Software maintenance outsourcing: models, SLAs and pricing explained
How to buy application maintenance services: corrective vs evolutive work, SLA structures that work, pricing models, knowledge transfer and exit terms.
Software maintenance outsourcing means paying an external partner to keep an application healthy after delivery: fixing defects, adapting it to changing environments, and evolving it with new features. It is bought under three main pricing models - retainer, ticket-based, and time & materials - governed by a service-level agreement (SLA) that separates response from resolution and tiers both by severity. The contract’s quality shows in three places: the SLA structure, the knowledge-transfer plan, and the exit terms. Here is how to evaluate all three before you sign.
What does software maintenance actually include?
“Maintenance” hides three different kinds of work, and a good contract names them separately because they behave differently:
- Corrective maintenance - fixing defects: production incidents, bugs found by users, data problems. Reactive, urgency-driven, impossible to schedule.
- Adaptive maintenance - keeping the software working as its environment changes: operating system and framework upgrades, security patches, third-party API changes, new regulatory requirements. Predictable in kind, less so in timing.
- Evolutive (perfective) maintenance - improving the product: new features, performance work, usability changes. This is really small-scale development, and it should be scoped and priced like development.
The distinction matters commercially. Corrective work needs an SLA with hard commitments. Adaptive work needs a planning rhythm. Evolutive work needs a backlog and estimates. A contract that lumps all three into one “support” clause guarantees arguments later about what the monthly fee covers.
Some buyers add a fourth category - preventive maintenance (monitoring, dependency updates, technical-debt reduction done before anything breaks). It is the cheapest category per problem avoided and the first one dropped when a contract is priced too thin.
How should the SLA be structured?
The single most important distinction: response time is not resolution time.
- Response time is how quickly a qualified person acknowledges the ticket and starts working. Vendors can commit to this firmly.
- Resolution time is how quickly the problem is fixed. It depends on the problem, so honest vendors commit to it as a target with defined escalation, not an absolute guarantee for every conceivable defect.
A vendor who guarantees a hard resolution time for any severity-1 incident, sight unseen, is either pricing in enormous margin or planning to argue about severity classifications later.
Severity tiers that actually work
| Severity | Typical definition | Response commitment | Resolution approach |
|---|---|---|---|
| S1 - Critical | Production down or unusable, no workaround | Fast, committed (measured in hours or less) | Continuous effort until restored; workaround acceptable first |
| S2 - High | Major function impaired, workaround exists | Committed, same business day | Target with escalation path |
| S3 - Medium | Minor function impaired, limited impact | Committed, in business days | Scheduled into the next work cycle |
| S4 - Low | Cosmetic, questions, nice-to-have | Acknowledged | Backlog, prioritized with the client |
Do not copy the numbers from a template - negotiate ones that match your business hours, your users’ geography and what the fee realistically buys. What must be non-negotiable is the structure: written severity criteria (who classifies, and how disputes are settled), measurement rules (business hours or calendar hours, which timezone), monthly reporting against the targets, and defined remedies - typically service credits - when targets are missed repeatedly.
One more clause worth its weight: who can declare an S1. If only the vendor classifies severity, every outage becomes an S2.
Which pricing model fits which situation?
Three models dominate application maintenance services (AMS), and the honest answer is that most mature contracts combine two of them.
| Model | How it works | Best when | Watch out for |
|---|---|---|---|
| Retainer | Fixed monthly fee for a defined scope or capacity | Steady workload; you want budget predictability and reserved capacity | Scope creep in both directions - unused hours, or “that’s not included” |
| Ticket-based | Price per incident/change, often tiered by complexity | Low, irregular volume; well-bounded systems | Perverse incentives - the vendor earns more when quality is worse |
| Time & materials (T&M) | Pay for hours worked at agreed rates | Evolutive work, unpredictable scope, transition periods | Needs your active prioritization; costs follow your decisions |
A common and sensible structure: a retainer covering corrective and adaptive work with the SLA attached, plus T&M or fixed-price mini-projects for evolutive work. The retainer buys you reserved capacity and response commitments; the T&M stream keeps enhancements from silently consuming the support budget.
Whatever the model, refuse to sign without knowing: what happens to unused retainer capacity (rollover, partial rollover, or lost), how out-of-scope work is approved and billed, and how rates are indexed over a multi-year term. No responsible article can tell you what a retainer “should cost” - it depends on the application’s size, technology, criticality and coverage hours - but a vendor who quotes a monthly figure without examining the codebase is quoting their hope, not your system.
What if the vendor didn’t build the system?
Taking over someone else’s application - after an in-house team dissolved, a vendor relationship failed, or a project rescue - is routine AMS work, but only when the takeover is treated as a project in itself. A serious vendor will insist on a knowledge-transfer (KT) phase:
- Assessment - code review, architecture walkthrough, documentation inventory, dependency and license check. Output: an honest report on the system’s state, including what is undocumented or fragile.
- Access and environments - repositories, CI/CD, staging environments, monitoring, credentials under your control. If a working development environment cannot be rebuilt from what exists, that is finding number one.
- Shadowing and reverse KT - sessions with the outgoing team while they are still available; where they are not, structured code archaeology and the vendor documenting as they learn.
- Supervised operation - a defined period where the new team handles real tickets with training-wheel SLAs before full commitments take effect.
Expect the SLA to phase in. A vendor promising full severity-1 commitments from day one on a codebase they have never run is telling you what you want to hear. And expect the KT phase to be paid work - it is the highest-skill part of the engagement.
What exit terms protect you?
Maintenance contracts create deeper lock-in than development projects, because over the years the vendor accumulates knowledge nobody wrote down. Negotiate the exit while you still have leverage - the same discipline we recommend for IP, NDA and exit clauses in any outsourcing contract:
- Living documentation as a contractual deliverable - runbooks, architecture notes and deployment procedures kept current during the engagement, not reconstructed at the end.
- Your infrastructure - code in repositories you own, tickets in a tracker you can export, credentials in your vault. Leaving should be a permissions change, not a negotiation.
- A defined handover package and transition period - the vendor supports your next team (in-house or another vendor) for a stated period at stated rates.
- Reasonable notice, both directions - long enough to transition, short enough that a failing relationship can end.
The one-line test: could you leave in 90 days without the application degrading? If the honest answer is no, fix the contract, not the relationship.
In short
- Maintenance is three services in one - corrective, adaptive, evolutive - and the contract should name and price them separately.
- A real SLA separates response from resolution, tiers both by written severity criteria, and defines measurement, reporting and remedies.
- Retainer + T&M is the workhorse combination: reserved capacity with commitments for support, flexible capacity for evolution.
- Taking over an unknown system requires a paid knowledge-transfer phase with phased SLAs - distrust anyone who skips it.
- Negotiate exit terms at signing: documentation as a deliverable, your infrastructure, a defined handover. Offshore AMS done this way delivers labor savings that may reach up to 70% versus onshore staffing, without the lock-in.
AD System Asia provides application maintenance and support from Bangkok - for systems we built and systems we inherit - with severity-tiered SLAs, a structured takeover process and exit terms we put in writing because we expect to earn the renewal. Request a quote: tell us what the system is and who maintains it today.
Frequently asked questions
What is software maintenance outsourcing?
Contracting an external partner to keep an application running and evolving after delivery: fixing defects (corrective), adapting to new environments and regulations (adaptive), and building improvements (evolutive). It is sold as application maintenance services (AMS), usually under a service-level agreement with defined response and resolution targets.
What SLA should I ask for in a maintenance contract?
Separate response time from resolution time, and tier both by severity: a production outage needs a fast, committed response; a cosmetic defect does not. Insist on defined severity criteria, measurement rules, reporting, and remedies for missed targets. A single blanket 'we answer within 24 hours' clause protects nobody.
How is application maintenance priced?
Three common models: a monthly retainer for a defined capacity or scope, ticket-based pricing per incident or change, and time & materials for evolutive work. Many contracts combine them - retainer for keep-the-lights-on support plus T&M for enhancements. The right mix depends on how predictable your workload is.
Can a vendor take over maintenance of software they didn't build?
Yes - it is a routine service, but it requires a structured knowledge-transfer phase: code and documentation review, environment access, shadowing the outgoing team where possible, and a period of supervised operation. Expect reduced SLA commitments during transition; a vendor promising full SLAs from day one on an unknown codebase is guessing.
What exit terms should a maintenance contract include?
Reasonable notice periods, an obligation to hand over up-to-date documentation, repositories and credentials, a defined knowledge-transfer package to your next team, and continued service during the transition. The test: could you leave in 90 days without the application degrading? If not, renegotiate before signing.
Is offshore maintenance cheaper than keeping it in-house?
Usually, because maintenance rarely justifies full-time onshore staff: workload is spiky, skills are broad, and idle capacity is expensive. An offshore AMS contract converts that into a predictable fee - labor savings may reach up to 70% versus onshore hiring - while keeping response commitments in writing.