Software maintenance and support keep an application usable after its first release. For a business buying that work, the important questions are practical: which systems are covered, who responds when a user journey fails, how changes are approved and what the monthly budget actually buys.
A development handover is the start of this responsibility. Someone still needs to review failures, update dependencies, test releases and keep access and operating instructions current. If the original developer leaves, those tasks do not disappear.
This guide explains how to commission application maintenance, compare support proposals and agree service boundaries. Use the editable project handover checklist to record the evidence a receiving team needs.
What should software maintenance and support include?
Separate the work into activities before choosing a retainer or an hourly allowance. A contract called “maintenance” can cover very different responsibilities. Ask who performs each activity, what triggers it and how the result is reviewed.
| Work area | Example responsibility | Evidence to request |
|---|---|---|
| Incident response | Investigate a failed critical journey, coordinate recovery and communicate progress | Reporting channel, severity rules, service hours and incident record |
| Defect correction | Reproduce an issue, fix the agreed behaviour and check for regressions | Reproduction steps, test result and release reference |
| Dependency and platform upkeep | Review supported versions, assess updates and test compatibility | Inventory, update decision and verified deployment |
| Operations | Review actionable alerts, capacity, backup jobs and restoration procedures | Named owner, runbook and recovery-test record |
| Product changes | Implement an approved enhancement or adapt an integration | Separate scope, estimate, acceptance criteria and approval |
A provider may handle some of these while your team owns the rest. Write that split down. Also identify the cloud provider, third-party vendors and business users whose cooperation is needed during an incident.
Start with an application inventory and takeover review
Before requesting an ongoing price, list the repositories, production environments, domains, databases, integrations, scheduled jobs and mobile store accounts involved. Identify who controls them and which person can approve access. Transfer secrets through an approved secure process rather than inserting passwords into a planning document.
Ask the receiving team to build and deploy the application in an appropriate test environment, run the important journeys and explain how a failed release would be recovered. A document that says “deployment works” is weaker evidence than a repeatable deployment demonstrated by the people taking over.
The review should produce a backlog with three categories: urgent operating risks, regular upkeep and separately estimated improvements. Keep the one-off takeover effort separate from the recurring allowance. An unsupported dependency or undocumented migration can require substantial initial work before a predictable support arrangement is possible.
If you are changing suppliers, agree a period of overlap, the outgoing team's responsibilities and a decision owner for unresolved items. Our SOW and acceptance guide covers deliverables, change approval and handover evidence.
Define service hours and severity before response targets
Service hours determine when a response commitment applies. State the timezone, working days, holidays, reporting channel and route for an emergency outside those hours. Automated monitoring running all night does not establish that an engineer is available all night.
Define severity by the effect on users and the business. For example, you might distinguish a critical journey unavailable to all users, a degraded journey with a workable alternative and a minor issue that does not prevent completion. Those are planning categories to agree, not service commitments offered by EmphaSoft.
For each category, ask who assigns severity, who can change it, when the clock starts and what happens while the team waits for access or a vendor. A screenshot in a private message may not count as a reported incident if nobody is assigned to monitor that channel.
Response time, recovery and resolution are different commitments
A response target describes the first qualified acknowledgement and investigation step. Recovery concerns restoring the affected journey, perhaps with a temporary workaround. Permanent resolution may need a code change, testing, a release or cooperation from another provider. Agree how each stage is recorded and communicated.
Do not read “respond within an hour” as “repair every issue within an hour”. If a proposal includes a resolution target, ask which conditions and dependencies make that commitment feasible, how exceptions are handled and what happens if the target is missed.
Reliability objectives are another decision. Google's SRE guidance on SLOs describes setting reliability targets around what users need and using them to prioritise engineering work. For your application, first choose a journey and a measurable success condition; then agree the target, observation window and responsibility for acting on the result.
For incident coordination, Google's incident response chapter recommends clear roles, communication and a working record. A small team can apply those principles by naming the incident owner, the person performing recovery and the person updating stakeholders.
How to estimate application maintenance cost
Estimate the activities and coverage you need instead of applying a universal percentage to the original build price. Two applications with similar development budgets can have very different support needs: one may have a small internal audience; the other may rely on several external services and serve customers across timezones.
The following is a hypothetical monthly planning example for a small web application. The hours and $60 blended hourly rate are assumptions used to show the calculation. They are not EmphaSoft prices, a market benchmark or a promise of availability.
| Planned work | Assumed hours | Illustrative cost |
|---|---|---|
| Dependency review and routine updates | 8 | $480 |
| Monitoring and operating review | 4 | $240 |
| Recovery exercise and runbook upkeep | 4 | $240 |
| Reserved investigation and defect work | 12 | $720 |
| Coordination and reporting | 4 | $240 |
| Total planned capacity | 32 | $1,920 |
At these assumptions, each line is hours multiplied by $60. If hosting and external services were an additional assumed $280, the planning subtotal would be $2,200. Taxes, one-off takeover work, major upgrades and new features are excluded. If the investigation allocation were exhausted and eight additional hours were approved at the same rate, engineering spend would rise by $480 to $2,400; the subtotal with those services would become $2,680.
Reserved hours are capacity, not unlimited incident insurance. Clarify whether unused time expires, carries forward or is replaced by an availability fee. Continuous on-call coverage requires a staffing arrangement that this 32-hour example does not price. Use the software development cost guide to compare the broader commercial models.
What changes for mobile application maintenance?
A mobile product adds operating-system compatibility, app-store releases, device coverage, push notifications and any native integrations to the support inventory. Agree which versions and devices are covered, who owns the store accounts and who prepares and approves a release.
Include the backend and web administration tools in that discussion. A mobile binary can appear healthy while an API, scheduled task or notification provider prevents the user from completing the journey. Separate the investigation of a store-review delay from a code defect and agree how external waiting time affects commitments.
The published NuStrength case describes a product spanning web, iOS and Android, with continued improvements after its initial releases. It illustrates the surfaces a support inventory needs to cover. The case does not publish a maintenance retainer or an uptime guarantee.
Plan updates, recovery tests and release evidence
Record the support lifecycle of important frameworks and services. For a Django application, for example, consult the project's official supported-version table when planning upgrades. Check it when making the decision rather than copying a version deadline into a long-lived contract.
Assign an owner to assess an update, test the important journeys, approve deployment and check the result. Review the rollback or recovery approach when a release changes stored data. Keep significant upgrades visible as planned work rather than allowing them to consume an undefined monthly allowance.
A successful backup job is not sufficient evidence that the application can be recovered. AWS's recovery-testing guidance recommends testing restoration against recovery-time and recovery-point objectives. Record what was restored, how long it took and what data or dependencies were missing.
The Sourcefit case describes joining an existing team, a project audit and continued releases across a multi-module platform. It is a useful example of the distinction between taking over responsibility and continuing product development, without implying a published support price or SLA.
Questions to settle before signing a support proposal
- Which applications, environments, integrations and user journeys are included?
- Who owns monitoring, triage, customer communication and vendor escalation?
- Which service hours, timezones and severity definitions apply?
- What does response mean, and how are recovery and permanent resolution treated?
- Which updates, fixes, tests and deployments use the allowance?
- Who approves additional spend and what happens when capacity is exhausted?
- Which accounts, runbooks and recovery evidence must be transferred?
- How can another team take over at the end of the engagement?
Use the supplier comparison worksheet to ask each provider the same questions. Keep gaps explicit and resolve the ones that affect the service you depend on before accepting the proposal.
Frequently asked questions
Is maintenance the same as developing new features?
They can share a team, but the scope and approval rules should distinguish them. Routine upkeep and incident work need reserved capacity; a new capability needs an agreed estimate and acceptance criteria.
Can another company maintain an existing application?
A receiving team first needs access, a working build and deployment process, operating context and a clear handover. Commission a bounded review before treating the recurring price as reliable.
Does a support agreement guarantee uptime?
Read its actual commitments. Support hours and response targets do not by themselves guarantee availability. Any availability commitment needs a defined measure, conditions and agreed consequences.
How much should a startup budget each month?
Start with its application inventory, planned upkeep, incident capacity and coverage needs. Price those responsibilities, one-off takeover work and external services separately. The worked example above shows the method; your own inputs determine the result.