Back to all posts
Process7 min read·October 26, 2026

What Should a Software Support SLA Include?

TB
ThynkBlox Team
Engineering

What should a software maintenance support SLA include?

A software maintenance support service-level agreement (SLA) should define which software is covered, when support is available, how incidents are prioritised, and what response each priority receives. It should separate response time from service restoration and permanent resolution, then name the escalation routes and responsibilities during an outage. Use the software maintenance support SLA checklist below to turn a broad promise of support into commitments both sides can assess.

What belongs in the SLA checklist?

Ask your build partner to complete each item before support begins. Replace phrases such as “prompt attention” with a measurable action and a clear owner.

ItemWhat to agree
Covered softwareApplications, integrations, supported versions and environments
Coverage hoursDays, hours, time zone, holidays and emergency arrangements
Severity levelsBusiness-impact criteria and who can change the classification
Initial responseDeadline for a qualified person to acknowledge and begin assessment
Restoration and resolutionSeparate commitments, targets and dependencies
Progress updatesFrequency, channel and responsible person
EscalationTechnical and management contacts, triggers and backup routes
MaintenanceIncluded work, scheduling, notice and approval requirements
ExclusionsWork outside the agreement and how it is authorised
Customer responsibilitiesAccess, evidence, decisions and designated contacts
MeasurementClock rules, reporting, missed commitments and dispute handling
HandoverDocumentation, access transfer and outstanding work when support ends

Keep the covered-software list specific enough to settle questions about ownership. If an integration fails, the agreement should say who investigates the connection and who contacts the outside provider.

When does the support clock run?

Specify a time zone and a working calendar. “Business hours” leaves too much open when the customer and support team work in different locations.

Distinguish ticket submission from staffed support. If customers can submit requests at any time but engineers work only during agreed hours, state when the response clock begins.

For emergency coverage, name the qualifying severity levels and the route customers must use. Explain whether monitoring alerts open incidents automatically or whether a customer must report the problem.

Also agree:

  • Which holidays affect coverage.
  • Whether lower-severity work pauses outside coverage hours.
  • What happens when an incident continues beyond the staffed period.
  • Whether emergency coverage is included or charged separately.
  • How a failed emergency contact attempt is escalated.

Choose coverage around the hours your customers depend on the product. A critical workflow used outside office hours deserves an explicit support decision.

How should severity levels be defined?

Define severity by business impact. Describe the affected workflow, how broadly it is affected, whether data is at risk, and whether a usable workaround exists.

The following is an illustrative classification, not a required standard:

SeverityExample impactExpected handling
CriticalA core service is unavailable, with no usable workaroundImmediate incident ownership and emergency escalation
HighAn important workflow is seriously impairedPriority investigation and regular updates
MediumA limited problem has a workable alternativeInvestigation within normal support coverage
LowA minor defect has little operational impactReview and scheduling through the support queue

Separate incident severity from feature priority. A requested improvement may be commercially important without being an outage.

Name who assigns the initial severity, what evidence supports it, and how disagreements are handled. Require a reason for any downgrade and keep the customer informed.

How do response, restoration and resolution differ?

Use separate definitions so everyone knows what a commitment actually delivers.

Initial response should mean a qualified person has accepted ownership, reviewed the reported impact and started assessment. Specify that an automated receipt does not satisfy this commitment.

Service restoration should mean the affected workflow is usable again under agreed acceptance criteria. A temporary workaround or rollback may qualify if the agreement allows it.

Permanent resolution should mean the underlying defect has been corrected and the fix has passed the agreed checks.

Consider this illustrative example: a critical incident has a response commitment of 30 minutes, progress updates every hour and a restoration target of four hours. An engineer responds within 20 minutes and restores service through a rollback after three hours. The permanent correction follows later.

Under those example terms, the response and restoration measures are met. Permanent resolution remains open.

For each measure, state whether it is a firm commitment or a target. If permanent resolution has no fixed deadline, require an owner, a next review point and a communicated plan.

What should happen during an outage?

Name an incident lead who coordinates technical work and a customer contact who can confirm business impact and approve consequential decisions. Provide backups for both.

Agree who can authorise a rollback, disable an affected feature or accept temporary reduced functionality. Define which actions the support team may take without waiting for approval.

Each progress update should cover:

  • The current customer impact.
  • What has been confirmed and what remains uncertain.
  • Actions underway and any decision needed.
  • The time of the next update.

Include a communication route that remains available if the application itself is down.

Define restoration checks and closure separately. The customer should know when service is usable, whether a workaround remains, and what follow-up work is still outstanding. Agree when an incident review is required and who owns the resulting actions.

How should escalation work?

Give escalation a trigger, a destination and a fallback. “Contact your account manager” is incomplete if that person is unavailable.

Set technical escalation triggers for stalled diagnosis, missing expertise or a threatened restoration target. Set management escalation triggers for missed commitments, disputed severity or decisions requiring additional authority.

Keep contacts in a maintained directory referenced by the agreement. Require both sides to update it when responsibilities change.

Escalation should preserve incident ownership. Specify who continues coordinating the response while specialists or managers join.

What maintenance work is included?

List the recurring work covered by the fee. Consider defect correction, dependency updates, security patches, backup checks, monitoring review and compatibility testing. For each item, say whether the provider performs it, advises on it or leaves it to the customer.

Separate routine maintenance from new features and substantial redesign. A maintenance agreement should explain how proposed work is assessed and approved.

For planned maintenance, agree notice, permitted windows, expected disruption, validation and rollback responsibilities. Define how urgent maintenance can proceed when the normal notice period is unsuitable.

If planned downtime is excluded from an availability measure, define the exclusion precisely. Require approval for extensions beyond the agreed window.

Keep longer-term improvements visible too. The guide to technical debt for founders offers a related starting point for discussing work beyond immediate incident repair.

Which exclusions and customer responsibilities need attention?

Consider unsupported versions, customer modifications, outside-service failures and work beyond the covered software. For every exclusion, explain what assistance remains available and how additional charges are approved.

For an outside-service failure, separate responsibility for repairing that service from responsibility for investigating its effect on your application.

Customer duties should include providing a reproducible description where possible, identifying affected workflows, granting appropriate access and making time-sensitive decisions. Agree a safe way to share evidence without sending unnecessary personal information or credentials.

If waiting for customer input pauses a clock, require a recorded request, a reason the input blocks progress and clear restart conditions. Avoid allowing an unrelated question to pause the entire incident.

Ask for a usable support handover. Documentation best practices in software projects is a useful companion when defining what the support team and customer need to maintain.

What common mistakes should you avoid?

Avoid a single “fix time” covering acknowledgement, temporary recovery and permanent correction. Give each stage its own definition.

Do not accept unlimited maintenance language paired with undefined exclusions. List what is included and how extra work receives approval.

Do not treat an availability promise as a substitute for incident handling. Define both if both matter to your product.

Finally, agree how performance is reported and what happens after a missed commitment. Specify any applicable credit process, corrective-action requirement and route for repeated failures. A report should show unresolved work as well as completed incidents.

What questions do founders ask about support SLAs?

Does continuous support mean every request receives immediate attention?

Define continuous coverage and severity handling separately. You may agree emergency handling at all hours while routine defects remain within normal support hours.

Should every defect have a guaranteed resolution deadline?

Choose commitments the provider can explain and support. Where diagnosis or outside dependencies make a fixed deadline unsuitable, require response commitments, progress updates and an explicit resolution plan.

Is support the same as maintenance?

Use separate scope lists. Treat support as handling reported problems and maintenance as agreed work to keep the software supported and usable. State where the two overlap.

What should we review before signing?

Walk through an outage scenario together. Check who responds, when clocks run, how decisions are approved and what happens if the first recovery attempt fails. Revise any clause that leaves those answers unclear.

Outside services change too; see what to do when a third-party API is being retired.

Ready to build?

Let's turn these ideas into your next product.

Start your project