SLA and Support Models for B2B SaaS
Enterprise contracts ask for uptime percentages, response times, and escalation paths before legal finishes the MSA. Small teams panic and promise 99.99% without on-call rotation or integration caveats. Support models for B2B software must match what you can actually operate: tenant count, integration dependencies, and whether ERP downtime is your incident or shared responsibility. This guide covers SLA design, severity matrices, support tiers, on-call, customer communication, and how SLAs connect to observability and production readiness. Use it when moving from pilot hypercare to paid support in commercial pricing discussions.
Why SLAs matter in B2B contracts
SLAs translate reliability into remedies: credits, termination rights, or simply trust. Procurement compares vendors on paper before evaluating UX. Overpromising creates legal exposure and burns engineers on pager duty for integration noise outside your control. Underpromising loses deals to competitors who document realistic commitments better. Good SLAs define scope: what is covered, excluded maintenance, customer obligations (timely ticket info, sandbox access), and measurement method.
- Uptime vs response time vs resolution time are different commitments
- Integration and IdP outages need carve-outs or shared status
- Severity definitions must be operational, not vague 'critical'
- Credits caps protect startup cash flow while staying honest
Severity matrix operators and IT understand
Severity 1: production down or data integrity risk for multiple users, no workaround. Severity 2: major feature impaired, workaround exists. Severity 3: minor defect or single user. Severity 4: questions, requests, cosmetic. B2B examples: cannot login via SSO for all users (Sev1); ERP sync stopped for one tenant (Sev2); report export missing column (Sev3); how-to question (Sev4). Align matrix with alerting rules so pages match contractual response clocks.
Uptime, SLOs, and what you measure
Measure availability on user-facing API and core UI, not only internal health checks. Exclude scheduled maintenance windows announced in advance. Monthly uptime of 99.5% allows roughly 3.6 hours downtime; 99.9% allows ~43 minutes. Pick targets your on-call can sustain with safe release practices. Document dependency on cloud provider, DNS, IdP, and customer ERP in SLA appendix. Status page honesty reduces angry tickets during third-party outages.
Support tiers and channels
Standard: email or portal, business hours, Sev1 response in hours. Premium: 24/7 Sev1, named CSM, quarterly reviews, faster integration escalation. Channels: ticket system with required fields (tenant, correlation ID, steps), not unstructured WhatsApp for enterprise accounts. Self-service: knowledge base, integration health in admin UI, release notes. Reduces Sev4 load and improves customer satisfaction cheaper than headcount.
- Business hours defined in customer timezone or contract region
- Escalation path to engineering with severity and impact
- Customer success for adoption, support for defects
- Professional services separate from break-fix support
On-call, hypercare, and handoff from build team
Hypercare after go-live: elevated monitoring, daily check-ins, fixed duration (often two to four weeks), included in implementation fee. Steady-state on-call: rotation, runbooks, handoff from migration programs and launch teams. Contractors can cover on-call if documented in contract terms. Burnout kills quality. If SLA requires 24/7, price premium tier to fund rotation or follow-the-sun partner.
SLA caveats for integrations and customer IT
When ERP or customer network causes failure, SLA credit may not apply if contract defines shared responsibility. Still communicate proactively. Customer-caused issues: expired certificates, firewall changes, rate limit misconfiguration. Support helps diagnose but not unlimited free consulting. Link to integration ownership boundaries documented at sale time.
Credits, remedies, and legal balance
Service credits (percentage of monthly fee capped at 100%) are common instead of direct damages. Define calculation, claim window, and exclusive remedy language with counsel. Track SLA performance monthly per tenant for renewal conversations, not only for credit triggers. Chronic breach may trigger termination rights; avoid settings you cannot meet for integration-heavy products early in lifecycle.
Pricing support into SaaS and services
Bundle standard support in base subscription; charge for premium SLA and implementation. Underpriced support destroys margin when customers treat you as integration consultant. Model cost in lifetime unit economics: tickets per tenant, on-call stipends, tooling for observability and ticketing. Professional services for training and custom reports stay separate SKUs.
Next steps
Draft severity matrix and measure current incident response times from last month. Compare to what sales promises. Align before next enterprise contract. Browse other resources, experience, book a call, or contact with tenant count, integration list, and current on-call setup if you need SLA drafting help.
FAQ
What uptime should a new B2B SaaS promise?
99.5% monthly is a common starting point for small teams with maintenance windows. Increase as on-call matures and architecture redundancy improves. Never promise tier you cannot measure and report.
Should ERP downtime count against our SLA?
Usually excluded or shared if your app is up but integration path fails due to customer ERP maintenance. Document measurement and status communication in the contract appendix.
When to hire dedicated support staff?
When ticket volume exceeds what engineering can handle without roadmap impact, often around dozens of active enterprise tenants or after premium SLA sales. Until then, structured tiers and self-service reduce load.
Do pilots need full SLA?
Pilots use hypercare terms: best-effort response, no credits, shared success goals. Convert to commercial SLA at contract signature with realistic targets based on pilot incident history.