections, financial reporting. Ask how the vendor preserves those controls during the transition, when old and new systems run side by side. A common reference for structuring security requirements is the NIST Cybersecurity Framework, updated to version 2.0 in February 2024. Ask whether their plan maps to a recognized framework or to your own control set.
Then ask about technical debt. Research from the Software Engineering Institute, including the 2016 book Managing Technical Debt, treats debt as something to measure and manage; complaints alone do not manage it. Ask the vendor how they will track shortcuts taken during migration and when those will be repaid. A vendor with no answer will hand you a new system with the same old problems.
Readers comparing security expectations across vendors may find this checklist on AI chatbot security questions to ask before buying useful for framing control questions, even though the subject differs.
Who Owns the Code, the Tooling, and the Knowledge?
Intellectual property disputes are common and avoidable. Ask who owns the modernized code, any accelerators or frameworks the vendor brings, and the documentation produced along the way. Plain-language guidance on intellectual property in software contracts exists precisely because these terms are so often left vague.
Then ask about exit. If the vendor holds key knowledge, what happens when the contract ends? A serious partner offers documentation standards, training for your staff, and a transition plan. A vendor that resists knowledge transfer is building dependency, not modernization.
For teams running a formal procurement, this managed detection and response checklist of questions to ask before buying shows how the same discipline applies to security vendors: named deliverables, measurable outcomes, clear exit terms.
What Evidence Can You Show, and Can I Check It?
Ask for references from projects similar in size, technology, and industry, then actually call them. Ask those references three things: did the vendor hit the cutover date, how did they behave when something broke, and would you hire them again.
Case studies with measurable outcomes beat logos. “Reduced batch processing from nine hours to forty minutes” is a claim you can verify. “Delivered digital transformation” is not.
A practical tip from procurement experience: ask for the name of the person who will actually lead delivery, not the person in the pitch. Then ask whether that person is available for the start date. Staffing swaps after signing are one of the most common sources of disappointment.
Buyers planning a longer program may also want a structured view of sequencing, which this application modernization roadmap from legacy to practical delivery lays out in phases.
What Does “Done” Mean, and How Will We Measure It?
Ask the vendor to define completion in writing. Good definitions include behavioral equivalence for defined scenarios, performance targets, migration reconciliation results, documentation handover, and a support period after go-live.
Then ask how progress will be reported. Weekly demos of working software beat status slides. If the vendor cannot show something running every couple of weeks, the project is drifting.
Finally, ask what happens if the estimate is wrong. Fixed-price contracts sound safe but often hide scope cuts. Time-and-materials contracts are flexible but need tight governance. Either can work if the vendor explains the trade-off honestly.
A Practical Hiring Sequence That Works
- Send the question list before the first call so answers are prepared, not improvised.
- Ask each vendor to walk through one real rule and one real data flow from your system.
- Request written answers to the discovery, cutover, testing, and data migration questions.
- Call two references and ask about failures, not successes.
- Confirm the delivery lead and the exit terms before signing anything.
The pattern across all of this is simple. Vendors who have done the work enjoy these questions. Vendors who have not will try to change the subject to technology. Pay attention to which one you are talking to.
Conclusion
Hiring a legacy modernization company is really an exercise in risk assessment. The right questions do not test whether a vendor knows modern frameworks. They test whether the vendor understands that the old system is the specification, that data is as important as code, and that a controlled cutover beats a dramatic launch every time.
Ask about discovery, cutover strategy, behavioral testing, data reconciliation, security controls, technical debt, IP ownership, knowledge transfer, and verifiable references. Then listen for specifics. The vendor who answers with methods, examples, and named people is the one worth hiring. The one who answers with adjectives is not.
FAQ
How long should a legacy modernization project take?
There is no universal timeline, but scope drives duration far more than technology does. A single well-bounded module can move in a few months, while a core system with heavy integration typically runs in phases over a year or more. Be wary of any estimate produced before discovery, because the undocumented rules are usually where schedules break. Ask vendors to commit to phase dates rather than one final date.
Should I choose a fixed-price or time-and-materials contract?
Both are legitimate, and the choice depends on how well the scope is understood. Fixed-price suits work with clear boundaries and stable requirements, but it often invites scope cuts when surprises appear. Time-and-materials suits discovery-heavy work but demands strong governance and regular demos. A phased hybrid, with fixed pricing per phase after discovery, frequently gives the best balance.
What if the vendor wants to rewrite everything from scratch?
A full rewrite can be the right call for small systems or genuinely obsolete platforms. For large systems, ask what happens to the business rules embedded in the old code and how they will be captured. Ask too about parallel running: how long both systems will operate side by side and who pays for that period. If the answer is vague, the risk is yours.
How do I evaluate a vendor’s references properly?
Ask for references from projects comparable in size and industry, then speak to operational staff in addition to the executive sponsor. Ask what broke, how quickly it was fixed, and whether the vendor’s team stayed stable. A reference who hesitates on the failure question is telling you something useful. Two honest calls beat ten polished case studies.
What documentation should I expect at handover?
Expect architecture diagrams, data dictionaries, rule specifications, environment setup instructions, test suites including characterization tests, runbooks for deployment and rollback, and a support transition plan. Ask to see samples during evaluation. Vendors who produce good documentation are usually happy to show it, and the samples reveal more about their standards than any promise.
Can we modernize in-house instead of hiring a vendor?
In-house teams can succeed, especially when they already own the domain knowledge and can dedicate senior engineers. The trade-off is capacity: modernization competes with daily feature work, and it rarely wins. Many organizations use a vendor for discovery and the first migration phase, then bring delivery in-house once patterns and documentation exist. That blended model often controls cost and dependency at the same time.


