App maintenance costs: planning for each year
An app continues to need attention after its first release. Even when the product scope stays unchanged, operating systems, dependencies and connected services can change. A useful annual maintenance plan describes the system, the work and the people responsible. It does not assume that a universal percentage of the original development bill describes every product.
Updated: · tealtap

Start with an operating inventory
List everything the product needs to complete its main journey: mobile clients, backend, database, administration tools, domains, identity services and external APIs. Record who owns each account, who pays its bills and who receives operational notices. Identify which failures would stop useful work and which would only remove a convenience. This makes the operating model visible before anyone estimates maintenance.
Separate recurring supplier charges from planned engineering work and incident handling. Paying for hosting does not necessarily include looking after the application. A crash reporting subscription does not mean someone investigates its alerts. For each category, describe the assumptions about usage, service availability and business impact. Confirm any actual tealtap maintenance offering during project discussions.
Budget for compatibility work
Android's official compatibility guidance recommends reviewing platform changes and testing affected behaviour. Apply that principle to the supported platforms in your product: inspect changes, identify exposed features and check real journeys. Login, notifications, background work and device connections may need particular attention. Record supported devices and system versions rather than promising unchanged behaviour on every future phone.
Test upgrades from an existing installation as well as clean installations. Stored data, permission decisions and expired sessions can reveal problems that a fresh device misses. Keep suitable test data and devices available. If you narrow platform support, assess who is affected and prepare communication. Compatibility maintenance is a continuing decision process, not a prediction that software will never require adjustment.
Keep store submissions possible
Assign someone to review current Apple and Google requirements and account notices. Submission tooling, privacy declarations and platform targeting requirements can change. Build the applicable work into the plan after checking official guidance; an old checklist is not a reliable source of permanent deadlines. A new SDK may also require another look at the information published about data use.
Maintain the ability to produce and distribute an update. That includes source access, documented build steps, signing arrangements and suitable account permissions. A completed fix is of limited help if nobody can submit it. Keep release records and confirm who approves public materials. Submission and availability are separate milestones because store review is an external process.
Treat dependency updates as engineering work
A library update can change behaviour, configuration or supported tooling. Keep an inventory of important dependencies and decide who reviews security notices and compatibility concerns. Distinguish urgent remediation from planned modernisation. Evaluate the affected user journey and backend interaction, then record what changed and which release contains it. Updating a version number alone is not evidence that the product still works.
Deferred work should have a named owner and an explanation. If an obsolete toolchain eventually prevents releases, restoring the build may become a separate project. That is a planning risk, not a basis for inventing a standard maintenance rate. Ask for visible technical debt and a proposed order of work so business decisions can account for it.
Understand hosting and supplier usage
Hosting demand depends on the type of operations, stored data, traffic patterns and recovery expectations. User counts alone may conceal the difference between small text requests and large file processing. Describe expected activity and identify usage based charges such as messaging or external API calls. Agree who receives unusual usage alerts and who can approve a service plan change.
Backups need a recovery procedure and an isolated restoration exercise. Identify who can access them and how the chosen recovery point affects subsequent records. Keep ownership of domains, certificates and billing contacts visible too. Administrative failures can interrupt a product just as effectively as code defects, so include them in the operating inventory.
Make monitoring lead to action
Watch signals that relate to useful operation: completion of core requests, failed jobs, queues and backup outcomes. Give alerts a recipient and a practical response path. An alert stream with no follow up creates little operational value. Review the business effect with a contact who can approve a temporary workaround or decide whether normal operation should pause.
Keep logs focused on diagnosis and avoid collecting unnecessary personal data or credentials. A request reference, timestamp and technical state may provide enough context. User reports should ask for reproducible steps and redacted screenshots rather than passwords. Agree support availability, severity assessment and acknowledgement expectations explicitly. A response commitment and a final repair commitment describe different obligations.
Define maintenance boundaries and review the plan
Repairing existing behaviour, adapting to a platform change and adding a new business workflow can be different work categories. Describe examples and the approval process in the agreement. When a case is ambiguous, assess its cause and scope before assigning it to a budget. This prevents the word maintenance from silently becoming an unlimited product development promise.
Review completed work, unresolved risks and the next decisions together. If the product is being retired, plan communication, data handling, account access and subscription closure. The purpose of annual maintenance planning is to preserve the agreed service while making changes and responsibilities visible. A clear inventory and a working release process provide a stronger basis for estimates than an unsupported industry average.
Frequently asked questions
How much should annual maintenance cost?
Estimate it from the system inventory, usage, condition and support needs. Request separate assumptions for supplier charges, planned engineering and incident work instead of relying on a universal percentage.
Does an unchanged app still need attention?
Yes. Platform behaviour, dependencies and submission requirements can change without a new business feature. The appropriate review and update work depends on the product.
Does hosting include operational support?
Only if the agreement says so. Ask specifically about monitoring, backup recovery, application incidents and account administration.
What helps with a maintenance assessment?
Provide a system inventory, known issues, supported platforms and release documentation. Use test data and redacted materials in the initial enquiry.
Discuss your project
Describe the app, its core journey, connected services and current support arrangements. Include known issues and planned changes. This gives a maintenance discussion a useful starting point; the available service scope and responsibilities need to be confirmed for the project.


